You are on page 1of 108

Tim Sweeney Archive - News
March 18, 2007

1 Prologue 10

2 1998 News 11
2.1 Oct 13, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Unreal Cheat-Finding Challenge . . . . . . . . . . . 11
2.1.2 UnrealScript Notes . . . . . . . . . . . . . . . . . . . 12
2.2 Oct 14, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Network Objects Follow-Up . . . . . . . . . . . . . . 14
2.2.2 AWT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.3 Cheat Challenge Update . . . . . . . . . . . . . . . . 15
2.2.4 Another Public Master Server . . . . . . . . . . . . . 15
2.3 Oct 22, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 Linux Unreal . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.2 Hired Guns . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Oct 24, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.1 Almost there... . . . . . . . . . . . . . . . . . . . . . . 16
2.4.2 More discussion on CORBA . . . . . . . . . . . . . . 17

Tim Sweeney Archive 2 News

2.5 Oct 25, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.5.1 Geting closer... . . . . . . . . . . . . . . . . . . . . . . 18
2.6 Oct 26, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.6.1 Direct3D Drivers . . . . . . . . . . . . . . . . . . . . 18
2.6.2 Known Issues . . . . . . . . . . . . . . . . . . . . . . 19
2.6.3 Unreal 218 Gaining Support . . . . . . . . . . . . . . 19
2.6.4 PlanetUnreal performance survey . . . . . . . . . . 20
2.6.5 More about 3D API’s . . . . . . . . . . . . . . . . . . 20
2.6.6 Unreal 218 Patch is ready! . . . . . . . . . . . . . . . 22
2.7 Oct 28, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.7.1 Latest Happenings . . . . . . . . . . . . . . . . . . . 22
2.7.2 Status of Direct3d - All Drivers . . . . . . . . . . . . . 23
2.7.3 Riva TNT . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7.4 Riva 128 . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7.5 Intel i740 . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7.6 ATI Rage Pro . . . . . . . . . . . . . . . . . . . . . . . 23
2.7.7 3dfx Voodoo2 . . . . . . . . . . . . . . . . . . . . . . 24
2.7.8 S3 Savage 3D . . . . . . . . . . . . . . . . . . . . . . . 24
2.7.9 Matrox G200 . . . . . . . . . . . . . . . . . . . . . . . 24
2.7.10 Rendition . . . . . . . . . . . . . . . . . . . . . . . . 24
2.7.11 Permedia . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.7.12 Drivers we consider playable, in best-to-worst order 25
2.8 Oct 30, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.8.1 Unreal 218 patch information (RTFM!) . . . . . . . . 25

Tim Sweeney Archive 3 News

2.8.2 Resources for Server Administrators . . . . . . . . . 26

2.8.3 Unreal Multiplayer Ramp-Up . . . . . . . . . . . . . 26
2.8.4 Other Servers . . . . . . . . . . . . . . . . . . . . . . 27
2.9 Nov 01, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.9.1 Unreal Grows... . . . . . . . . . . . . . . . . . . . . . 27
2.10 Nov 02, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.10.1 Unreal 219! . . . . . . . . . . . . . . . . . . . . . . . . 28
2.10.2 The testing is going well . . . . . . . . . . . . . . . . 28
2.10.3 Testing 219 . . . . . . . . . . . . . . . . . . . . . . . . 29
2.11 Nov 10, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.11.1 Latest News . . . . . . . . . . . . . . . . . . . . . . . 29
2.11.2 The Roadmap . . . . . . . . . . . . . . . . . . . . . . 30
2.11.3 Server Administration . . . . . . . . . . . . . . . . . 31
2.12 Nov 19, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.12.1 Unreal 220 Internet Improvements . . . . . . . . . . 31
2.12.2 Unreal Editor Progress . . . . . . . . . . . . . . . . . 32
2.12.3 Web Site Notes . . . . . . . . . . . . . . . . . . . . . . 33
2.12.4 Unreal mod notes . . . . . . . . . . . . . . . . . . . . 34
2.13 Dec 10, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.13.1 No News . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.14 Dec 11, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.15 Dec 12, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.16 Dec 14, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.16.1 Hard at work . . . . . . . . . . . . . . . . . . . . . . . 36

Tim Sweeney Archive 4 News

2.16.2 That’s . . . . . . . . . . . . . . . . . . . 37

2.17 Dec 15, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.17.1 Unreal Networking Code: Status . . . . . . . . . . . 38
2.18 Dec 18, 1998 . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.18.1 Unreal Tournament notes for mod authors . . . . . 41

3 1999 News 43
3.1 Jan 01, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.1.1 Happy New Year . . . . . . . . . . . . . . . . . . . . . 43
3.1.2 Other Updates . . . . . . . . . . . . . . . . . . . . . . 49
3.2 Jan 28, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.1 Unreal 221 Development Update . . . . . . . . . . . 49
3.2.2 Unreal Tournament . . . . . . . . . . . . . . . . . . . 50
3.3 Feb 10, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3.1 Latest News . . . . . . . . . . . . . . . . . . . . . . . 51
3.3.2 What we’ve been programming . . . . . . . . . . . . 51
3.3.3 Linux Notes . . . . . . . . . . . . . . . . . . . . . . . 52
3.4 Mar 19, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4.1 Latest News . . . . . . . . . . . . . . . . . . . . . . . 53
3.4.2 Level-Of-Detail Texture Mapping . . . . . . . . . . . 54
3.4.3 Future of Programming Languages . . . . . . . . . . 54
3.4.4 Visible Surface Determination . . . . . . . . . . . . . 57
3.4.5 Commentary on Cool Stuff . . . . . . . . . . . . . . . 58
3.5 Apr 14, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

Tim Sweeney Archive 5 News

3.5.1 3dfx Voodoo3 Totally Rocks! . . . . . . . . . . . . . . 59

3.6 Apr 15, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.6.1 Important Unreal 224 Note To Mod Authors . . . . . 60
3.6.2 Maps, Textures, Sounds, Music are backwards com-
patible . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6.3 Latest News . . . . . . . . . . . . . . . . . . . . . . . 62
3.6.4 Cool Stuff . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.7 Apr 19, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.7.1 Getting closer... . . . . . . . . . . . . . . . . . . . . . 64
3.7.2 Network Cross-Compatibility . . . . . . . . . . . . . 64
3.8 May 01, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.8.1 224 Progress Continued . . . . . . . . . . . . . . . . 65
3.9 May 02, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.9.1 STAT NET: Diagnosing your connection to the server 66
3.9.2 Server Alert . . . . . . . . . . . . . . . . . . . . . . . 68
3.9.3 Unreal 224v C++ headers released! . . . . . . . . . . 69
3.9.4 Unreal 224v Released to the public! . . . . . . . . . . 70
3.10 May 03, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.10.1 Known Bugs . . . . . . . . . . . . . . . . . . . . . . . 71
3.10.2 Future Patch . . . . . . . . . . . . . . . . . . . . . . . 71
3.10.3 Check out the Unreal news sites . . . . . . . . . . . . 72
3.10.4 OpenGL UnrealEd . . . . . . . . . . . . . . . . . . . . 72
3.11 May 05, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.11.1 Maps that broke since 220 . . . . . . . . . . . . . . . 73

Tim Sweeney Archive 6 News

3.11.2 Quick Notes . . . . . . . . . . . . . . . . . . . . . . . 73

3.12 May 08, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.12.1 Latest progress . . . . . . . . . . . . . . . . . . . . . 74
3.13 May 19, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.13.1 Unreal 225 musings . . . . . . . . . . . . . . . . . . . 75
3.13.2 ”Make Something Unreal” Contest . . . . . . . . . . 76
3.13.3 Things That Are Cool . . . . . . . . . . . . . . . . . . 76
3.14 May 29, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.14.1 For server administrators only: Unreal 225e Patch . 77
3.14.2 On engine licensing . . . . . . . . . . . . . . . . . . . 77
3.14.3 UnrealEd ”Runtime Error 20005” Experimental Fix . 78
3.15 May 31, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.15.1 Latest News . . . . . . . . . . . . . . . . . . . . . . . 79
3.15.2 Holiday? . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.15.3 New 225f Patch for server administrators only . . . . 80
3.15.4 Server 225e Patch Feedback . . . . . . . . . . . . . . 80
3.15.5 UnrealEd Experimental Fix 2 . . . . . . . . . . . . . 80
3.16 Jun 01, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.16.1 UnrealEd Follow-Up . . . . . . . . . . . . . . . . . . 81
3.16.2 Server Cheats? . . . . . . . . . . . . . . . . . . . . . . 81
3.16.3 UnrealEd for 3dfx Voodoo3 bundle owners . . . . . 82
3.17 Jun 06, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.17.1 Direct3D Improvements . . . . . . . . . . . . . . . . 82
3.18 Jun 19, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

Tim Sweeney Archive 7 News

3.18.1 Unreal Direct3D . . . . . . . . . . . . . . . . . . . . . 83

3.18.2 Summary of things I learned about Direct3D . . . . 84
3.19 Jun 30, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.19.1 Direct3D Is Happening . . . . . . . . . . . . . . . . . 85
3.20 Jul 01, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.20.1 226 Progress . . . . . . . . . . . . . . . . . . . . . . . 86
3.20.2 Recommended Reading . . . . . . . . . . . . . . . . 87
3.21 Jul 03, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.21.1 On Our World Domination Plans . . . . . . . . . . . 87
3.21.2 What’s really happening here? . . . . . . . . . . . . . 88
3.21.3 The Economics . . . . . . . . . . . . . . . . . . . . . 89
3.21.4 The Result: General-Purpose Solutions Win . . . . . 90
3.21.5 Where we go from here . . . . . . . . . . . . . . . . . 91
3.22 Jul 10, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.22.1 The AMD Athlon Rocks! . . . . . . . . . . . . . . . . 92
3.23 Jul 12, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.23.1 New Downloads & Links Pages . . . . . . . . . . . . 93
3.24 Aug 14, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.24.1 DirectX7 . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.25 Aug 16, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.25.1 Unreal Tournament development note . . . . . . . . 94
3.26 Sep 14, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.26.1 Unreal Tournament demo schedule . . . . . . . . . 95
3.26.2 DirectX7 . . . . . . . . . . . . . . . . . . . . . . . . . 95

Tim Sweeney Archive 8 News

3.27 Sep 22, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

3.27.1 DirectX7 . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.27.2 What about OpenGL? . . . . . . . . . . . . . . . . . . 97
3.27.3 Experiencing poor Internet play on your Voodoo3
3500TV? . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.28 Sep 29, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.28.1 More Hardware Troubleshooting . . . . . . . . . . . 100
3.28.2 Direct3D Anomalies . . . . . . . . . . . . . . . . . . 101
3.28.3 Athlon In The House . . . . . . . . . . . . . . . . . . 102
3.29 Oct 01, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.29.1 TNT Users, Try This! . . . . . . . . . . . . . . . . . . 102
3.29.2 Latest Findings . . . . . . . . . . . . . . . . . . . . . 103
3.29.3 Links . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.30 Oct 02, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.30.1 Unreal Tournament players overtake QuakeWorld
at #5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.31 Oct 06, 1999 . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.31.1 Windows 2000 RC2 Rocks! . . . . . . . . . . . . . . . 105
3.31.2 Windows 2000 Internationalization . . . . . . . . . . 106

4 2000 News 108

4.1 Feb 22, 2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.1.1 Common Questions I’m Asked . . . . . . . . . . . . 108
4.1.2 Programming Languages . . . . . . . . . . . . . . . . 110
4.1.3 Also... . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Tim Sweeney Archive 9 News

4.2 Feb 28, 2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Chapter 1


This is a collection of (almost) all the news posts by Tim Sweeney from
the original Unreal Technology archive. Posts not by Tim Sweeney or
not really having anything engine-related may have been left out. The
archive continues to 2003, but Tim Sweeney unfortunately stopped post-
ing in early 2000.

Chapter 2

1998 News

2.1 Oct 13, 1998

2.1.1 Unreal Cheat-Finding Challenge

I’ve seen two reports that people have found a way to cheat in Internet
play. One person reported seeing a little tiny 1-inch tall player walking
around named ”LilTinyChessPiece”, and another reported a player who
never took damage. I’ve been scouring the code, and haven’t found any
holes which would enable this kind of cheating, but I’d love to hear if
anybody has found a way.
If you find a way for players to cheat or crash servers over the Internet
in Unreal, please email The first person to
report a particular security hole in Unreal will receive credit and thanks
here for helping to protect the Unreal community from cheaters.
People who cheat in Deathmatch are lamers, but people who are moti-
vated by simply discovering ways to cheat are doing valuable research.
Note: I don’t consider the following things Unreal cheats:

• Server administrators cheating on their own server. Since the ad-

min is in control of the server, there are many ways he could dis-

Tim Sweeney Archive 12 News

advantage players (in any game, not just Unreal), but fortunately
people who go to the effort of setting up their own server tend to be
honest and respectful of the community.

• Any of the standard TCP/IP tricks people can use to confuse Web
& FTP servers, such as SYN flooding, packet flooding, DNS spoof-
ing, and so on. These aren’t Unreal-specific cheats, and I can’t stop

2.1.2 UnrealScript Notes

There was recently a ”to script or not to script” debate about the merits of
UnrealScript vs pure C/C++. In this debate, several misconceptions were
put forth, which I want to correct:

1. UnrealScript replicated functions (described in the http://unreal. indeed return instantly, rather than
blocking execution while waiting for network acknowledgement.
Function calls are sent across the network using an asynchronous
RPC-like approach, and are executed upon reception. They can ei-
ther be sent wih guaranteed delivery and sequencing (by using the
”reliable” keyword), or without sequencing (using ”unrelible”). Un-
reliable replicated functions are most efficient, and should be used
for non-vital events like hearing sounds.
If one were to assume UnrealScript functions were synchronous
and blocking, one would indeed draw incorrect conclusions such
as this statement: ”That must make AIs in multiplayer games really
fun, they fire at you and do an RPC, then freeze until they get a ping
back from the client.”

2. As our licensees know, UnrealScript classes and C++ classes fully

interoperate. You can derive C++ classes from UnrealScript classes,
and vice-versa. For all actor classes, running ”Unreal.exe -make”
automatically parses the UnrealScript classes, and generates C++
stubs for them, including:

2.1. OCT 13, 1998

Tim Sweeney Archive 13 News

• C++ variable declarations mirroring the UnrealScript ones, al-

lowing variables to be accessed identically in either language.
• C++ -> UnrealScript calling stubs, enabling C++ functions to
call into UnrealScript functions.
• UnrealScript -> C++ calling stubs, enabling UnrealScript func-
tions to call into C++.

Indeed, we invested a lot of effort into providing dual programming

interfaces, enabling seamless sharing of objects and code between
the two languages. Among other things, I studied COM, MFC/VB/ActiveX,
CORBA/IDL, and JNI in order to determine the most efficient and
least error-prone methods of cross-language programming. There
is a lot of great research material on the net, for those who are in-
I suspect that people who have a low opinion of mixed-language
programming have likely drawn their conclusions from JNI and CORBA/IDL
which do indeed involve a lot of messy handle/id references and
error-prone redundent declaration of objects. It doesn’t have to be
this way. For an example of ”cross-language programming done
right”, check out the ActiveX support in Microsoft Visual J++ and
Visual Basic – they’re ultra-clean, automatic, and they involve no
redundency. MFC also does a pretty good nearly-automatic job of
cross-language ActiveX object handling. Mixed-language program-
ming can be very fruitful, if it’s implemented properly.

3. Unreal performs all file loading and saving using serialization. Se-
rialization is automatic for objects defined in UnrealScript. For ob-
jects and structures defined only in C++, such as the following func-
tion that serializes a vector.
friend FArchive& FVector::operator<<( FArchive& Ar, FVector& V )
return Ar << V.X << V.Y << V.Z;

For programmers just learning about object-orientation, a must-

read is:

2.1. OCT 13, 1998

Tim Sweeney Archive 14 News

qid=908310340/sr=2-1/002-5033601-8164246 (Addison-Wesley Pro-

fessional Computing) by

2.2 Oct 14, 1998

2.2.1 Network Objects Follow-Up

I’ve chatted with several people lately who have added insight to the dis-
cussion of multi-language programming and networking. One interest-
ing topic mentioned was using the CORBA protocol to coordinate net-
work games. CORBA is a high-reliability protocol with a fairly high amount
of overhead and latency, due to its exacting nature and use of TCP. I see
CORBA as a very suitable protocol for network game backbones. For
example, if you were doing an Ultima Online style world with multiple
servers that stay in synch, CORBA would be good for linking servers to-
gether in a high-reliability, non performance-critical way. But, CORBA
isn’t suited for an actual realtime modem-based gameplay protocol, since
it’s TCP based (which adds terrible lag under lossy conditions), and the
protocol has enough overhead that it’s not going to be able to handle
large amounts of data over the modem.

2.2.2 AWT is a graphical user-interface layer for Un-

realScript, which has been making some cool progress lately.

2.2.3 Cheat Challenge Update

I’ve received a bunch of tips from two Unreal enthusiasts. Congratula-

tions go to Mek @ PlanetUnreal and RisingSun for discovering and veri-
fying two holes in Unreal, and to Mike Lambert for suggesting a potential

2.2. OCT 14, 1998

Tim Sweeney Archive 15 News

third one. Also, thanks all the players who provided observations they
saw of potential cheating online. I’ll write more about these techniques
after the next patch has been released, and the holes plugged.
In the meantime, if you too have figured out how to cheat in Unreal,
pat yourself on the back, you outsmarted me! But also, please resist the
temptation to cheat on the public deathmatch spoils the fun.

2.2.4 Another Public Master Server

This just in:

How about a mention of the OGN Public Unreal Master Server?

It’s been available for almost two weeks and has no access re-
strictions, bandwidth or otherwise. The server is running @ Server lists can be retrieved @ http: //
master. ogn. org: 6669/ list/ game/ unreal . A page describ-
ing how to link to it can be found @ http: // www. ogn. org/
– sky G.
http: // www. qhunter. com/
http: // www. golightly. org/
http: // www. aurorablue. org/

2.3 Oct 22, 1998

2.3.1 Linux Unreal has reported a way to run the existing ver-

sion of Unreal under Linux, using the Wine Windows emulator for Linux.
They’ve included a screenshot and more information at this http://www. page.

2.3. OCT 22, 1998

Tim Sweeney Archive 16 News

2.3.2 Hired Guns

Want to see something cool and unique that’s being done with the Un-
real engine? Check out the
98/13:01:36.html preview. The team is
really pushing the engine, and has implemented a beautiful user-interface
design featuring multiple game windows. I look forward to playing this
game. (Thanks for pointing out the pre-

2.4 Oct 24, 1998

2.4.1 Almost there...

The latest Direct3D support is now integrated into the Unreal codebase.
It will be in the upcoming 218 patch in ”public beta” form. It hasn’t gone
through wide testing yet, but we and our partners have been testing it
a lot internally on Windows 95 and 98. It’s looking good on the next-
generation 3D cards (Riva TNT, Matrox G200, Savage 3D), though high-
level optimizations are planned that will boost performance further.
The Voodoo2 multitexture code is in, thanks to the efforts of Jack Math-
ews at 3dfx.
I’m doing some final Internet play testing and tweaking on the patch. My
”torture test” environment is an ISP connection at 24.4K with 400 msec
ping. If it’s playable on that, it should be playable on anything. :) Yes-
terday’s testing revealed a few last-minute problems that I’m working on

2.4.2 More discussion on CORBA provided this insight into CORBA with respect to

realtime apps:

2.4. OCT 24, 1998

Tim Sweeney Archive 17 News

In general, CORBA causes too much overhead and latency to

be a server/clients gaming protocol. This is due to IIOP’s (the
default inter-ORB protocol) use of TCP/IP, and the use of large
packet sizes (compared to the dedicated bit-tuned gaming pro-
tocols being used in, for example, Quake2 and Unreal). Though
support for IIOP is mandatory, an ORB vendor can deploy more
efficient (or more secure) protocols between ’his’ ORBs, option-
ally using other network protocols. By default, CORBA calls are
synchronous (that is, the calling object is blocked until a re-
sponse is received from the called object). Such a connection-
oriented call is best implemented using TCP.
CORBA also supports so-called ’one-way’ operations, that en-
able the caller to continue processing immediately after mak-
ing the call - the call is non-blocking. Here, it might make sense
to use UDP as an underlying protocol, provided you deal with
the unreliability at the application level (the ’call’ might not
arrive at the client). Some specialized high-performance ORBs,
such as Nortel’s RCP-ORB, <> ,
use other transport protocols than TCP for one-way calls. Most
ORBs implement optimized mechanisms for ’intra-server’ and
’intra-process’ communication (not that relevant for multi-player

I’ll post more thoughts on the topic in the future. With Unreal, distributed
object models are already playing a role in multiplayer gaming, but only
in a ”tip of the iceberg” sort of way. This kind of technology will expand
very interestingly in the future.

2.5 Oct 25, 1998

2.5.1 Geting closer...

We’re gearing up for a big internal test session with 218 right now. The
release date of the 218 patch will depend on the outcome. Note that 218

2.5. OCT 25, 1998

Tim Sweeney Archive 18 News

is not network-compatible with 217, so if you see our 2 test servers in

GameSpy, please don’t try to join (you won’t be able to).

2.6 Oct 26, 1998

2.6.1 Direct3D Drivers

To use Unreal’s Direct3D support, you must have

com/directx/download.asp installed. You also need the latest Direct3D
driver from your 3D hardware maker. Because Unreal uses many new Di-
rect3D features, many hardware makers had to update their drivers for
Unreal to work. Therefore, some of the drivers distributed with DirectX
6.0 have problems with Unreal (crashes or visual anomalies). Here are
driver download sites for various brands:

• Permedia

• Voodoo, Voodoo

Rush, Voodoo2, Voodoo Banshee

• Rage

• i740

• G200

• Riva 128 (known problems), Riva


• Verite

• Savage 3D

• (untested, prob-

ably doesn’t work)

• (untested, prob-

ably doesn’t work)

2.6. OCT 26, 1998

Tim Sweeney Archive 19 News

Please keep in mind that Unreal’s Direct3D support is an early beta and
isn’t yet working reliably on all 3D hardware.

2.6.2 Known Issues

I’ve update the list of, summa-

rizing the problems people have found with 219. There are a bunch of
items, but most are non-showstoppers. I’m working on them now. Over-
all, though, 218 is by far the best-received version of Unreal to date! New
servers are popping up rapidly (50% increase in the past 18 hours), with
player counts rising even faster.

2.6.3 Unreal 218 Gaining Support

After releasing the patch at midnight Sunday night, it was amazing to

see how many Unreal servers had it up and running in less than an hour
It will be interesting to watch new Unreal servers spring up on the net
now that our Internet play quality is under control. In the not-too-distant
future, we’ll have many more improvements in Unreal network play: per-
formance enhancements, new features requested by the community, and
several cool upcoming surprises. GameSpy currently shows 27 Unreal
servers, and a couple thousand for Quake. Clearly, we got off to a very
slow start while we learned the the ins-and-outs of Internet game pro-
gramming; now Unreal is on a solid footing and we’re really beginning to
pick up the pace!
Level designers in the Unreal community have already started producing
some great Deathmatch maps (see
for downloads), so we expect to see things accelerate in this area, too.

2.6. OCT 26, 1998

Tim Sweeney Archive 20 News

2.6.4 PlanetUnreal performance survey

If you’re using the new 218 patch, please fill out http://www.planetunreal.
com/ Unreal performance survey. We’ll be watching Unreal players’ feed-
back in the survey, and looking at all the 218 bug reports sent to the account. Everyone’s feedback will be
taken into account for future improvements to the Internet play and other

2.6.5 More about 3D API’s

In an interview on, I gave my opinion in

the ongoing OpenGL vs Direct3D debate, and I wanted to add some more
information here.
First of all, both our DreamForge partners working on the OpenGL sup-
port, and our Direct3D partner have both done a very high-quality job,
given the constraints of Unreal and the 3D card drivers available. My
feelings on the two 3D standards are based completely on their API’s and
the quality of drivers available.
Next, I wanted to give some more detail on the code.

API support layer Lines of C++ code

3dfx Glide 1711
Direct3D 3123
NEC PowerVR 4049
OpenGL 5697

Now, the Glide is the simplest, because it’s aimed at the Voodoo family
of 3D cards, which are very standard and straightforward to support. For
programming to 3dfx cards, Glide is the ideal API. For people thinking
”If Glide is so great, why don’t other 3D cards support it?” The short an-
swer is, if Glide supported other manufacturer’s 3D cards, it would pick
up all the complications involved in real-world Direct3D and OpenGL
programming: Testing for core capabilities, working around driver bugs,

2.6. OCT 26, 1998

Tim Sweeney Archive 21 News

The Direct3D code is next in simplicity. It began as a working 1600 line

driver, then expanded as it was optimized and support was added for
3D cards that lack key features. The Direct3D API follows the traditional
Microsoft model of being not very beautiful code, but dealing with real-
world hardware robustly, by providing a capability-querying mechanism
and well-tested drivers.
The PowerVR code is large because the PCX2 chip’s rendering approach
is very different than what Unreal was designed for. This makes the im-
plementation fairly complex, though the next-generation PowerVR, cur-
rently on-hold in the PC market while NEC focuses on the Sega Dream-
Cast, is a great chip that’s more traditional in its architecture, and more
optimal for Unreal.
The OpenGL code is large because of optimizations and support for lots
of real-world hardware. If reduced to plain-vanilla OpenGL, this would
probably be the simplest and most stable driver of them all, because the
OpenGL API is very straightforward.

2.6.6 Unreal 218 Patch is ready!

You can read the release notes and download the patch from the http:
//! Our test session went well, and
the participants (including a lot of modem players, and even some over-
seas modem players) reported major success with the performance im-
Epic’s two deathmatch servers have been upgraded to 218: unreal:// and unreal:// However, since
our T1 line is going to be very saturated with people downloading the
patch, performance probably won’t be very good on our server for the
next several days.

2.6. OCT 26, 1998

Tim Sweeney Archive 22 News

2.7 Oct 28, 1998

2.7.1 Latest Happenings

Since the Unreal 218 patch was released, the number of Unreal servers
showing up in GameSpy has increased by 225%. A huge thanks to every-
one who is running Unreal servers now. It’s awesome to see the commu-
nity grow and begin to reach critical mass.
At any time, about 20% of the servers listed in GameSpy are shown as
”Not Responding”. I’ve been looking into this, and there are three reasons
for this.
One, the server has been shut down or crashed but hasn’t timed out and
been removed from the list. Two, the server is switching levels, which
prevents it from responding to GameSpy queries for 5-20 seconds. Third,
the server is shown in GameSpy because a user running a LAN server
(dedicated or non-dedicated) enabled the ”DoUplink” option to publi-
cise their server to the world without realizing the consequences, but ei-
ther exited or is on a LAN blocked off by a firewall. We’ll work with the
GameSpy guys to adjust the timeouts so that fewer ”Not Responding”
servers are listed.

2.7.2 Status of Direct3d - All Drivers

Some issues across-the-board with switching between fullscreen and win-

dowed mode, and switching resolutions during gameplay. We’re looking
into theses issues.

2.7.3 Riva TNT

Has lots of potential, but there are major problems with their Direct3D
driver: rainbow colors, time-lag, w-buffering issues, missing textures. We
are hoping nVidia will issue an updated driver soon. The ball is in their
court now.

2.7. OCT 28, 1998

Tim Sweeney Archive 23 News

2.7.4 Riva 128

Their latest driver should run Unreal OK. The framerate isn’t very good,
and there will be a lot of ”stuttering” since texture downloads are very
slow on this hardware. Some users are reporting that creature polygons
appear corrupted. Hopefully nVidia will fix thie error in their driver.

2.7.5 Intel i740

There are some problems with their currently-available Direct3D driver,

but they have an internal beta driver which fixes many problems. Hope-
fully this will be ready for the public soon.

2.7.6 ATI Rage Pro

They have a beta Direct3D driver which runs Unreal pretty well (though
it doesn’t appear to be available for public release yet). Multitexturing
currently isn’t working with Unreal, because of a hardware limitation (it
can’t multitexture a paletted base texture map along with a non-paletted
light map).

2.7.7 3dfx Voodoo2

There isn’t any reason to use Direct3D for Voodoo2, since we support
Glide natively and that tends to be more stable and faster, but their latest
Direct3D driver does run Unreal pretty well. Their detail texture blend-
ing (D3DBLEND DIFFUSEALPHA) doesn’t work, which breaks detail tex-
tures. There are reports of missing lightmaps in some areas (not sure
what’s wrong). 3dfx could get a performance increase here by imple-
menting subrect Blts and hardware triangle-fan setup.

2.7. OCT 28, 1998

Tim Sweeney Archive 24 News

2.7.8 S3 Savage 3D

Works well with their internal beta Direct3D driver. Not sure when this
will be available via the web.

2.7.9 Matrox G200

Works pretty well.

2.7.10 Rendition

Works pretty well.

2.7.11 Permedia

Only supports monochrome lighting due to blending limitations. Explo-

sions are displayed with a black background (due to a bug in the current
Unreal code). Fairly slow texture downloading – performance isn’t great.

2.7.12 Drivers we consider playable, in best-to-worst or-

• Voodoo2: Fastest of all.
• Savage/G200: Reasonable performance. Not as fast a fill rate as
Voodoo2, but texture downloads are fast, and they support palettes,
making for a steady frame rate.
• Rage Pro/Rendition: Reasonable performance. High display qual-
ity in truecolor mode.

The others should be playable pending public release of improved Di-

rect3D drivers for their hardware.

2.7. OCT 28, 1998

Tim Sweeney Archive 25 News

On Permedia, Riva 128, and Rendition, you may get better performance
using Unreal’s software renderer, especially if you have a Pentium II class
processor. This isn’t such a bad thing, because the software renderer sup-
ports all of Unreal’s key features, including truecolor lighting and volu-
metric fog.

2.8 Oct 30, 1998

2.8.1 Unreal 218 patch information (RTFM!)

If you downloaded the 218 patch, please

versions.htm. If you’re still experiencing problems with Unreal on your
modem connection, you need to use the NETSPEED command described
there, so that Unreal doesn’t try to use more bandwidth than you have

2.8.2 Resources for Server Administrators

Here are some resources to aid new Unreal server administrators in get-
ting up and running:

• How to get listed in Game-

Spy and other querying tools (new).

• Basic information

about running an Unreal server.

• New ”Server Administration”

forum. Steve Polge, Brandon Reinhart, and I will be answering ques-
tions from server administrators and people writing server-querying

2.8. OCT 30, 1998

Tim Sweeney Archive 26 News

2.8.3 Unreal Multiplayer Ramp-Up

A lot of news sites have been picking up on Unreal’s multiplayer improve-





The number of public Unreal servers listed in GameSpy is now up to 51

as I write this, and has been growing at a rate of over 20% daily. While
working on the upcoming 219 patch, I’ve been spending a couple hours
each day playing Unreal online, especially on the servers running user-
created levels. The community has made some really impressive levels,
and I’m glad to see that a lot of servers are now making use of download-
able levels, textures, sounds, and music.

2.8.4 Other Servers

In addition to the public servers listed in GameSpy, there are 14 extremely

high quality, public servers on listed in their
own web-based ”server finder” site.
I should also mention our other online partner, http://www.mplayer.
com/action/games/unreal/, and the
news site on MPlayer.

2.8. OCT 30, 1998

Tim Sweeney Archive 27 News

2.9 Nov 01, 1998

2.9.1 Unreal Grows...

More public servers keep popping up every day. Awesome!

I noticed that the percentage of servers shown as ”not responding” has
gone up, though I’m not sure whether that’s due to server crashes, or peo-
ple accidentally having their LAN server listed in GameSpy when it’s not
meant to be public. Server admins, if your server crashes unexpectedly,
please email your Unreal.log file to so
we can investigate it.
A reminder, server administrators can get help in the new ”Server Ad-
ministration” forum on the
According to the feedback we receive, and in http://www.planetunreal.
com/’s survey, 218 has fixed the past network problems and made Un-
real playable on the Internet for the large majority of people. Still, we
have lots of improvements in the works, which will further increase the
game’s smoothness and responsiveness during network play, as well as a
few new features that I’m not allowed to announce yet. :) The next patch,
219, will be purely bug fixes to 218 (no speed difference). After that, 220
will contain some more high-level improvements to many areas of the
network code.
Currently, there seem to be two sets of people who are still having some
problems with Unreal on the Internet. One set has 28.8K-33.6K modem
connections that experience an unusally high amount of packet loss. 220
should improve this a bit, but when you have upwards of 20% packet loss,
gameplay is bound to suffer. The other set of problems is being reported
by people with some unusual connections we haven’t been able to test
on internally: people with dual simultaneous modem connections, peo-
ple with connections that receive via satellite and send via modem, and
some similar configurations. I haven’t determined what’s going wrong in
these cases yet, but I suspect packet-order problems.

2.9. NOV 01, 1998

Tim Sweeney Archive 28 News

2.10 Nov 02, 1998

2.10.1 Unreal 219!

It’s now available for download

htm. The Epic Deathmatch servers will be shut down for the night, to pro-
vide more bandwidth for downloads.

2.10.2 The testing is going well

Ok, ok, you can stop now, that’s enough testers. :) Server performance
has definitely improved since I last benchmarked it. My dual-processor
Pentium II-300 is serving a 20-player game, copying 600 megs across the
network, and running a 3dfx Unreal client at 800x600. The server is keep-
ing up a steady 15 fps. I have more optimizations planned, though.
On my 28.8K connection, gameplay has remained pretty smooth until I
get into a firefight with 5-6 people, at which point the movement of other
players starts to become unsmooth. We should be able to improve this
significantly further in 220, with two protocol improvements. First, I’ll be
improving the scheme for managing temporary objects like projectiles,
which should reduce their bandwidth usage by around 30%. Second,
Steve Polge and I working out a new variable priority predictor/corrector
scheme for transmitting the movement of other players.

2.10.3 Testing 219

I have an Unreal 219 server up and running at unreal://

in case a few people want to help test. You can join in and play with 218;
it’s fully compatible. 219 should be available soon, unless new bugs are

2.10. NOV 02, 1998

Tim Sweeney Archive 29 News

2.11 Nov 10, 1998

2.11.1 Latest News

The latest survey shows a significant gain

in gamers’ satisfaction with the latest 219 patch, compared to the previ-
ous ones. We’re now hammering on version 220, which includes signif-
icant new Internet play improvements, plus some core game improve-
ments that the rest of the team has been working on. 220 won’t be out
this week, but should be ready soon after.
There seems to be one Internet problem in 219 that is affecting 10-15%
of players. I haven’t figured out the cause yet, but the symptom is a ping
time (visible with STAT NET) that swings around wildly, going from 200 to
1000 or more over the course of several seconds. Many of these players
report that other games like Quake 2 don’t exhibit the problem–which
indicates it’s not just a bad connection or ISP. I suspect packet-ordering
problems, though I haven’t been able to duplicate it here.

2.11.2 The Roadmap

In the remainder of 1998, we will be releasing:

• Unreal 219: Latest version. Download http://unreal.epicgames.


• Unreal 220: Further Internet improvements, core game improve-

ments, bug fixes.

• Unreal 221: Incremental fixes

• Unreal 222: Offical (non-beta) patch; shareware version will be re-


• Additional patches as necessary.

2.11. NOV 10, 1998

Tim Sweeney Archive 30 News

We also have some cool things in the works for 1999. Since Unreal first
shipped, I’ve been working on Internet improvements and bug fixes, while
the rest of the team has been working on a top-secret project we are re-
ferring to as ”Unreal Tournament”. There have been a lot of rumors and
speculation about what this project will be. Some of the rumors are com-
pletely incorrect. For example, we are not
using any alien technology recovered from crashed spaceships on Tour-
nament. (We’re saving that for our next project).
I’m also working on the new, enhanced UnrealEd, with lots of new fea-
tures, improved stability, a plug-in interface, and other goodies. We don’t
have a date for this, but I’m expecting a public beta around new years,
with final release early in 1999.
At some point, we’ll be publically releasing Unreal’s C++ interface so the
community can start on .dll plug-ins for Unreal and UnrealEd. This will
happen no later than the new UnrealEd public beta. It might be possible
to release it sooner, however there are changes in the works which won’t
be binary compatible, so user-created .dll’s would break and need to be
What can you do with Unreal .dll’s? Lots. Here are some examples:

• Define new kinds of algorithmic textures that can be created in Un-

realEd and used in levels.

• Expose anything you can do in C++ to UnrealScript, for example the

file system, ActiveX, more Internet functionality, etc.

• Write performance-critical game support routines, such as AI.

While it’s also possible to write game code directly in C++, the preferable
way to do that is UnrealScript, since there is already a lot of game support
infrastructure built into the language.

2.11. NOV 10, 1998

Tim Sweeney Archive 31 News

2.11.3 Server Administration

Remote server administration isn’t working in 219, but will be addressed

shortly. 219 does have a few new features for local server administration.
While running a server, click on the Unreal server icon and choose ”Show
log window”. In the log, you can type:

• SERVERTRAVEL < levelname> : Jump to a new server.

• SAY < message> : Broadcast a message to all users.

-Tim Sweeney

2.12 Nov 19, 1998

2.12.1 Unreal 220 Internet Improvements

I’m now down in Raleigh, North Carolina with the rest of the Unreal team.
Things are really coming together for the top-secret project we’re working
on, and Unreal 220 is starting to solidify. We don’t have a release date yet,
but it could be as early as next week.
The Internet play improvements in 220 are quite noticeable. Steve and I
are using a new physics prediction scheme based on motion compen-
sation. This improves the smoothness of other players’ movement in
deathmatch games, especially in large, open levels like DmDeck16, where
lots of players are visible.
Steve has redesigned the moving brush code so that moving brushes now
are now handled client-side. Platforms and doors are very smooth now.
Finally, I’ve created a new scheme for managing temporary actors like
projectiles, which reduces their bandwidth by about 30%.
Steve and Brandon have also added a lot of other improvements, such
as showing ping times, server name, and level name in the player list,
showing the name of opponents when you point the crosshair at them in

2.12. NOV 19, 1998

Tim Sweeney Archive 32 News

a network game, a beep when chat messages come in, and color coding
of messages.

2.12.2 Unreal Editor Progress

I’ve also been working on the new Unreal Editor, featuring a user-interface
rewrite (moving over to C++ from Visual Basic). It’s now more than half-
way done. Architecturally, the new editor code is far more general than
the old UnrealEd. The editor itself is very abstract, and almost all func-
tionality is implemented through plug-ins. There are plug-in interfaces
for brush builders, views of objects (like 3D scenes and the 2D editor),
view controllers, import tools, export tools, browsers, and operations on
arbitrary kinds of objects. Plug-ins can even define their own sub plug-in
interfaces, etc.
This new approach will provide a more solid foundation for Unreal 1 edit-
ing, but the big gains will come in the next-generation engine. A ma-
jor bottleneck in creating the Unreal 1 engine was the complex coupling
between the rendering engine and the editor, which made it difficult to
modify the renderer. The new approach generalizes the concept of edit-
ing, and will make it easier to add cool new features which require editor
I plan to start releasing public betas of the new UnrealEd as soon as it’s
really stabilized, with the first beta in the first week or two of 1999 if all
goes well.
We have long planned to release the editor as a full retail product through
GT Interactive after it has been cleaned up and finalized, but there doesn’t
seem to be enough mainstream retail interest to justify it. So, unless
something changes real soon, we’ll plan to release the final version freely,
with the manual provided in HtmlHelp format rather than printed.

2.12. NOV 19, 1998

Tim Sweeney Archive 33 News

2.12.3 Web Site Notes

A request of web sites providing collections of Unreal related files, such as

the cool page: Please link to the of-
ficial page for the Unreal
patches, rather than mirroring specific patch versions locally. A bunch of
web sites have old Unreal patches in their download sections (as old as
the 1.01 patch), which are no longer relevent. If everyone links to the cen-
tral page, users are more
likely to get the latest patch. Thanks!
Cool Unreal related things I’ve seen recently...

•, a great ”overhead

view” mod including some custom levels. It feels a lot like the clas-
sic arcade game Gauntlet.
•, a totally cool stats
tool for server administrators.
• has grown to contain more and more
useful UnrealEd resources.

2.12.4 Unreal mod notes

Mod developers who get the cryptic ”sandbox” error when trying their
mods in network play: This error means you have a .upkg file that says
server-side-only because you copied the one from IpDrv.u (which is SUP-
POSED to be server-side-only). This usually prevents the mod from work-
ing, but sometimes just causes actors defined in the mod to not be visible
to players. If you don’t have a .upkg problem, you won’t have a problem.
If you do have a .upkg file, it should read:


2.12. NOV 19, 1998

Tim Sweeney Archive 34 News

-Tim Sweeney

2.13 Dec 10, 1998

2.13.1 No News

We’re still testing and tweaking Unreal 220. It’s getting closer... has a cool
1.16/bent/ that compares, contrasts, and critiques some next-generation
game development tools, and focuses on the strengths and weaknesses
of UnrealEd and 3D Studio Max. The author makes a lot of good points
about features that would improve UnrealEd, many of which are already
planned for the upcoming C++ UnrealEd rewrite.

2.14 Dec 11, 1998

Here is the file from the

upcoming 220 patch.
We’re doing some final testing on it now. If that goes well, 220 will be re-
leased Saturday. Otherwise, we’ll spend a few more days fixing and test-

2.15 Dec 12, 1998

The team played a huge coop game through all the levels of Unreal, and
220 is looking good – ”almost there”. Here’s the list of bugs we found dur-
ing the test session, which we’re fixing now:
tim’s issues:

2.13. DEC 10, 1998

Tim Sweeney Archive 35 News

• ”brightness” txt isn’t shown

• brightness setting not saved

• alt-enter loses menu in w95 glide but ”advanced options” doesn’t

• verify ipdrv not loaded for normal play

• avoid losing savegames! (firstrun detect *.usa)

• no music when enter dig coop

• spinning death anim plays superfast

• ”connecting” msg shows current level name not next

• ”netspeed” setting doesn’t persist across level switches in lan game

• cliff says screenshot crashing

• make sure all mover subclasses are bAlwaysRelevent

• text wrap: additional lines have leading spaces

steve issues:

• no pickup sound for kevlar vest

• dude in chair isn’t screaming

• sparking wire anim in vortex rikers isn’t showing


• coop scoreboard needed, to show who’s in the game

2.15. DEC 12, 1998

Tim Sweeney Archive 36 News

2.16 Dec 14, 1998

2.16.1 Hard at work

We’re hard at work putting the finishing touches on Unreal 220. It’s look-
ing good, but the extent of the changes and improvements has required
quite a lot of testing and refinement. It won’t be ready for public con-
sumption this week or over the weekend, but hopefully early next week.
A lot of mod authors have emailed asking for early access to Unreal 220 so
they can get their stuff is up and running in the new version. The answer
right now is ”maybe”...if the code comes together over the weekend, I’ll
post a public ”for mod authors only” version. 220 will break compatibility
with most UnrealScript mods, but I’ve modified UnrealEd so that mods
can still be loaded (this is a change: 219’s UnrealEd gives a fatal error
when loading an incompatible .u file).
Upgrading most mods will be easy; the most significant changes are func-
tion parameter changes all over the place, and splitting off the network-
related variables in Pawn into a new PlayerReplicationInfo class, to make
scoreboard-programming simpler and more expandable. In terms of mod
support, the week following 220’s release will be chaotic, with mods be-
ing repaired and re-released. Sorry, we would have liked to avoid this, but
the improvements we wanted to make were just too cool to stick with full
Levels, sounds, texture files, music files, and audio files remain backwards-
Shortly after 220, I’ll be releasing the C++ headers required for users to
write C++ add-ons. I’m planning on throwing together a very simple
SampleTexture package showing how to make algorithmic textures, and
a SampleActor package showing how to write intrinsic C++ functions for
actors. This stuff should give mod authors a very significant amount of
new stuff to play with, though the more exciting C++ stuff will come in
a later release with the UnrealEd plug-in interface. It will be interesting
to see the response to the C++ headers, because the Unreal engine’s C++
architecture is very, very different than what people are used to in Quake

2.16. DEC 14, 1998

Tim Sweeney Archive 37 News

and other engines. It’s very much a framework rather than an API.
Jack ”Mek” Porter has come here from Australia for a few months to help
with the Unreal Tournament programming, and is currently working on
some cool new mouse-based user interface code. He is a big Linux fan,
and is talking with Brandon Reinhart about cooperating on a native Linux
port of the Unreal server.

2.16.2 That’s

The cool independent Unreal editing site I mentioned is actually http:

//, not

2.17 Dec 15, 1998

2.17.1 Unreal Networking Code: Status

Unreal 220 incorporates the majority of the low-level bandwidth opti-

mizations I’ve been planning for Unreal 1 and Unreal Tournament. I was
just looking back at the initial shipping version of Unreal and comparing
packets, and realized that we’ve achieved a bandwidth usage improve-
ment of over 250% in both directions.
This doesn’t mean we’re going to stop now...there are some minor proto-
col improvements planned, and also high-level game play optimizations
that Steve and Brandon are working on, moving more logic to the client-
side to reduce lag and save bandwidth. But, the foundation is now very
solid, and the rest of our work mostly consists of building on that foun-
The cool thing about Unreal’s networking support is that we brought the
real-world Internet play performance up to the state-of-the-art without
sacrificing generality. When I first posted the http://unreal.epicgames.
com/Network.htm, it was simultaneously praised for being a powerful and
very general framework, and flamed because Unreal’s initial Internet play

2.17. DEC 15, 1998

Tim Sweeney Archive 38 News

performance was below gamers’ expectations. The critics were delighted

to claim, ”this networking engine is a Grand Scheme that will never per-
form well in the real world”. They sure were wrong! Iit turned out that all
the performance bottlenecks were in the low-level protocol, and were all
completely fixable given sufficient R&D time.
It took a lot of hard work to identify all of the inefficiencies in the proto-
col and come up with solid, general solutions. The key problems in game
networking are data compression, error correction, and latency reduc-
tion. They turn out to be mutually opposing:

• Standard data compression schemes (i.e. Huffman and arithmetic

coding) work best when you have a large, well-known compres-
sion window, and large symbol tables. Sending symbol tables isn’t
practical when you have a low-bandwidth, high-latency connec-
tion, and when your window history is questionable due to trans-
mission errors.

• Error correction schemes are tough to implement when you have

low-bandwidth, high packet loss conditions (for example., most gamers’
modem-based Internet connections!) I went through 6 separate er-
ror correction schemes in Unreal, 5 of which all fell apart in various
conditions. One big trap is getting into a feedback loop that saps
away bandwidth and increases ping times. Another trap is retrans-
mitting time-sensitive data when it gets lost, causing lag. These is-
sues are tricky in bandwidth-intensive action games, because you
need to support multiple tiers of reliability in packets, and some
level of sequencing between them.

• Latency reduction is another tricky issue, because of the way modems

queue data: The larger a packet is sent, the more latency is induced
in the buffering scheme. So, you want to send small packets to keep
latency down. But, UDP packets have significant built-in overhead,
so you want to send large packets to keep bandwidth up.

In the end, I opted to go with several a custom bit-encoding compres-

sion scheme, which adapts itself to the type of data being sent, and intel-
ligently crams it into the least amount of bits possible. This has worked

2.17. DEC 15, 1998

Tim Sweeney Archive 39 News

very well, because the previously large overhead present in packet head-
ers (which identify the game objects being updated) was reduced very
My initial attempts at error correction were based on sliding-window schemes.
These are still supported in the protocol, because they are great for send-
ing latency-insensitive data, for example when you download a map upon
entering an Unreal server. However, for the far majority of game data, I
use an asynchronous scatter-gather data replication scheme which guar-
antees that game objects will all be updated correctly under all packet-
loss conditions, sacrificing guaranteed ordering of updates in order to
improve bandwidth usage.
To trade off latency and bandwidth, I adjust packet sizes based on the
bandwidth available, and hand-tuned the factors to get an acceptable
Behind the scenes, we’ve also optimized a lot of the prioritization logic.
Visible objects are now updated based on a priority queue that is weighted
according to a viewing probability, and a predictor/corrector algorithm
which updates the objects which ”need it most” (based on extrapolated
position error and gameplay priority).
To test everything, I wrote an ”ISP From Hell” simulator into Unreal that
lets you toy around with various packet loss, variable packet lag, and
packet disordering combinations, and the ”STAT NET” display. There’s
also an option built in (C++ source required) which writes an extremely
detailed network traffic analysis to the log file. Having and using these
visualization tools turned out to be invaluable. Without the tools, the
wasted bits, complex feedback loops, and race conditions in the proto-
col would have never have been found.
So, where do we go from here? The majority of the network-related im-
provements coming in the Unreal Tournament timeframe (still unan-
nounced) are high-level, in areas like gameplay rules, client-side predic-
tion, ease of use, and user interface.
Further off in the future, there are a lot of new possibilities that can be
directly extrapolated from the
htm document. Unreal 1 networking can be described simply as ”a small

2.17. DEC 15, 1998

Tim Sweeney Archive 40 News

number of game clients connected to a single server”. However, the pro-

tocol’s low-level design is peer-to-peer rather than client-server. You might
say that Unreal just looks like a client-server engine, because our Un-
real 1 game code favors the server’s logic over the client’s. For the next-
generation engine, I’m going to experiment with a lot of new networking
configurations. For example, the protocol could be used for a server-to-
server backbone, enabling servers to be clustered together with objects
moving seamlessly between them–can you say Ultima Online 3D? There
are also other interesting possibilities for servers, such as a spanning-tree
chat backbone similar to IRC, or UnrealScript application servers (for
example, managing user accounts on one machine, chatting on a sec-
ond, skin repository on a third, supporting a cluster of dedicated game
servers). These aren’t promised features at this point, just possibilities...but
to me, very exciting possibilities.

2.18 Dec 18, 1998

2.18.1 Unreal Tournament notes for mod authors

I’ve had several inquiries from mod makers about how hard it will be to
”port maps and mods from Unreal 1 to Unreal Tournament”. The answer
is, port? What’s this talk of porting? This stuff just works.
Steve Polge put a lot of effort into updating the Unreal 220+ game code
so that it integrates seamlessly with Unreal Tournament. Both products
install into the \Unreal directory, and are mutually compatible. So, you
can install one, the other, or both – and existing maps and mods which
are compatible with the latest version of Unreal, will be fully compatible.
For convenience, Unreal Tournament ships with all of the script pack-
ages from Unreal 1 (including UnrealShare.u and UnrealI.u), all of the
.utx files, and all of the .uax files. Unreal Tournament is a superset of the
Unreal 1 with the exception of the Unreal 1 maps and music. All new UT
content (lots of scripts, textures, sounds, game types, and new actors) are

2.18. DEC 18, 1998

Tim Sweeney Archive 41 News

designed to interoperate with the existing ones in Unreal 1.

The new player code is implemented in a subclass of UnrealIPlayer, which
means that Unreal Tournament players will work in user-created death-
match maps for Unreal 1.
However, the new UnrealTournamentPlayer classes and meshes include
new features and animations, so you can choose to take advantage of this
(and require UT), or not. The new game types are GameInfo subclasses.
The new weapons and pickups interoperate with the old ones.
UnrealEd supports Unreal Tournament maps without any modifications.
This ”just works” in the same way that most of the licensees use UnrealEd
without modifications.
One huge Unreal Tournament feature that script programmers will be in-
terested in is the all-new user interface code. While Unreal 1’s menus are
functional, the ”look and feel” is basically derived from Doom’s menus,
which were derived from Wolfenstein 3D’s menus, which probably date
back to some even earlier id Software game.
The new game UI is a cutting-edge redesign, and continues our tradition
of creating code and tools which licensees and mod makers can endlessly
customize and tweak. I can’t say more about it yet, but the ongoing joke
is that we’re going to rename it UnrealOS (it’s just a joke, but the anal-
ogy will become clear later). Maybe we’ll release UI screenshots at some
-Tim Sweeney

2.18. DEC 18, 1998

Chapter 3

1999 News

3.1 Jan 01, 1999

3.1.1 Happy New Year

The team is now getting back together from the holiday and we’ll get back
to Unreal 221 development first thing tomorrow (Saturday). We were on
the verge of releasing the patch right before Christmas, but we’re now
back to being a few days away, since I added some new features that need
re-testing, and some additional bugs have been found.
For me, the holidays were fun and productive. I hauled my computer up
to Maryland for a one-week visit with my parents, and that provided a
good excuse to work on engine improvements and new R&D in a zero-
distractions environment. Without getting to deep into the details, here
are some of the things I worked on:

• Unicode Unreal. I finished up 16-bit Unicode support; now the

Unreal code base can be recompiled for the ANSI (8-bit) or UNI-
CODE (16-bit) character sets. This has several advantages. First of
all, localizing a game for non-Roman languages (such as Japanese,
Chinese, Korean, Hebrew, Arabic) is easy with Unicode. Second,
we’re really interested in Windows CE, and that is a pure Unicode

Tim Sweeney Archive 43 News

OS. Windows CE is the native operating system of the Dreamcast,

so Unicode is a natural stepping-stone along the Dreamcast port-
ing path which we’re exploring. Windows CE is also used in tons of
palm-top computers. Now, none of these palm-top computers are
currently interesting from a gamer’s perspective, but if you extrapo-
late their grown–in terms of video resolutions, RAM, and processing
power–along a Moore’s Law curve, they’re not that many years away
from being comparable to low-end Pentium II PC’s today. I like to
think that the GameBoy’s of the future won’t be a closed platform,
but rather run a standardized OS like Windows CE.
All Unreal data files are ”character set neutral”, meaning they can
contain any mix of ANSI and Unicode characters, and they’re auto-
matically converted/truncated at load time. Network play similarly
”just works” between the Unicode and non-Unicode versions.
Regarding Unicode font issues (”who’s going to paint all those Japanese
characters?”), Jack Porter has built a TrueType font importer into
UnrealEd, which converts any Windows TrueType font to an Unreal-
compatible font bitmap. This is an essential part of our Unicode ef-
forts, but also ties into the new user interface he and Brandon are
building for Unreal Tournament. There are lots of Unicode fonts
with support for nearly all Windows-supported languages, so local-
izing games into all languages should be fairly painless.

• Data compression. I spent a couple days experimenting with vari-

ous compression schemes and found some interesting results. First
of all, it’s fairly easy to achieve PKZIP-quality compression–the al-
gorithms are well documented and easy to implement. Second,
it’s nearly impossible to beat PKZIP by much of a margin. Beyond
the standard algorithms (Huffman and LZH compression), I exper-
imented with many variations, including:

– ”Double” Huffman-compressing each character in a file then

recursively merging compressed regions together into spans,
with the goal of compressing ”different types of data” with dif-
ferent encoding tables. This approach came out neutral: there
were huge gains in compression, but they were offset by hav-
ing to store lots of Huffman tables.

3.1. JAN 01, 1999

Tim Sweeney Archive 44 News

– Trying Huffman compression in bases other than 2 (such as 3,

5, and 7): no overall gains.
– ”Infinite sliding window”: An LZH-style scheme that spends
a huge amount of CPU time on compression by analyzing an
entire file for differences. Here, the gain was typically 10%, but
it was unbearably slow to work with.
– Arithmetic encoding. This is a Huffman-style scheme that em-
ploys some basic number theory to generate a compression
table using arbitrary bases (other than base 2). This gets real
gains of around 10% compared to Huffman, but is much slower
to decode because it relies on arbitrary-precision integer math.

I didn’t get the compression code into the Unreal master source (it
was standalone R&D). I’m considering supporting compression for
client-downloadable files in Unreal Tournament, to make it easier
to go onto servers running custom maps. My most usable compres-
sion scheme got 4:1 compression on .unr and .u files (nice!) but no
significant compression on textures and sounds.
Unreal’s textures compress poorly for an interesting reason: they
are already stored using palettes, which are a form of compression.
Furthermore, the process of generating an optimal palette for an
image yields an array of bytes which are distributed near-uniformly,
which is the worst possible case for variable-code-word compres-
sion. I think the state-of-the-art in texture compression will be S3’s
real-time S3T/DXT scheme, which is fairly lossy, but optimal for
real-time decompression while rendering. And it’s 6:1 for all pos-
sible kinds of data.
One interesting technique I played around with for texture com-
pression is wavelet compression. The 2D wavelet scheme for tex-
tures is quite similar to mipmaps: to oversimplify a lot, you repre-
sent an nxn image as a sequence of 1x1, 2x2, 4x4, 8x8 ... nxn. Each
bitmap is stored as ”signed differences” relative to the lower-res ver-
sions of the bitmap. Wavelets are a good starting point for lossy
compression schemes, because the ”most visually noticeable” de-
tails are in the low-res bitmaps, and in the high-frequency portions
of the high-res ones. So, it’s fairly easy to filter out unimportant de-
tails and compress what remains. Wavelets are also nice because

3.1. JAN 01, 1999

Tim Sweeney Archive 45 News

you can generate all your texture’s mipmaps from a single wavelet
without any overhead. While I don’t think wavelets will play a role
in our future texture-mapping plans, they look more useful as a way
of storing and manipulating height maps and displacement maps.
Wavelets provide a very efficient, ”level of detail”-aware form of
storing 2D surface geometry.

• 3D API’s. I spent a few days learning the basics of curved surface

rendering, which turned out to be very interesting. I was going
to do my research with Direct3D (under Windows NT 5.0 Beta 3
RC0) but the 3D hardware support isn’t quite there yet, so I took
the opportunity to learn OpenGL using the software rasterizer and
3dfx’s OpenGL minidriver. Both renderers were extremely stable
and easy to use. To my amazement, I never experienced a ”blue
screen of death” or had to reboot–quite a difference from my past
work with Glide and Direct3D! (Note: I haven’t been very closely
involved in the Unreal Direct3D and OpenGL support, which has
been developed by our partners, so this is my first really in-depth
experience).OpenGL’s approach to window management turns out
to be extremely simple and reliable. I went from zero to having
a spinning cube up and running in under 30 minutes. The tools
GL provides for polygon drawing are extremely simple and power-
ful. For just drawing a few lone polygons, glVertex() and its associ-
ated functions are super-easy. For drawing complex meshes, GL’s
approach to providing separate client-side arrays for vertices, col-
ors, and texture coordinates is very easy and efficient for multi-pass
rendering, since you can swap out one table (say, texture coordi-
nates) without affecting the others.
OpenGL’s texture approach (glTexImage2D and glBindTexture) are
also very simple and powerful. I remember back under DirectX5,
trying to upload a texture to hardware. I spent 4 hours trying to fig-
ure out the code and rebooting my computer as it repeatedly locked
up. I spent less than 15 minutes figuring out OpenGL texturing
and implementing it. Comparing both API’s texturing approaches,
OpenGL’s texture management is absolutely the right way to go.

– It’s extremely foolproof–you call 3 textures to create, upload,

and delete textures. Nothing can possibly go wrong!

3.1. JAN 01, 1999

Tim Sweeney Archive 46 News

– There is no bizarre/mysterious emulation going on. For exam-

ple, in Direct3D you can modify a texture by Lock()ing it and
writing to ”video memory” directly. But, on some hardware
(like the Riva TNT), the hardware stores textures internally in a
”swizzled” format you can’t access directly. So some emulation
thing has to go on behind the scenes where multiple copies of
the textures are maintained and copied around.
– OpenGL hides the inner details of texture management from
the app, enabling the driver to be optimized for whatever style
of texture management is best for a particular 3D card. In OpenGL,
a driver writer has total control over texturing and can opti-
mize the hell out of it. In Direct3D, the API hides the appli-
cation from the hardware and vice-versa, which makes it im-
possible for a game or a hardware driver to implement good
texture caching.

Here’s a link to and

• Curved Surfaces. I became familiar with a bunch of different tech-
niques. Curved surfaces pose the same kind of dilemma I felt when
writing Unreal’s software renderer: there are lots of mutually in-
compatible approaches to choose from, but no really clear favorite.
Each technique has some desirable strengths, but no technique com-
bines them all.My first experimentation was with bicubic bezier patches,
which are a great primitive for building curvy shapes without holes
between patches (as with polygons, continuity can be guaranteed
by having adjacent patches share vertices and control points). How-
ever, they require a fair amount of math for tessellation and, though
continuity can be guaranteed, it’s hard to join patches together smoothly
(with a continuous normal vector across adjacent patches), espe-
cially during animation. I’m pretty sure these will be a key primitive
in the next-generation engine. Beziers are very intuitive to model
The tensor-product math for bicubic bezier quad patches can eas-
ily be modified to enable ”bicubic bezier triangles”, which can share
edges continuously with quads. I haven’t been able to find any ref-
erences on this, but I derived the math, and it seems to work.

3.1. JAN 01, 1999

Tim Sweeney Archive 47 News

Then, a totally different approach can be used to generate smoothly

curved surfaces of arbitrary topology, using a technique known as
”subdivision surfaces”. This looks promising: approximately the
same overhead as Beziers, but a bit more general.
All of the above approaches can also be modified to support dis-
placement maps, for creating more organic geometry. While it will
be a long time before 3D hardware supports sufficient triangle den-
sities for displacement maps to be practical for fine surface detail
(such as rocks and bricks in walls), they’re very applicable to terrain
and coarse surface detail.
There are two even wilder curved techniques I experimented with
and found to be quite promising, but I don’t want to get too far off
on that tangent.
No, Unreal Tournament will not contain any curved surfaces. While
curves are cool and we’d love to have them, we’ve been totally fo-
cused on core game play, networking, and user-interface enhance-
ments for UT.
One thing that has become extremely clear in my curved-surface
rendering research is that editing tools will be a make-or-break fac-
tor in the success of next-generation level design. UnrealEd is a
pretty feature-filled polygon based editing tool, but we’re going to
need a lot more than that for the next generation. So many features
that were trivial with polygon engines–such as texture alignment,
light map placement, primitive building tools, object alignment,
and freeform object editing–become more complex when curves
are involved. The editing tools will need to be more powerful and
intuitive in order to compensate. Editing tools will be a huge focus
for us going into the next project.

-Tim Sweeney

3.1.2 Other Updates

• We updated and coolified the
htm page, which lists all the announced Unreal engine licensees.

3.1. JAN 01, 1999

Tim Sweeney Archive 48 News

• Alexander Brandon updated the


• Dave Carter’s contains lots of cool

resources for Unreal modelers and animators.

3.2 Jan 28, 1999

3.2.1 Unreal 221 Development Update

Unreal 221 development continues, with quite a lot of new features slip-
ping into this version.
Right now, Mark Poesch (the lead programmer of Wheel of Time) is visit-
ing us here in Raleigh, merging in some of the improvements the Legend
team has added into the Unreal engine for WoT and Legend’s upcoming
Unreal level pack.
I’m working on Unreal’s OpenGL support, and will soon get a version of
the code to the 3D hardware makers who have some driver issues to work
out. The Riva TNT is currently the fastest and most stable card, with the
ATI Rage 128 looking very promising, especially in 24-bit color.
I’m also extending the Unreal installer which shipped with the 220 patch,
turning it into a full-blown installation program for future patches, share-
ware versions, and retail versions. I’m also contemplating extending it to
support Unreal mods, which could then be distributed in a self-contained
format (.uzip?) which Unreal automatically knows how to install. This
would put cool mods just 3-4 mouse clicks away from users who are
browsing mod sites like, a great im-
provement over the current process of ”download a file, run PkZip, ex-
tract the .utx files to the Unreal\Textures directory, extract the .unr files
to the Unreal\Maps directory, run Unreal, then try to figure out how to
launch the mod”. This has the potential to bring mods to a much wider
While I don’t enjoy the painful and redundant task of writing an install

3.2. JAN 28, 1999

Tim Sweeney Archive 49 News

program, I see it as worthwhile because it’s such a critical piece of a game–

especially one which is supported by an online community and third-
party licensees.

3.2.2 Unreal Tournament

The rest of the team is hard at work on Unreal Tournament, and we’re
really happy with the way the game is shaping up. Unreal was such a
huge, multi-faceted project that we weren’t able to focus on deathmatch
as much as we would have liked. Unreal Tournament, with its identical
gameplay style between single-player and multiplayer (the only differ-
ence being bots vs. human opponents) has given us a far more opportu-
nity to polish the gameplay and the look-and-feel.
-Tim Sweeney

3.3 Feb 10, 1999

3.3.1 Latest News

Progress on Unreal 221 and Unreal Tournament continues. We are mak-

ing changes and improvements at such a fast pace that it has been hard/impossible
to stabilize the code for the 3-5 days required to fully test a new patch. I
apologize for the earlier false start we had with 221.

3.3.2 What we’ve been programming

I’ve been juggling work on the installer, OpenGL, and some core Unre-
alScript improvements. The installer is coming together, and is com-
plete in terms of installing the retail game and patches, but more work
remains to be done on uninstallation issues, and improved support for
installing/distributing mods. Some of the latest UnrealScript improve-
ments include a command-line script compiler, which can be used from

3.3. FEB 10, 1999

Tim Sweeney Archive 50 News

any programmer’s editor (or Visual C++) with standard output of error
messages; UnrealScript execution optimizations; and support for dynamic
UnrealScript’s string support has been pretty solid all along, as a result of
using BASIC style functions (Left, Right, Mid, InStr, Chr, Asc...) instead of
C’s error-prone string handling functions (strcpy, strcat, strcmp, sprintf).
The one annoying limitation, which has been the source of some bugs
in the past, is that strings were fixed length. 221 supports dynamically
sized strings with no practical size limitation (just available memory).
This makes it easier to write text-processing routines, which the new in-
game server browser is really stressing.
Steve Polge has been improving the AI and the flexibility of the game
code. Because Unreal Tournament’s gameplay can be far more complex
than Unreal 1’s (from a bot’s perspective) as a result of team play and Cap-
ture The Flag, Steve has been teaching the bots a lot of new techniques
for coordinating together and better understanding the environments.
It’s sooo cool watching Cliff and Myscha playing CTF with a few bots on
their teams.
Jack Porter is hammering away on the new user interface code and the in-
game server browser. In the past, first-person action games haven’t been
known for having good user interfaces. We’re hoping to change that, and
go for the quality level of an EA Sports interface. Sticking to our general-
purpose roots, Jack isn’t just creating user-interface code for Unreal Tour-
nament, he’s creating a flexible, object-oriented UI framework which will
be very suitable for licensees and mod makers.
Brandon Reinhart is working hard, implementing a lot of the Unreal Tour-
nament game code, the single-player progression, stats tracking, and much
more. The stats tracking is going to be a very cool feature for Internet
play; perhaps I can convince Brandon to write a note about it here.
Erik de Neve, who created Unreal’s algorithmic fire and water effects and
optimized much of the rendering code, is working on a next-generation
project that will make its debut in Unreal 2.

3.3. FEB 10, 1999

Tim Sweeney Archive 51 News

3.3.3 Linux Notes

We now have a command-line (windowless, console based) script com-

piler and server for Unreal, which has minimal dependencies on Win-
dows. This is a first step towards a Linux port, and it also runs very well
under Wine, with minimal overhead. This might be a good interim so-
lution for Linux server administrators, before the full port is done. I just
noticed in GameSpy that the top server on the list is titled ”CHiX Unreal
NT Server (220) PORT UNREAL TO LINUX”. Hint taken. :)
We actually have the Unreal 221 Core source compiling under EGCS 2.90.29,
C++ templates and all. Getting the thing linking and running successfully
will require a lot more work, which we won’t be able to focus on until UT
ships, but the experience thus far has been very positive.
-Tim Sweeney

3.4 Mar 19, 1999

3.4.1 Latest News

We’ve been going through many iterations of internal testing and im-
provements on the upcoming patch, which is now known as 223. Here is
the latest The story be-
hind the version number is that each time we make a new version for
internal testing, we give it a new version letter (for example, 221a, 221b).
Between the previously released 220 patch and the current one we’re work-
ing on, we’ve gone through more than 52 internal versions. We’ve beta-
tested two release candidate patches, but each one has required more
work. I don’t have a release date for this patch yet, but we’re working
hard to get it finished.
One unfortunate side-effect of the recent UnrealScript improvements (dy-
namic strings, multi-skin support, other engine enhancements) is that
223 will have to break mod compatibility with previous versions, and re-
quire at least a recompile (and code changes in many cases). We hate

3.4. MAR 19, 1999

Tim Sweeney Archive 52 News

breaking mods, because we understand how much pain it causes for mod
makers and the community to get back in synch with the latest version,
but sometimes it’s a necessary evil, as a side-effect of progress.
One bonus we’re putting into 223 to help compensate for the pain and
suffering is Jack ”Mek” Porter’s windowed GUI code, the high-level Unre-
alScript windowing framework we’re using for Unreal Tournament. This
includes the graphical ”server browser” for finding good servers to play
on – sort of an in-game GameSpy. The windowing code will give the com-
munity a chance to experiment with windowing well before Unreal Tour-
nament ships.

3.4.2 Level-Of-Detail Texture Mapping

I finally got the chance to write a rendering optimization I’ve been think-
ing about for a long time now, level-of-detail texture mapping.
Unreal’s biggest performance problem in OpenGL has been texture man-
agement overhead, which causes frame rates to vary a lot (for example,
jumping from 30+ fps down to 15 fps for a frame or two when new tex-
tures come into view). This is especially noticeable on on the Riva TNT,
which has a very good fill rate and can handle large polygon counts, but
is around 4X slower transferring textures to the card than Voodoo2.
Solution: I create redundant versions of textures, scaled down 4X (in
memory usage), 16X, 64X, and 256X. When an object being rendered is
far away, I use a lower resolution version of the texture – which doesn’t
cause visual problems (since the texture would be mipmapped anyway)
but requires less texture data to be transferred.
This has enabled us to boost the Unreal Tournament player models’ tex-
ture usage to 256K of texels per model. More info about additional level-
of-detail features later...

3.4. MAR 19, 1999

Tim Sweeney Archive 53 News

3.4.3 Future of Programming Languages

Lately, I’ve been doing research on programming languages, in the inter-

est of gaining new ideas to implement in the successor to UnrealScript
for our next-generation (after Unreal Tournament) engine. Some of the
topics I’ve been formulating are:

• Garbage collection (see the

html for background). UnrealScript uses a simple recursive mark-
and-sweep garbage collector which is only executed when chang-
ing levels, along with some special-case code to garbage-collect de-
stroyed actors during gameplay. Most implementations of Java use
a background thread to run a generational garbage collector. Java’s
approach is more general, but can cause pauses during execution
as large amounts of garbage are purged, which is bad for frame-rate
Java garbage collection presents a problem for object finalizers (the
equivalent of C++ destructors), because the finalizer is executed at
some random time after the object in question effectively becomes
garbage, and finalizers of multiple objects are executed in random
order. This limitation makes finalizers dangerous and not terri-
bly useful. One work-around is to not support finalizers. Another
work-around is to support them, but publish a list of ”things that
aren’t safe to do from within a finalizer” (such as cause finalized ob-
jects to become rooted again, make assumptions about finalization
order, etc). UnrealScript’s current solution is a mix of the two; final-
izers must be native (implemented in C++) and are subject to those
safety guidelines.
Another general conceptual problem I have with garbage collection
is that, often, the physical lifetime of an object (as defined by the
garbage collector) often exceeds the ”practical” lifetime of an ob-
ject (as defined by the semantics of the program you’re writing). For
example, a window object in a windowing system is only meaning-
ful when the window is open on screen; after the window has been
closed, the object becomes meaningless even though it still exists
and other objects might still be referring to the window. The general
problem is that programmers often would like a ”way of explicitly

3.4. MAR 19, 1999

Tim Sweeney Archive 54 News

destroying an object and removing any other objects’ references to

it”. C and C++ avoid this problem by making all memory manage-
ment the program’s responsibility. Java avoids it by making the pro-
gram responsible for determening whether an object you’re refer-
ring to is meaningful (for example by calling Window.IsOpen()). It
would certainly be nice if object lifetime management and notifica-
tions became the full responsibility of the programming language,
enabling programmers to instead focus on higher-level problems.
The garbage-collection strategy that would be ideal, from the stand-
point of predictable code and memory usage, is to finalize and de-
stroy objects the exact moment they become garbage, i.e. when
the last variable still referencing it changes. I’ve tried, very unsuc-
cessfully, to find efficient approaches to manage such as system
by tracking multiple reference counts, doubly-linked lists of two-
way references between objects, etc etc etc, and haven’t been able
to find an approach more efficient that performing a brute-force
mark-and-sweep pass every time a reference to an object changes
(plus optimizations such as reference counting to collect objects
without cyclic references). Stated in graph theory, the question boils
down to ”Given a graph and a particular directed edge of the graph,
does removing that edge break the graph into two disjoint pieces?”
Does anybody know if there is a graph-theory algorithm for deter-
mining this, which enables you to add a directed edge to the graph
in constant time, and also remove an edge and determine disjoint-
edness from that removal in constant time?

• LISP defines a new kind of object called a ”function closure” which

basically represents a pointer-to-a-member-function with several
of its parameters bound to specific values (for example, binding
the implicit ”Self”/”this” parameter to a particular object. This is
a great object-oriented generalization of function-pointers, but the
syntax becomes fairly complex in a statically-typed language (some
issues you run into are recursively-defined closure types, and cyclic
closure types). Microsoft’s Java extensions in Visual J++ include a
simpler and more limited type of object called a ”delegate” which is
a pointer-to-a-member-function bound to a specific instance of an
object. I plan to experiment with these possibilities and implement
”something along these lines” in the next engine.

3.4. MAR 19, 1999

Tim Sweeney Archive 55 News

• Inner classes, as defined by Java, are a great feature for attaching

objects together hierarchically. This is definitely a feature I want for
the future. One example area where inner would simplify and gen-
eralize the code is Unreal’s actor lighting support. Currently, the
Actor class has a bunch of variables related to lighting (LightBright-
ness, LightHue, LightSaturation, LightEffect, etc), and an enumer-
ation LightType that describes the light’s overall mode of opera-
tion. While this is an improvement over past engines, and general
enough for 1998-1999 games, it could be much better. Actors could
instead reference a separate Light object which describes the light-
ing in a more general way. Instead of having a hardcoded enumera-
tion for light effects (such as LE TorchWaver, LE Cylinder, LE Spotlight),
all the different light types could be represented by separate classes
(TorchWaverLight, CylinderLight, SpotLight) with the class supply-
ing the appropriate spatial and volumetric lighting math.
Additionally, the lighting objects could be made hierarchical, so you
could modulate a cylinder light by a spotlight, and add a torch wa-
vering effect on top. New light classes could be defined modu-
larly to supply new lighting functions without modifying the engine

While my plan for the next scripting language is gaining more clarity,
there are still a lot of wide-open issues. For example, whether Java-style
interfaces are necessary (since most of the same functionality can be ob-
tained from inner classes), what the final garbage-collection strategy is,
how similar do we want to make the language to Java, etc.

3.4.4 Visible Surface Determination

One of the biggest next-generation challenges for 3D programmers will

be in solving the ”visible surface determination” problem for complex
scenes. Doom basically used a densely portalized BSP scheme; Quake
used a BSP in conjunction with precomputing the visibility from each
BSP leaf in the world to every other leaf, trading off precomputation time
in exchange for rendering performance. Unreal uses a sparsely portal-
ized BSP tree and performs a software span-buffering pass to determine

3.4. MAR 19, 1999

Tim Sweeney Archive 56 News

which polygons are visible prior to actually rendering them. None of

these approaches are clearly superior to the others; they each have strengths
and weaknesses. Unreal’s tradeoff enables us to render more complex
multi-pass surfaces (very high-res textures, detail textures, light maps,
and fog maps) but limits our polygon counts, since rendering perfor-
mance grows approximately as ”n log n” with polygon counts, compared
to Quake’s basically linear ”n” scaling.
The visuals of Unreal Tournament vs. Quake 3 Arena will provide a very
useful equal-footing comparison between the two techniques, with id
Software is pushing the limits of polygon counts, and us pushing the lim-
its of texture detail.
After this generation of games, I think we’ll all have to fundamentally
change the way we approach the visibility problem. As 3D worlds grow
larger, more seamlessly interconnected, and more vibrant in terms of dy-
namic and procedural geometry, we’ll need more of a no-tradeoffs ap-
proach than precomputed visibility or BSP’s. Portal and anti-portal clip-
ping schemes look promising. Hardware-assisted visibility tests (such as
test.txt) look promising. Even the good old DFE scheme becomes more
practical with increased 3D hardware performance. After Unreal Tourna-
ment, I’ll be experimenting again with all of these techniques.

3.4.5 Commentary on Cool Stuff

I just wanted to point out the following cool technical things.

• OpenGL
fragment_lighting.txt and
Version1.2/EXTspecs/light_texture.txt extensions. These exceed-
ingly clear, general, and well thought-out rendering capabilities are
going to be the next quantum leap in 3D hardware acceleration ac-
celeration, enabling true Phong/Blinn bump mapping and a tremen-
dous variety of other features. My reaction to seeing this spec for
the first time is reminiscent of first experiencing 3dfx’s Voodoo ac-
celerator and the Glide API–it’s ”I can’t believe how powerful yet

3.4. MAR 19, 1999

Tim Sweeney Archive 57 News

simple this is to work with”.

• This is the first

generally useful new GUI control I’ve seen in years. Imagine the
possibilities for using hyperbolic trees to browse the actor lists and
server interconnectivity of huge interconnected online worlds.

• A new, unpatented

approach to data compression; trivial implementations can easily
beat PKZIP.


3.5 Apr 14, 1999

3.5.1 3dfx Voodoo3 Totally Rocks!

I just picked up a carload of the new Voodoo3 3000’s; we’ve put them to
good use, as the team is now testing the upcoming 224 patch.
The performance and graphical quality of the Voodoo3 are simply amaz-
ing. Though we’ve had a board for over a month, beta drivers and pre-
release board problems prevented us from realizing just how far ahead of
the pack this hardware is.
On the PC, most technological improvements come in little incremental
improvements, such that we seldom have a chance to experience a sin-
gle major leap. Back when 3dfx introduced the Voodoo 1, that was one
of those rare leaps. Since I’ve been off working on ”OpenGL land” on the
Riva TNT, ATI Rage 128, and other accelerators, I haven’t had a 3dfx card
in my machine since soon after Unreal shipped. So, for me at least, get-
ting a Voodoo3 and being able to play at those ultra high resolutions at a
great frame rate, is another one of those leaps.
While the Voodoo3’s fill rate is outstandling, where the card really clob-
bers all others is its texture management performance. This is a very

3.5. APR 14, 1999

Tim Sweeney Archive 58 News

important characteristic, because it determines how smooth the perfor-

mance is from frame-to-frame. Unreal, though it shipped nearly a year
ago, still pushes texture limits harder than other 3D action games, and
Unreal Tournament pushes them even harder.
Unreal 224 performance is also being helped by some new lighting opti-
mizations, and a major new engine feature, which Erik de Neve (Unreal
optimization and algorithmic texture guru) has been working on. But I’m
not allowed to talk about that since Mark Rein is writing a press release
about it now.
Note: Unreal 220 is incompatible with Voodoo3’s due to some changes
that were made to Glide; people who have Voodoo3 2000’s will need the
upcoming 224 patch.
-Tim Sweeney

3.6 Apr 15, 1999

3.6.1 Important Unreal 224 Note To Mod Authors

Unreal 224 .u files aren’t backwards compatible with past versions of the
engine. So you’ll need to take special steps to upgrade your .u files.
To upgrade mods to 224, you’ll first need to export your scripts to *.uc
files. You can do this in all versions of UnrealEd like this:

1. Bring up the UnrealEd class browser on the right side of the screen.

2. Click on the ”Load” button and load your mod’s .u file.

3. Click on the ”Export All” button and have UnrealEd export all of the
loaded scripts to .uc files (including yours and the engine’s), creat-
ing a directory structure for all the .uc files: for example, the files
from the Core package are in \unreal\core\classes\*.uc. If your
mod is named XYZ, your .uc files will be exported to \unreal\XYZ\classes\*.uc.

3.6. APR 15, 1999

Tim Sweeney Archive 59 News

In Unreal version 221 and earlier, you can rebuild your mod named ”XYZ”
from the command line by typing:

cd \unreal\system
del XYZ.u
unreal -make

You can do that now, prior to the release of 224, to verify that your mod
has been safely exported and can be rebuilt from the command line. I
highly recommend doing this now, because if you install 224, you will no
longer be able to load your old .u files, so you’ll have to downgrade back
to 220 in order to export them. Some developers prefer to edit scripts in
UnrealEd, but many of us actually edit scripts in an external editor (like
Visual C++) and use the command-line interface for all production work.
In Unreal 224, we have a new command-line compiler with a slighly dif-
ferent syntax:

cd \unreal\system
del XYZ.u
ucc make


3.6.2 Maps, Textures, Sounds, Music are backwards com-

patible there shouldn’t be any problem with existing user maps built with
past versions of UnrealEd. I checked this by downloading a bunch of
maps from and http://www.planetunreal.
com/nalicity/ and testing them. The one ”gotcha” is that some user
maps being distributed contain .u files, and those aren’t backwards com-
patible, so they will refuse to load until the .u files they depend on are

3.6. APR 15, 1999

Tim Sweeney Archive 60 News

3.6.3 Latest News

I’ve been awake for way too long, so please forgive my spelling and gram-
Internal testing of 224 continues. Things are looking pretty good, though
we need to go through a couple more iterations before everything’s in
perfect shape.
The has a good article on the latest http://avault.
com/news/displaynews.asp?story=4141999-221754. They spill the beans
on the major new feature I alluded to below:

Another major enhancement to be included in 224 is continu-

ous level of detail,which varies polygon counts on objects such
as player characters and power-ups relative to their distance
from the camera, or player’s perspective. In Unreal Tourna-
ment, this will enable the engine to place more characters and
objects on the screen and still maintain a consistent and fluid
framerate. Rein said this is an important feature not just for
Unreal and Unreal Tournament but also for engine licensees
since it also opens the engine to new game types, such as real-
time strategy titles, that were not previously obvious choices for
the engine.

3.6.4 Cool Stuff

While I wait for 200,000 lines of Unreal code to recompile, I wanted to

point out some more cool things that seem to be under-appreciated.

• is a great MP3 music player for Windows;

and its companion, is a streaming ”In-
ternet radio” companion. While I know MP3’s and Internet radio
has been around for a while, WinAmp really deserves credit for its
grassroots following and overall coolness factor.

3.6. APR 15, 1999

Tim Sweeney Archive 61 News

• Windows 2000 internationalization support. This pure Unicode OS

will mark the first time Microsoft ships a single binary worldwide,
which will be a great leap forward in making it easy to develop games
(and applications, and websites...) for the international market.

• 2400 bits-per-second phone-

quality voice compresion. Speech compression research has been
making major leaps and bounds lately, as quality voice compres-
sion ratios skyrocket. The latest set of algorithms analyze an audio
stream and (to simplify things a bit) transform it backwards to vocal
cord excitation. The resulting compression ratio is in the neighbor-
hood of 100-to-1. This is a research subject Carlo Vogelsang (Unreal
sound engine programmer) and I have been looking into lately, and
it turns out to be very complex and math intensive.


3.7 Apr 19, 1999

3.7.1 Getting closer...

But we’re not quite there with 224 yet. The guys have been beating on
internal test versions for the past few days, and we’ve been polishing the
code. It’s both a blessing and a curse that the Unreal code is evolving
by leaps and bounds, gaining major new features all the time. This ap-
proach makes testing each new patch a major effort. Though, 224 will
be especially worth the effort because of the in-game server browser and
level-of-detail optimizations.
224 will probably happen late this week.
Here is a where
Mark Rein shamelessly plugs the patch.

3.7. APR 19, 1999

Tim Sweeney Archive 62 News

3.7.2 Network Cross-Compatibility

(warning: nitty-gritty technical info)

An interesting technology feature in 224 is a new method of evolving .u
files in a way that is backwards-compatible for network play.
One of the things that has always bothered me about past versions of
Unreal is that clients and servers had to have exactly-matching .u files in
order to play together. This has meant that every recent Unreal patch has
broken network compatibility, leaving some players orphaned. The rea-
son behind this is that Unreal’s network code (described in detail http:// uses the positions of objects within
.u files as a ”common point of reference” which clients and servers use
to map objects between each others’ address spaces. Some scheme of
this nature is necessary when network code is generalized like Unreal’s
is, with the same scripts executing on the client and server, passing ar-
bitrary data back and forth, with the network code transparently coordi-
nating the game world.
I’ve been aware of some solutions to this problem all along (such as ”send
all object names as strings”), but they all sacrificed bandwidth.
Recently, I found a no-compromises way to enable .u files to evolve incre-
mentally without losing compatibility. Now, when we release a new ver-
sion, we copy the .u files into a new developers-only subdirectory \Unreal\SystemConform.
Then we can make minor changes to our scripts and recompile new ver-
sions that are network-compatible with the old ones. When recompiling
and saving a .u file such as \Unreal\System\Engine.u, the script com-
piler now examines the existing file \Unreal\SystemConform\Engine.u,
and makes the new version ”conform to the layout of the original”. In
addition, when a client connects to a server, they both negotiate to fig-
ure out who has the earliest version of each .u file, and both talk about
objects ”in relation to that old .u file”. This enables compatibility without
any loss in performance or bandwidth usage.
Really major updates will always break code assumptions and require
downloading an update. This approach just makes minor incremental
updates possible, enables us to test spot-fixes on the net, and generally

3.7. APR 19, 1999

Tim Sweeney Archive 63 News

makes our programming lives easier.

3.8 May 01, 1999

3.8.1 224 Progress Continued

The team found about 10 bugs in Unreal 224t that are significant and
merit fixing before release, so I’m just wrapping up fixes for them now,
and we’ll have another release candidate (224u) in internal testing this
afternoon. Thanks for the patience...we’re almost there. posted a huge
screenshots/ut/utshot1.shtml of a bunch of Unreal Tournament char-
acters. These are the actual meshes (same poly count as the game), but
textures of this resolution will only be available on cards that support
S3TC texture compression such as the Savage4 (though other cards will
have support soon). Current 3D cards will have to make due with four
256x256 textures per player. Our new skinning method enables player
meshes to be mapped with four textures, so we can have separate faces
and armor. This Ultima Online inspired approach lets players customize
their look more than past 3D action games.
-Tim Sweeney

3.9 May 02, 1999

3.9.1 STAT NET: Diagnosing your connection to the server

When playing a multiplayer game of Unreal, press TAB and type STAT
NET to bring up the network stats display. You’ll see something like this:

137 PING

3.8. MAY 01, 1999

Tim Sweeney Archive 64 News

1395 2031 BYTES/SEC
2600 2600 NETSPEED

The numbers on the left represent incoming statistics (for example, how
many packets from the server to you were lost), and the right outgoing
statistics (for example, how many packets from you to the server were
lost). There interesting stats are:

• PING: Round-trip time between you and the server, in milliseconds.

Modem users will typically see 200-350 for in-country gameplay.
ISDN/T1 users will be in the 50-200 range. Playing overseas results
in enormous times, 600-1000 or more. This number represents the
quality of the connection between you and the server, and really
should be 300 or less for good gameplay performance. A bad con-
nection on either end can hurt the ping time.
• CHANNELS: Roughly, the number of objects that the server is ac-
tively updating to you. Proportional to how much combat is hap-
pening around you.
• PACKET LOSS: Percentage of packets lost between the client and
server. The higher the packet loss, the worse the performance will
be, and the more you’ll see objects skipping around rather than
moving smoothly. High packet loss can be caused by:
– Using too aggressive a network speed on the client side. Try
typing ”NETSPEED 2600” on the console to reign in bandwidth
– The server’s connection is overtaxed. This will result in heavy
packet loss and a bad gameplay experience for everyone. Servers
will limited bandwidth available need to set their play limit
conservative (expect 5000 bytes per client per second), and also
lower the MaxClientRate variable in Unreal.ini. A value of 5000
should be good for everyone.

3.9. MAY 02, 1999

Tim Sweeney Archive 65 News

– A bad connection anywhere between you and the server.

– A bad connection between you and your ISP.

• PACKETS/SEC: Number of packets being sent per second. The num-

ber on the left should be 13 or higher. If it’s significantly less, that
means the server is either running too slowly, or experiencing high
packet loss. Gameplay will be bad at 10, and unbearable at 5. The
number on the right is proportional to your frame rate and you
don’t need to worry about it.

• BUNCHES/SEC: Number of objects being updated per second. Not

important for tweaking.

• BYTES/SEC: Bytes being sent per second.

• NETSPEED: The network speed setting you set using the console
NETSPEED command. Recommended settings are:

– 2600 for a typical modem connection.

– 2400 for a poor modem connection.
– 5000 for an ISDN connection.
– 20000 for a T1 connection.

3.9.2 Server Alert

Unreal 224 servers are popping up very rapidly, and becoming filled to
capacity rather quickly. From the last couple hours of cruising the net
and playing on various servers, we’re seeing wildly varying performance.
As with all games, the quality of Unreal netplay is a function of how well
the server is performing and how good your connection is (bandwidth,
packet loss, and latency).
If you’re having a bad experience, please try out a few other servers. This
can affect performance hugely. In my game sessions here, I had some
great performance with a 300 msec ping and 28.8Kbps bandwidth, and
some astronomically bad performance with a 100 ping and 50Kbps band-
width. The bad performance fell into two categories.

3.9. MAY 02, 1999

Tim Sweeney Archive 66 News

Some servers have more users than bandwidth available, so they are drop-
ping packets, experiencing 50%+ packet loss and escalating ping. Some
are just running very slowly, around 5 fps (I’m not sure why, perhaps
other processes were running?). This tends to happen whenever we re-
lease a patch, as people rush to get into gameplay, flooding the servers
that have upgraded, users setting up servers on cable modems out of
desperation to get into a game... It can get chaotic. In the future, the
new backwards-compatible network code should mean there are always
lots of servers available when a patch is released.
So, bottom line: Please don’t bitch if netplay performance is erratic over
the next couple days as servers get set up and stabilized. If netplay per-
formance sucks after a couple of days, THEN commence bitching. :-)

3.9.3 Unreal 224v C++ headers released!

For hardcore mod authors who want to experiment with Unreal’s C++ in-
terface, here it is! This code requires Microsoft Visual C++ 6.0 with no ser-
vice packs or SP1. Read the
htm for more information and distribution terms.
Now it’s time for a disclaimer about the C++ code. For mod authors, Un-
realScript is the best way to write game code. The fact that 100% of the
Unreal 1 and Unreal Tournament game-specific code is in UnrealScript
illustrates our strategy here, that UnrealScript is a fully-featured language
designed to simplify game programming compared to C++. The C++
code is significantly more complex than UnrealScript, and is only neces-
sary for programming things that just aren’t possible under UnrealScript,
such as algorithmic textures, performance-critical AI decision making,
and that kind of low-level work. Furthermore, our distributing just the
headers (and not the C++ engine source), means that many key aspects
of the engine are hidden in implementation details, so it’s going to be
hard to understand what’s happening ”under the hood”. Another major
downside to C++ is that C++ mods will break with almost every upcoming
patch, whereas UnrealScript is more stable (having broken only 3 times

3.9. MAY 02, 1999

Tim Sweeney Archive 67 News

in 11 months and approximately 12 public betas released). Summary:

This C++ code is for very, very hardcore mod makers only.
That said, there is some cool stuff here. I’ve included all of the core and
engine header files and .lib link libraries in this release, so you can com-
pile your own DLL’s and link them to Unreal. I’ve also included the C++
code for some modules which we don’t feel are confidential: Unreal’s
OpenGL support layer (OpenGLDrv), the engine startup code (Launch),
and the installer/uninstaller (Setup).
The OpenGL code will be of interest to OpenGL driver writers who want
to see what we do in Unreal, in order to optimize their drivers, and pos-
sible to enthusiasts who want to experiment with rendering.
We will probably be able to publically release the Direct3D driver code
soon too, also to aid driver writers and experimenters, but our Direct3D
development partner is going to first spend some time cleaning up that
This code is provided as-is: we don’t have much time for answering ev-
eryone’s questions about it, so you’re mostly on your own.
The Unzip it into
your \Unreal directory, and the proper subdirectories will be created (de-
scribed in the The C++
code compiles there and links to DLL’s in the \Unreal\System directory,
read to run. Enjoy!
-Tim Sweeney

3.9.4 Unreal 224v Released to the public!

The Unreal 224v patch is now available for download from our http://, along with the release notes.
-Tim Sweeney

3.9. MAY 02, 1999

Tim Sweeney Archive 68 News

3.10 May 03, 1999

3.10.1 Known Bugs

Having a problem with Unreal? Please visit the http://unreal.epicgames.

com/Bugs.htm page to see if we know about it and are addressing it, and
if there any any workarounds.

3.10.2 Future Patch

There will be another Unreal 1 patch. It will be 225. It will contain fixes for
the significant bugs reported that are under our control (3D card drivers
are dependent on the hardware makers). By popular demand, our focus
on future Unreal 1 patches will be making bug fixes and small, incremen-
tal improvements. So, we’ve split off the Unreal 1 code base from the Un-
real Tournament code base, and won’t be major new features added to
the Unreal 1 code base, just fixes.
This code split gives the server community some room to get back to-
gether and grow healthily. The server community was fragmented in
the past because of our Unreal 224 delays, at one point having 4 ma-
jor versions available (220 for the PC, 219 for the Mac, 222 for U.S. 3dfx
Voodoo3 users, 223 for European 3dfx Voodoo3 users. From here on, we’ll
keep network compatibility between versions, and keep code changes to
a minimum. And we’ll keep Unreal 1 script compatibility (but not UT
script compatibility).
The split also gives us more flexibility in releasing future patches. In the
past, we’ve been adding so many new features to the engine for Unreal
Tournament, some necessitating architectural changes, that we’ve had a
long internal testing cycle for each patch. From now on, the Unreal 1
changes will be simple and localized.
The split is also something our Unreal tech licensees have wanted. From
here on, our partners who want to ship games soon (before or soon after
Unreal Tournament) will prefer to stay on the 224 track to gain stability
and minimize changes, and those with further-off products may want to

3.10. MAY 03, 1999

Tim Sweeney Archive 69 News

move forward with the Unreal Tournament code track, to keep up to date
with the latest features.

3.10.3 Check out the Unreal news sites

• has lots of news and tips about 224.

• is a news site dedicated to the Un-

real engine.

• is a master index of Unreal resources on

the Internet, and has tons of cool Unreal related links.

I’ll be making a summary of reported bugs, workaround, and related

feedback from the community later tonight.

3.10.4 OpenGL UnrealEd

Thanks for pointing out this feature I added

months ago but basically forgot about. :)
Though it theoretically works, the consumer OpenGL drivers I’ve tried
aren’t yet stable enough for it to work reliably on real-world hardware.
But for people who like experimenting, if in-game OpenGL support works
flawlessly, you can try editor OpenGL support by changing the Windowe-
dRenderDevice= line in Unreal.ini to WindowedRenderDevice=OpenGLDrv.OpenGLRenderDe
Problems I’ve experienced on my Riva TNT and ATI Rage Pro (2 months
ago) include crashing in the display driver when resing windows, failure
to create HRC’s when opening windows, and texture corruption.
If there is enough demand, perhaps we can pressure 3D hardware mak-
ers to test with UnrealEd. Hopefully this will be easier now that we’ve
released Unreal’s OpenGL source code to the public.

3.10. MAY 03, 1999

Tim Sweeney Archive 70 News

3.11 May 05, 1999

3.11.1 Maps that broke since 220

Some user maps such has broke with version 224. This was caused by a little bug
in the file backwards-compatibility code, that I’ve fixed for 225. Map au-
thors don’t need to do anything, this will automatically be fixed with 225.
The specific problem I fixed was an exit with an assertion:

Assertion failed: Field [File:..\Inc\UnClass.h] [Line: 139]

If there are any user maps which don’t require .u file mods, and the maps
refuse to load with a different error message than this, please email the
map (or a URL where we can download the map) to mailto:unreal224@

3.11.2 Quick Notes

We’ve been going through the Unreal 224 bug reports and addressing lots
of issues people have pointed out. Here’s the latest status.

• Creative Labs EAX 3D sound support is fixed for 225.

• We’re working with Aureal to get A3D 2.0 support into 225, which is
very likely to happen.
• I fixed the cause of the ”out of memory” errors that tended to hap-
pen when the console was pulled down on some machines. There
was a problem in UnrealScript with switch statements operating on
strings, which caused this and some similar problems reported by
licensees. Thanks to Mark Poesch of
for tracking this one down.

3.11. MAY 05, 1999

Tim Sweeney Archive 71 News

• Under some conditions, alt-tabbing between fullscreen and win-

dowed play would cause the in-a-window game window to be col-
lapsed to zero size. Fixed.

• I optimized the network visibility code by about 25%. Server ad-

mins are reporting notably more CPU usage per player with 224
than 220. I haven’t tracked this down, but I’m going through a lot
of code to measure and improve the server performance.

Many other fixes, tweaks, and incremental improvements are in the works.
See the page for more info.

3.12 May 08, 1999

3.12.1 Latest progress

All the known server-crashing problems in 224 have been fixed and are
in internal testing now. Stability has improved quite a bit. There is still
significantly higher server CPU usage in 224 compared to 220 on heavily
populated servers, and I’m still tracking this down.
The Direct3D code has been cleaned up and optimized a bit, giving a few
FPS improvement on the Riva TNT, with support for detail textures now
enabled. 225 will include the new D3D driver, and the 225 public source
release will include the source.

3.12. MAY 08, 1999

Tim Sweeney Archive 72 News

3.13 May 19, 1999

3.13.1 Unreal 225 musings

The recent work on 225 has been focused on improving server perfor-
mance and stability, and it has been paying off really well. It turns out
that 224’s server code was noticeably slower than 220, as a result of some
code improvements I made but didn’t test and time rigorously. The two
bottleneck routines in the server are the visibility tracer and the actor
replicator. I’ve been making the visibility tracer faster by processing all
actors simultaneously and optimizing the code for a pure ”yes/no” vis-
ibility test, rather than using the more general (and slower) line trace
routine built into the engine. The actor replicator has gained signifi-
cant speedups from a combination of low-level optimization (to the bit-
stream writer, and replication condition caching code), and a new time-
stamping system that speeds up replication by up to 50% on servers with
high player counts.
Now, the 225 server is faster than all previous versions, and I’m working
on some further improvements over the next few days.
As many server admins have pointed out to us, if we can double Unreal’s
server performance, admins will be able to run twice as many servers on
their machines!

3.13.2 ”Make Something Unreal” Contest

Check out the site for the latest info and
news on the contest for Unreal map makers, mod authors, artists, and TC

3.13.3 Things That Are Cool

• Anybody who is programming
on a large project without using a source-control system is nuts!

3.13. MAY 19, 1999

Tim Sweeney Archive 73 News

SourceSafe provides a central repository for all code, and enables

programmers to check out files and work on them without their
changes ”colliding” with those of other programmers. SourceSafe
also tracks all changes to all files since the beginning of a project,
which has been incredibly useful in developing Unreal, because it
enables us to examine our past code changes when tracking down
bugs. Finally, SourceSafe is excellent for managing ”branched”
projects: when we split Unreal 224 apart from the Unreal Tourna-
ment codebase, SourceSafe managed the files’ shared history auto-

• Now that Microsoft has publicly com-

mitted to dropping the ball on 3D hardware driver distribution with
DirectX7, GLsetup is the only hope for game developers who want
their game to work on users’ computers out-of-the-box. The PC
is already plagued by driver problems, and the classic symptom
is users having to go to the web and download new drivers to get
new games working. The world would be a much better place if
Microsoft took a leadership role in driver updates, by keeping the
DirectX redistribution continually refreshed with new QA-certified
drivers, thus assuring that new games always work ”out of the box”
– if that happened, PC gaming would be a lot closer to the level
of reliability of console games. But Microsoft isn’t doing that, and
GLsetup is a grassroots effort to solve the problem.

P.S. Check out, the cool in-depth Unreal
news site.

3.13. MAY 19, 1999

Tim Sweeney Archive 74 News

3.14 May 29, 1999

3.14.1 For server administrators only: Unreal 225e Patch

This is a test version incorporating all of the latest server-side stability

and performance improvements. It’s backwards compatible with Un-
real 224 clients. If you’re running a 224 server and are experiencing per-
formance or stability problems, please upgrade. If you experience any
server problems or server crashes with 225, please email your Unreal.log
file to so we can investigate the prob-
lems. The results on our test server here at Epic indicate that this ver-
sion is significantly faster (i.e. supports more players) than 224, and is
more stable but I don’t expect it to be perfect, because I rewrote about
a thousand lines of server code to improve performance, and there will
probably be some bugs to shake out.
Only download this if you are running an Unreal server: http://unreal.
The general Unreal 225 patch for everybody including users will be avail-
able in a few days.

3.14.2 On engine licensing

In this, I think John Car-

mack hit the nail on the head on the ”build vs. buy” decision that goes
into licensing an engine:

”If you want to aim for something that’s 100 percent propri-
etary and you’re willing to take 18 months longer on the project
to develop the technology while running a significant risk of
abject failure, you make your own game engine. If you’ve been
to a few E3’s, you’ve probably seen a dozen instances of some-
body making a really, really cool demo that’s showing off some
new technical direction - then the game never comes out.”

3.14. MAY 29, 1999

Tim Sweeney Archive 75 News

Though game engine licensing was occuring over two years ago, we’re
still very much in the early days of this rapidly growing and evolving mar-
ket: Game developers are learning what an engine can and can’t do for
them, and engine developers are still learning the ropes of the business
models, engine modularity, support, and version management.
I’m happy to see the press treating engines level-headedly: everyone seem
to recognize that an engine is a useful tool that can speed development
and enable a developer to focus on game development rather than tech-
nology; but also that an engine is just a tool and isn’t a miracle solution
for game development. When an exciting new technology is evolving,
there’s always a danger of it being massively overhyped and pushed as
a cure-all solution (see Java), but engines seem to have received a fair
showing, not unrealistically positive or negative.

3.14.3 UnrealEd ”Runtime Error 20005” Experimental Fix

People who are getting the infamous ”Runtime Error 20005” or other
problems when UnrealEd is initializing, please install this http://unreal., then try running UnrealEd again.
Does this solve the problem for you?
Please email and let us know either
way. Note: This is a feedback email address, not tech support.
Webmasters: Can we please not post this on big general gaming sites?
I want to get feedback from users on the effectiveness of the fix before
widely publicizing it. I’ll post the results here in a couple of days. Thanks.

3.14. MAY 29, 1999

Tim Sweeney Archive 76 News

3.15 May 31, 1999

3.15.1 Latest News

• 4 out of 5 users have reported success with the experimental Un-
realEd fix – it made those nasty ”Runtime Error” messages go away.
If you’re having problems with UnrealEd not starting up, grab this

• The other one reported an error message relating to Threed32.ocx

which I fixed in

• There was a master server outage/problem today. If you weren’t

seeing servers in Unreal’s new server browser or GameSpy today,
don’t worry, the problem was at the master server and it has been

• Aussies, check out the for Aus-

tralian level designers!

3.15.2 Holiday?

Here at Epic HQ, we suspect today is some type of holiday. The tell-tale
sign is that the parking lot was empty when we all came to work this
morning. That usually gives it away. However, we haven’t figured out
the identity of the occasion – it seems too warm outside for Easter, and
our publisher isn’t calling us three times a day about shipping our game
so it’s probably not Christmas.

3.15.3 New 225f Patch for server administrators only

This should fix the UActorChannel::ReplicateActor crash and some smaller


3.15. MAY 31, 1999

Tim Sweeney Archive 77 News

It’s backwards-compatible with Unreal 224-225 clients. If you have any

server crashes, please email your Unreal.log file to mailto:unreal225@
Only download this if you are running an Unreal server: http://unreal.

3.15.4 Server 225e Patch Feedback

Server admins are reporting much better performance, but 90% of the
bug reports point to a crash in UActorChannel::ReplicateActor. I’ve been
fixing this (and a few other things) and will have a new server-only patch
later tonight. Thanks for everyone’s feedback!

3.15.5 UnrealEd Experimental Fix 2

Based on feedback, Saturday’s experimental UnrealEd fix solved the prob-

lems for only about 20% of users (wild guesstimate) who were previously
unable to run UnrealEd. On further investigation, we’ve found a set of
Visual Basic runtime DLL’s that are more likely to work for most users. If
UnrealEd still isn’t working for you, please try http://unreal.epicgames.
com/Files/UnrealEdFix2.exe. Please let us know what happens by email-
with a subject of ”Experimental Fix 2”.
IMPORTANT: After installing the fix, reboot your machine before running
UnrealEd. The proper files won’t be refreshed until you reboot.
Webmasters: Please only post this info on Unreal news sites. I don’t want
it to go on general gaming sites until we have feedback on patch suc-
cess/failure from users.

3.15. MAY 31, 1999

Tim Sweeney Archive 78 News

3.16 Jun 01, 1999

3.16.1 UnrealEd Follow-Up

If UnrealEd is crashing at startup and you are getting an error message

regarding ”UEditorEngine::FinishAllSnaps < - WM USER EXEC < - Vb-
WindowProc < - glxDetectOutput”, that is a separate problem that seems
to be caused by some sound card drivers.
You can avoid this crash by running ”UnrealEd -nosound”. You don’t get
any sound, but UnrealEd otherwise works fine for editing.

3.16.2 Server Cheats?

A few server admins have been reporting a cheat that enables clients
without administrative rights to change the speed of the server’s game
(as if typing ”slomo 0.5” from the console in a local game). What’s the
trick? First person to give the exact steps to duplicate the cheat gets a
free signed copy of Unreal Tournament as soon as it ships. Send reports
with a subject ”SLOMO CHEAT”. Include your shipping address.
Are people seeing any other cheats occuring on the public Unreal servers?
If so, let us know, (no prizes for this,
we’ll just fix the security holes!)

3.16.3 UnrealEd for 3dfx Voodoo3 bundle owners

The version of Unreal that shipped with 3dfx’s Voodoo3 bundle didn’t in-
clude UnrealEd.
So, we now have the
exe which installs UnrealEd.exe and all of its DLL’s into the appropriate
directories. For this to work, you must have Unreal installed, and you
must point the UnrealEd installer to the same directory Unreal is in (for

3.16. JUN 01, 1999

Tim Sweeney Archive 79 News

example c:\Unreal). Then just run c:\Unreal\System\UnrealEd.exe to

start editing.

3.17 Jun 06, 1999

3.17.1 Direct3D Improvements

Since Riva TNT users are reporting that Direct3D is more than 30% faster
than OpenGL on their cards, I’ve been spending some more time improv-
ing the D3D code, adding better support for video mode switching, and
optimizing the texture management code a bit more.
Another motivation for the Direct3D improvements is the Matrox G400.
I just received one, and it’s an awesome video card–fast fill rate, great
graphics quality, and decent texture management performance. The G400
has a great Direct3D driver, but their OpenGL is hideously bad.
One interesting new feature I managed to implement in Direct3D is re-
cursive, fractal detail textures. With this feature enabled in Direct3D, you
never see any individual texels, no matter how close you get to a surface.
It’s a really interesting contrast to the blurry bilinear filtering you see near
surfaces in most games. Considering how fast current 3D hardware has
become, there’s no reason for games to reveal any individual texels any-

3.17. JUN 06, 1999

Tim Sweeney Archive 80 News

3.18 Jun 19, 1999

3.18.1 Unreal Direct3D

Our Direct3D performance and stability have increased significantly since

version 225. For the past few days, I’ve been swapping 3D cards in and
out, tuning Unreal’s Direct3D performance on all of them. I’m not in the
benchmarking biz, so I’m just going to give my personal reaction from
playing, and say that the Riva TNT, TNT2, Matrox G400, and ATI Rage
128 are all very nice for playing Unreal now!
(Don’t ask about the Riva 128, Rage Pro, and Permedia 2, they are worse
for gameplay than Unreal’s software renderer).
Everyone’s complaint with Direct3D support in past versions of Unreal,
on good cards like the TNT, has been ”the average frame rate and bench-
mark numbers are fine, but there is major hitching and pausing during
gameplay”. This was due to several factors, which I tracked down and
fixed with help and advice from some driver writers. The key improve-
ments are:

1. Dramatically less memory usage. My Direct3D code wasted tons of

hidden, ”behind the scenes” memory while swapping textures into
video memory, leading to lots of virtual-memory swapping.

2. New texture management code, better optimized for Unreal’s tex-

ture usage patterns. I had been relying on Direct3D’s built-in tex-
ture manager, which is slowed down by its generality.

The next patch (”when it’s done”) will incorporate the new Direct3D code.
This will be out before Unreal Tournament ships, and we’ll be looking for
feedback from players on its performance and stability.
I’d like to thank Ben de Waal, Sim Dietrich, and Doug Rogers at NVidia;
Sameer Nene at Microsoft; and Eric Le at Matrox for providing cool ad-
vice and performance tips.

3.18. JUN 19, 1999

Tim Sweeney Archive 81 News

3.18.2 Summary of things I learned about Direct3D

Never ever Lock() a video memory surface. That is amazingly slow, espe-
cially on TNT cards. When I went to a pure Blt() based texture handling
scheme, my worst-case frame rate went from 8 fps to 20 fps.
Games that use large amounts of textures, palettized textures especially,
should not use DDSCAPS2 TEXTUREMANAGE. Write your own texture
manager, and optimize it around your game’s usage patterns. You can
get much better performance than using the general-purpose one that’s
built in.
Realize that DDSCAPS2 TEXTUREMANAGE makes system-memory copies
of all your textures as backing-store for the ones cached in video memory
temporarily. If your native textures are palettized and the 3D card doesn’t
support paletted textures (a very common case with Unreal), realize that
you’re going to end up with major memory waste. Unreal 225 and earlier
kept around (a) its own 8-bit copy of each texture, (b) the D3D texture
manager’s 16-bit copy of each texture, and (c) the 16-bit video memory
copy. In Unreal in some cases, this wasted 40+ megs of system RAM!
When I dumped DDSCAPS2 TEXTUREMANAGE, I went down to about
12 megs of system RAM. Better yet, a lot of cool new cards like the TNT2
and G400 have 32 megs of video memory, so you can effectively store all
your textures there, and free the system memory copies (which I do), to
bring the waste down to 0. This improves smoothness very significantly
by reducing virtual memory usage.
Realize that DDSCAPS2 TEXTUREMANAGE can potentially do evil stuff
to your frame rate, for example if it tries to free many small textures to
make room for a big texture.
(It’s totally general-purpose, so it has to handle all the bizarre inputs you
might throw at it).
Never allocate or free video-memory textures during gameplay in Di-
rect3D. This operation is slow. Do it at init time. I do this in Unreal now,
caching all textures into fixed-size bins in video memory. I swap textures
into the bins in realtime, but never reallocate the bins.
Don’t be afraid to constrain your engine, texture formats, texture sizes,

3.18. JUN 19, 1999

Tim Sweeney Archive 82 News

etc to get optimal Direct3D performance. GENERAL==SLOW, especially

when dealing with texture management, where you are already pushing
RAM and memory bandwidth limits.
Overall, Direct3D has come a long way. The API and drivers are quite
stable now. In my 100 hours of rewriting the Direct3D code and testing
it on 10+ different 3D cards under Windows 98, I had 5 lockups. This is
about the same as my experience with OpenGL on Windows NT.

3.19 Jun 30, 1999

3.19.1 Direct3D Is Happening

We’re testing 226a internally now. The results of are good so far. If no
significant ”non-driver” problems are found, we’ll have Unreal 226a out
later this week. Here is the kind of feedback we’re seeing on D3D Unreal

First client impressions, after zonking it with a couple of really

nasty, detailed, clippy, high polycount maps: incredible im-
provement! I got through DMDowntown, my personal Unre-
alRavager(tm) with nary a clip, pop or tick. Switching resolu-
tions works well now, but you have to have the latest Hercules
driver (now using 1.70, dated 6-28-99) for it to work, or you get
a critical error. ALT-ENTER works fine too, thanks Tim (now
I can get multiple screenshots without exiting and restarting
Unreal). DMBreakfast, a new map which uses a great deal of
UnrealEd rendered graphics, played very well too.
This is a quantum leap for the ’in game’ video speed on the Her-
cules TNT2. Most impressive. I’ll try to get this patch on the
Nali City game server (unreal:// tomorrow so we
can all beat it up.
-Pete, http: // www. planetunreal. com/

3.19. JUN 30, 1999

Tim Sweeney Archive 83 News

This isn’t a miracle improvement, but TNT, G400, and Rage 128 owners
should see clear speed and smoothness gains. Just make sure you get the
latest drivers available for your card.

3.20 Jul 01, 1999

3.20.1 226 Progress

Testing of 226a brought in mixed results: Direct3D performance improved

for most testers, but some had crashes and other weirdness. I’m looking
into all of this now; I think most of the problems are driver/hardware re-
lated, but I should be able to work around most of them in the code. We’ll
definitely go through a couple more internal test versions before releas-
ing 226 though.
I’m also adding a new ”wizard” based user-interface for switching 3D
drivers and changing ”safe mode” settings, to improve the user-friendlyness
of the Direct3D/OpenGL/Glide/Software switching.

3.20.2 Recommended Reading
is an excellent book for people trying to design object-oriented program-
ming languages (or just learn the theoretical background of OOP). It does
get pretty technical, defining a lambda-calculus variant in terms of ob-
jects, and analyzing OOP concepts within that framework. But it sheds
a lot of light on the general language-design problems that UnrealScript
After writing a big piece of code like the UnrealScript compiler, it’s cool to
examine some of the design decisions that I made without really know-
ing what I was doing, only to discover there’s a well-developed theoretical
foundation for what I basically hacked together until I felt it worked right.

3.20. JUL 01, 1999

Tim Sweeney Archive 84 News

For example, it turns out that UnrealScript’s type-compatibility checking

(the rules for determining whether a variable is compatible as an assign-
ment parameter or function call parameter), exactly follow the rules of
type covariance, contravariance, and invariance.

3.21 Jul 03, 1999

3.21.1 On Our World Domination Plans

Ok, the headline is exaggerating a bit. Still, there are some cool things
happening with the Unreal engine that we find very exciting–uses of the
technology that we never dreamed would exist a couple years ago.
For example, as the latest news story de-
scribes, the team is building a ”virtual recon-
struction” of the
interviews/images/vrnd1big.jpg. And it looks jaw-droppingly amaz-
The team here, busy working on Unreal Tournament, is very jealous about
the polygon counts these guys are using (the lucky bastards don’t have to
worry about the CPU impact of 12 players deathmatching in their cre-
Also, the rapidly-growing independent news site, http://www.unrealengine.
com/, is covering lots of other Unreal engine projects that fall way out-
side the ”first person action” stereotype, such as Hasbro’s http://www. (see
news/displaynews.asp?story=6301999-102335) – as well as the other well-
known projects from our
There are other interesting things happening with the tech, behind the
scenes. A number of schools (ranging from high schools to universities)
are using UnrealScript to each real-time object-oriented programming.
Several university masters-degree thesis projects are centered around projects
involving the engine. Some architects are using the editor to generate
real-time architectural walkthroughs. At least two Silicon Valley startups

3.21. JUL 03, 1999

Tim Sweeney Archive 85 News

are pitching completely non-game projects incorporating the Unreal en-

gine to investors.

3.21.2 What’s really happening here?

What we’re seeing is part of a larger trend that will continue to grow. Re-
member ten years ago, when PC’s were very weak compared to the state-
of-the art: SGI workstations, and the Unix workstations from HP, Sun,
and Apollo?
Now, a $2000 Pentium III PC with a Voodoo3 or TNT2 card eclipses the
performance of a $30,000 SGI for real-time rendering. The CPU is faster,
the fill rate is faster. When I first saw an SGI Reality Engine, my impres-
sion was, ”Holy cow, I can’t believe how much better this is than my ’286!”
But nowadays, the best 3D games look far cooler than anything you see
running on a Reality Engine.
That entire segment of the high tech industry, the one containing non-PC
workstations, SGI’s high end hardware, simulation tools vendors, and so
on, is falling off a cliff and disappearing. One-time leaders like SGI and
Intergraph are turning into niche players, who are losing out badly to the
new guys like Dell, Gateway, NVidia, and 3dfx who were born and bred in
the consumer, high volume, low-price PC era.
The result is that projects, like the Notre Dame reconstruction, architec-
tual walkthroughs, and training simulations–which were once the realm
of high-end SGI’s–are now coming to the PC in droves. Soon, they’ll all
be here.

3.21.3 The Economics

What we’re seeing in processors, 3D accelerators, and games is a funda-

mentally different level of technological progress than happens in other
markets. Industrial era businesses, like car manufacturing, and limited
by raw materials and labor. Most of what you pay for a car goes to materi-
als, labor, and the other economic components of the industrial pipeline

3.21. JUL 03, 1999

Tim Sweeney Archive 86 News

(like transportation, insurance, and so on). Years of tweaking the process

have made manufacturing near optimally efficient, so future progress
goes very slowly, gaining perhaps 2%-3% productivity per year. That’s
why there is no Moore’s law for car performance.
But in our business, the cost of raw materials is insignificant, so innova-
tion is driven by volume and R&D investment. A 500 MHz Pentium III
now costs the same to manufacture as a 60 MHz Pentium cost in its day,
or a 386-16 long ago. A retail game contains around $2.50 of raw mate-
rials (box and CD) and sells for around $49.95. So, manufacturing isn’t a
limiting factor here.
Still, games, processors, and 3D cards require a tremendous up-front in-
vestment in R&D and facilities. It takes lots of people and a lot of capital
to develop a game, design a chip, or design and build a manufacturing
plant. The difference is, this investment is a fixed cost–you pay it once,
then you have the ability manufacture as many games or chips as you
desire at an (approximately) insignificant cost.
The remaining piece of the puzzle is the ”winner take all” nature of tech-
nology markets. Since manufacturing costs are small, price differences
are minor, leadership goes to the companies that have the best high-
volume products. Therefore, everyone invests heavily in R&D, and at-
tempts to produce the best product in hopes of selling tons of chips or
games or engines. When they do this successfully, they make a profit and
plow it back into the business to get further ahead.
Most R&D and capital-investment hungry markets work this way; it’s why
you only see a handful of prominent CPU makers, 3D card makers, and
engine developers. The ones who delivered were successful early on and
reinvested in their continued growth; and those who didn’t deliver either
out-of-business or are falling apart now.

3.21.4 The Result: General-Purpose Solutions Win

The result is that high-volume general-purpose hardware and software

is quickly gaining the lead in absolute performance. This trend will only
accelerate as the PC market becomes larger and the R&D investments

3.21. JUL 03, 1999

Tim Sweeney Archive 87 News

grow. The ramifications are:

• Specialized software and hardware loses out to general-purpose.

The company that focuses on generality (for example, a CPU, or
a 2D+3D graphics accelerator, or a general-purpose game engine)
has the expectation of higher sales, and can make so much more of
an R&D investment that special-purpose solutions can’t compete
(for example, 3D-only accelerators, or a once-off engine powering
a single game).

• Volume and increasing R&D investment enable 3dfx and NVidia to

sell millions of $150 graphics cards that outperform SGI’s $50,000
solution which sells 1000 units. That has the effect of crushing SGI’s
business model, even though companies like 3dfx aren’t even con-
sciously trying to compete with SGI.

• When a company like Epic or id Software builds a 3D engine, we do

it with the expectation of it powering games that sell a few million
copies, between our games and licensee projects. We make money
from each one, and that realization drives investment in making
more and better tools. This gives our engines a price, performance,
and feature advantage over ”in-house” engines made specifically
for one game, and over less-capable engines designed for niche mar-
kets. We amortize our R&D investment over ~10X more units sold.
This licensing model has already proven successful with mainstream
game developers, but now it’s starting to overflow into non-game
markets too. We have this great 3D engine, why not use it for your
architectural walkthrough?

• As a result of widespread interest, thriving ecosystems develop around

successful general-purpose products. For example, there are hun-
dreds of web sites dedicated to 3dfx hardware, and lots of develop-
ers optimizing their games for 3dfx. There are many hundreds of
web sites covering Unreal and licensee projects; thousands of kick-
ass aspiring level designers building maps and making them avail-
able online; many licensees building games; and lots of other cool

3.21. JUL 03, 1999

Tim Sweeney Archive 88 News

projects that tie into the engine, such as mods and TC’s, and the re-
search projects like Notre Dame. The community has a multiplica-
tive effect on growing the platform.

3.21.5 Where we go from here

When you look at the big picture, what’s happening now with 3D graphics
on the PC is just the tip of the iceberg. In the past, we’ve been limited
to the realm of hardcore gamers, but now 3D acceleration is becoming
much more mainstream, and 3D engines are becoming recognized as a
viable development tool for a wide range of projects. The coming years
will be interesting.

3.22 Jul 10, 1999

3.22.1 The AMD Athlon Rocks!

My new 550 MHz

(K7) just clocked a jaw-dropping 68.5 Unreal timedemo at 1024x768, run-
ning on a Voodoo3 3000 card. Even more telling, at no point did the frame
rate ever drop below 38.0 fps. That’s astounding, considering the intense
lightmap and geometry usage in the timedemo level. Even while play-
ing Unreal Tournament’s most texture and polygon intensive level (Shane
Caudle’s DmGothic), the frame rate hardly ever went below 60 fps.
The Athlon’s 128K L1 cache is awesome for memory-intensive games like
Operations like visibility determination, which thrash on the Pentium
III’s 32K cache, now run at full speed on the Athlon. This CPU truly shows
a generational performance improvement, like going from my old 486 to
my first Pentium.

3.22. JUL 10, 1999

Tim Sweeney Archive 89 News

When I saw AMD’s K7 spec, I was pretty skeptical. The K6 had been hyped
up, but in reality it was slower for Unreal than a Pentium II of comparable
clock rate, due to its poor non-SIMD floating point performance. The K7
claimed to fix all of that, and debut a new architecture with 3 execution
pipelines. I decided to wait and see, without getting my hopes up.
Bottom line: I waited, and now I have seen! The Athlon is clearly the
fastest x86 CPU at any clock speed.
Congratulations go to AMD.
My wish list:

• I want to be able to buy dual-processor Athlon workstations from

major manufacturers like Dell and Gateway.

• I want MMX, 3DNow! and SSE code generation support in Visual

C++, with native SIMD C/C++ datatypes like float2, float4, and short4,
making the compiler manage all register allocation and code gen-

• I want! 8-)


3.23 Jul 12, 1999

3.23.1 New Downloads & Links Pages

By popular request, the links on the tech page have been updated and
cleaned up. I added:

• New page: An index

of all the official files here on the tech page, plus links to dedicated
Unreal download sites. If you’re looking for the latest PC or Mac
patch, UnrealEd patch, Fusion maps, or whatever, look here.

3.23. JUL 12, 1999

Tim Sweeney Archive 90 News

• New page. Links to some

of the major ”hub” web sites related to the Unreal community, Epic,
our partners.

I also wanted to mention the supercool new download site, http://www., check it out.

3.24 Aug 14, 1999

3.24.1 DirectX7

Some people have been asking about our plans for supporting the up-
coming DirectX7. Unreal Tournament will ship with DirectX6 support,
and will be compatible with (but not optimized for) DirectX7. As soon as
Microsoft has released DirectX7 and we’ve had a chance to optimize the
code for DX7 and test it across a wide range of cards, we’ll release an Un-
real Tournament patch with complete DirectX7 optimizations. This has
been the plan all along.
DirectX7 has lots of cool new features. The ones which Unreal Tourna-
ment will exploit are:

• Improved texture management.

• Hooks to enable 3D card drivers to perform their own texture man-

agement to improve performance.

• Optimized polygon path for significantly faster polygon rates.

• Windows 2000 support.


3.24. AUG 14, 1999

Tim Sweeney Archive 91 News

3.25 Aug 16, 1999

3.25.1 Unreal Tournament development note

In response to the pirate punk who has been claiming to have a ”final”
version of UT...
Unreal Tournament isn’t finished yet, so there isn’t a final version. We’re
still polishing, fine-tuning, and optimizing the game and will ’till it’s done.
Believe me, when UT goes gold, we’ll be announcing it here prominently.
Until the demo and release version are available, any copy of UT you see
is an unfinished beta version; unless you’re a reviewer who received such
a copy direct from Epic or GT, having such a copy is illegal. These versions
are clearly labeled ”confidential beta version” or ”release candidate - not
for distribution” in at least 5 places in the program. In addition, they
have time-bomb logic which will cause them to stop functioning after a
certain amount of time. So if anybody gets hold of a version of UT before
release, it should be pretty clear that it’s not legitimate.

3.26 Sep 14, 1999

3.26.1 Unreal Tournament demo schedule

To quote quoting me :), here’s the Un-

real Tournament demo release plan:

• Windows 3dfx demo this week.

• Windows Direct3D/OpenGL next week.

• Linux Glide/OpenGL soon after.

• Mac timing not figured out yet.

3.25. AUG 16, 1999

Tim Sweeney Archive 92 News

3.26.2 DirectX7

Microsoft’s new DirectX7 API will be released soon. I’ve already ported
Unreal Tournament’s Direct3D code to DirectX7 and have noticed a nice
speedup on the TNT2 and GeForce256, primarily due to improved tex-
ture management.
The API’s simplicity has also improved, which is something you don’t of-
ten see: usually code just gets more complex as it evolves. Porting Unreal
Tournament’s code from DirectX6 to DirectX7 only took 3.5 hours, and
mostly consisted of deleting now-redundant code and changing func-
tion calls and interfaces. I’m very glad to see the IDirect3DTexture, IDi-
rect3DViewport, and IDirect3DLight interfaces gone, and replaced by much
simpler state-setting code. Direct3D’s abuses of object-oriented program-
ming are now gone.
With DirectX7, Microsoft did something I welcome, and would like to see
more of: they designed the new Direct3D interfaces to not be backwards-
compatible with the old ones. This enabled them to remove from view a
bunch of the old baggage that obfuscated Direct3D: execute buffers are
gone, unnecessary intermediate objects are gone, and much less weird
COM QueryInterface stuff is necessary.
Ripping out old code and replacing it with new, better designed code is a
great practice which too many software developers are afraid of.
-Tim Sweeney

3.27 Sep 22, 1999

3.27.1 DirectX7

Microsoft has released DirectX7. You can read about it and download it
from and download it on their http:
Now that DirectX7 is available, we’re doing final testing and tweaking

3.27. SEP 22, 1999

Tim Sweeney Archive 93 News

on the upcoming Unreal Tournament demo with Direct3D and OpenGL

support, coming in the next few days.
One thing we’ve found on DirectX7 is that NVidia TNT2 performance is
not fill-rate limited until you get to very high resolutions. During testing,
performance at 640x480 seemed a bit slow, but as we increased the res-
olution, the frame rate was hardly impacted at all. So 1024x768 seems to
yield the best overall experience on this card.
I’ve enabled support for 32-bit color textures under Direct3D, which sig-
nificantly improve the graphical quality, with a 10-15% frame rate im-
The Direct3D code now uses vertex buffers, which speeds up mesh and
text drawing a bit.

3.27.2 What about OpenGL?

We’re still maintaining OpenGL sypport, though it’s not as much of a pri-
ority at the moment because texture management performance in GL
on Windows is significantly behind Direct3D. It’s still useful in NT4 and
Linux of course. I’ve been advocating a GL extension which would en-
able the fine-grained control over texture management that has led to
such an improvement in Unreal performance under Direct3D. Here is a
suggestion I sent to the OpenGL ARB.

I’ve been maintaining the Unreal / Unreal Tournament render-

ing code under both Direct3D and OpenGL simultaneously for
a while. They used to have approximately equal performance
back when I was using Direct3D’s ”driver-managed textures”
(similar to the one and only OpenGL option for texture man-
agement, which is transparent to the application). Unfortu-
nately, the performance wasn’t very good, and the memory us-
age was nuts because our game uses a huge amount of tex-
tures, and the automatic texture management in Direct3D and
OpenGL had to keep duplicate copies around as ”backing store”
to the copies in video memory.

3.27. SEP 22, 1999

Tim Sweeney Archive 94 News

However, I recently rewrote my Direct3D code to manage tex-

tures explicitly. I create a limited number of textures of all pos-
sible sizes, forced to be in video memory, and then at runtime
swap my actual game textures into those ”video memory tex-
tures”. This way, I have complete control over texture manage-
ment, and can make optimal decisions about where to put tex-
tures, based on the constraints of my app, such as:
• Texture usage patterns.
• Decompressing textures at glTexImage2D time using my
own internal formats.
• Having a background thread load them off disk specula-
tively and stick them in video memory when needed.\
• Other nonlinear, possibly time-variant priority factors.
When I replaced D3D’s default texture manager with my own
code, Unreal’s performance and memory usage under Direct3D
improved very significantly, to the point where it’s not worth
bothering playing the game in OpenGL anymore.
So, my question is, does anybody have plans to add an ”ex-
tended version of glBindTexture” which lets the application tell
the OpenGL driver, ”allocate this texture in video memory, keep
it there, and never swap it out”? If an option like this existed,
I could do the same kind of high-level optimizations as I did
in Direct3D, and probably get a comparable increase in perfor-
mance. Obviously, video memory is a limited resource, so at
some point those glBindTextures will fail if video memory fills
up, and have to return an error code.
This is a bit lower-level than GL’s existing glBindTexture mech-
anism, but I think it’s justified by the need to support apps which
use huge quantities of textures and have to manage them care-
fully in order to maintain realtime frame rates.
For example:
posed, then you can replace a call to glBindTexture( GL TEXTURE 2D,
MyBindId ) with glBindTexture( GL TEXTURE 2D VIDEOMEMORY,
MyBindId ) and the texture is guaranteed to be allocated in

3.27. SEP 22, 1999

Tim Sweeney Archive 95 News

video memory and never moved around by the GL driver’s tex-

ture manager.
For my application, glPrioritizeTextures(?) isn’t a viable solu-
tion, because it still requires extra backing-store copies of tex-
tures to be kept in system memory so they can be swapped in
and out. In addition, implementing my own texture manager
on top of the existing, unextended glBindTexture isn’t at all effi-
cient, because I don’t have any way of knowing how many tex-
tures I can allocate before they start spilling out of video mem-
ory, and in addition most GL drivers do lots of extra internal
memcpy’s in the process – and with big textures, having lots of
unnecessary 256K memcpys (for 256x256x32-bit textures) just
kills performance.
Such a solution would benefit Unreal engine games on Win-
dows, Linux, and possibly Mac such as Unreal Tournament,
Duke Nukem Forever, Deus Ex, Wheel of Time, etc etc etc. As
well as other developers’ texture-intensive games that use OpenGL,
such Starsiege: Tribes (the developers have been running into
this same issue).

3.27.3 Experiencing poor Internet play on your Voodoo3


As reports, the Voodoo3 3500TV’s WebTV

installer does some evil things to your Internet setting, causing many
games (including Unreal Tournament and Quake 3 Arena) to experience
poor Internet play:

Hello, We are aware of the problems that the Voodoo3 3500TV

is experiencing with online games. Many of these problems can
be cured by removing WebTV, however that will disable the Vi-
sual Reality software. The problem usually can also be cured
by removing the Internet Explorer 5 upgrade from Windows98.
While these are just a work around, and not really fixes, both
options usually will cure the problem until 3dfx releases a real
fix. We are currently working with Microsoft to try and figure

3.27. SEP 22, 1999

Tim Sweeney Archive 96 News

out whether the problem is due to Microsoft software (WebTV

and IE5) or 3dfx software (3500TV drivers and Visual Reality).
Thank you for your patience,
Aaron D. Patton
3dfx Interactive Email Support


3.28 Sep 29, 1999

3.28.1 More Hardware Troubleshooting

• We received some more confirmations of Unreal working better on
TNT’s with the Creative Labs unified drivers using Glide than Di-
rect3D. If you have a Creative Labs TNT card and have having per-
formance problems, try that. Especially if you have under 128 megs
of RAM.

• Lots of reports of slowdowns from users with Monster Sound’s who

have enabled the ”Advanced Options / Audio / Use3DHardware”. If
you’re having this slowdown, disable Use3DHardware and try again.
Then please email and let us know
whether that helped. I’ve seen three confirmations from users that
their problems (huge slow down) went away after disabling this.

• Jack and I have been tracking down the performance problems with
the TNT on 64-meg (and lower) machines. To our great surprise, it
appears that the TNT is keeping duplicate system memory copies
of all our textures–we saw system memory grow and shrink in al-
most exact proportion to our ”supposed” video memory texture al-
locations. Thus the game uses an extra 12-26 megabytes of system
memory–a very inefficient allocation of resources. This is a big sur-
prise, because the NVidia guys has always told us this isn’t the case.
Problem in
I don’t think so, but I’d love to be proven wrong. I’ll follow up with

3.28. SEP 29, 1999

Tim Sweeney Archive 97 News

our henchmen at NVidia and Microsoft and see if we can track this

3.28.2 Direct3D Anomalies

We’ve been looking at the feedback on Direct3D performance and inves-

tigating some strange reports. Recently, we’ve mainly been testing on
96-meg and 128-meg machines (I have a Celeron 400, Jack Porter has a
K6-2 450). On these machines, TNT1 performance is good – average 28
fps at 648x480, 25 fps at 800x600. The TNT2 performance is significantly
However, upon removing some RAM and testing Direct3D on a 64-meg
K6-2, the ”precache” time increased by about 5X, and performance dropped
to a few frames per second. These performance drops don’t occur in the
software renderer, and don’t occur in Glide. Something is going wrong
between Unreal, Direct3D, and the TNT’s Direct3D driver, and we’re in-
Overall, the feedback indicates a very wide variance in performance among
TNT users, much more so than with any other card. Our internal testing
has indicated this too; for example, we’ve found (and worked around) a
lot of driver bugs that only happen on one machine, and not others with
otherwise similar configurations.
Don’t Try This At Home Dept.: Some TNT users have reported that tweak-
ing their BIOS’s ”AGP Aperture Size” improves performance on 64-meg
machines. We have tried this and couldn’t find any differences on our
64-meg test machine. Others report that the Creative Labs unified drivers
(with TNT Glide support) outperform Direct3D on their cards. If anybody
finds definite improvements or workarounds, or has insight into what’s
happening, please email
and let us know.

3.28. SEP 29, 1999

Tim Sweeney Archive 98 News

3.28.3 Athlon In The House

Mark Rein walked into Best Buy and picked up a 650 MHz Athlon off the
shelf – an IBM Aptiva with a TNT2 Ultra bundled in. Fastest off-the-shelf
machine we’ve ever seen! Click
wav for our in-depth Athlon review.

3.29 Oct 01, 1999

3.29.1 TNT Users, Try This!


• Play in 16-bit color, which is significantly faster than 32-bit color.

• Turn the detail options down in the Preferences/Video menu.

• Use the ”-” key to reduce the HUD size.

• In Preferences/Game, turn off the display of your weapon.

• If you have enabled Advanced Options / Audio / Use3DHardware

audio support, TURN IT OFF for now! Big driver performance prob-

Experimental improvement for the brave:

• Try (thanks

to site for the link).

• Try putting this new into

your \TournamentDemo\System directory. Please back up your ex-
isting one first!

3.29. OCT 01, 1999

Tim Sweeney Archive 99 News

These two downloads are beta and haven’t been rigorously tested yet, but
in our gameplay sessions here, they’re a huge improvement on low-RAM
machines with the TNT, TNT2, and GeForce256.
Please let us know how these work for you, by emailing mailto:utbugs@

3.29.2 Latest Findings

We’ve been benchmarking UT performance and memory usage on many

3D cards, and we’ve found the source of the performance problems on
64-96 meg machines with TNT/TNT2 cards.
Our measurements indicate that the TNT/TNT2 drivers consume up to
25 megabytes of page-locked system memory with UT, while other Di-
rect3D cards don’t. On machines with 64-96 megs, that kills performance,
as one would expect. The game runs, but data is swapped to the hard
drive so frequently that playability is compromised. We’ve been compar-
ing with other 3D cards and aren’t seeing the issue on the Rage 128 or
Voodoo3 under Direct3D.
System-wide memory allocations measured with SYSMON.EXE on a 64-
meg Celeron running Windows 98 SE:

• TNT D3D: 51 MB baseline, 122 MB playing UT in 16-bit color, 134

MB playing UT in 32-bit color.
• Rage 128 D3D: 55 MB baseline, 107 MB playing UT in 16-bit color,
109 MB playing UT in 32-bit color.
• Voodoo3 D3D: 51 MB baseline, 100 MB playing UT in 16-bit color.

We’re working with the NVidia guys to track down the source of the extra
25 megabytes of memory consumption. I’m going to hold off on posting
our benchmarks till we’ve had a good chance to deal with the memory
consumption issue, so everyone is on a level playing field. The NVidia
cards already perform well on PC’s with lots of RAM, so they should be in
good shape once the memory issues are solved.

3.29. OCT 01, 1999

Tim Sweeney Archive 100 News

3.29.3 Links

Unreal Technology in the news:


• - Newsday

Cool Sites:

• - Cool Unreal site & community host in


• - The GameSpy empire’s new portal site.

Check out their page,
where Unreal Tournament is #6 and rising quickly. Hopefully some-
day we’ll overtake that 3-year-old QuakeWorld thing. :-)


3.30 Oct 02, 1999

3.30.1 Unreal Tournament players overtake QuakeWorld

at #5

Top Game Servers By Players from

@ 7:52 PM EST

Game Servers Players

================== ======= =======
Half Life 1502 4344

3.30. OCT 02, 1999

Tim Sweeney Archive 101 News

Quake II 1628 1711

Starsiege TRIBES 589 1162
Quake 3: Arena 1040 1156
Unreal Tournament 423 520
Quakeworld 623 517
Kingpin 127 256
Unreal 134 144
Turok 2 20 43
Descent 3 8 23


3.31 Oct 06, 1999

3.31.1 Windows 2000 RC2 Rocks!

I recently upgraded to Windows 2000 Release Candidate 2, and wanted

to mention that it’s extremely solid! With NVidia’s latest Windows 2000
TNT/GeForce drivers, DirectX7’s Direct3D and OpenGL both work flaw-
lessly. What’s cool about the new OS:

• High-performance DirectX7 Direct3D and OpenGL support.

• USB mouse support – mouse movement is very smooth.

• Much more stable than Windows 98, especially for software devel-

• Good administration tools, more powerful than NT4 and far easier
to use.

• Great international support.

• Dual processor support.

• Plug and play.

3.31. OCT 06, 1999

Tim Sweeney Archive 102 News

3.31.2 Windows 2000 Internationalization

This is a topic we’re very interested in, because we’re always trying to
make our games and engine as widely accessible as possible. As a mat-
ter of fact, today I’m working out the font management issues for the
Japanese version of Unreal Tournament.
The Windows 95/98 OS had very weak international support, and a con-
fusing array of incomplete features such as supporting Unicode in a few
API’s but not all, weird and undocumented multibyte character set han-
dling, and mixed ANSI/OEM file system conventions. Localizing Win-
dows 9X apps is a very difficult undertaking.
On the other hand, Windows 2000 is by far the most thoroughly inter-
nationalized piece of software I’ve ever seen. All international versions
sharing the same binaries, fonts, and input support. Everything is 100%
Unicode. So, you can open up a Japanese language document on your
English computer, visit Arabic web sites, use an Input Method Editor to
type in Chinese text, and so on. While most people might not care about
that, it’s infinitely beneficial for developers to have that power at their
fingertips, because it enables you to do all of your international devel-
opment and most of your testing without jumping through hoops and
installing multiple OS’s.
When Windows 2000 makes its transition into the consumer OS market,
it will be a great day for worldwide software development.

3.31. OCT 06, 1999

Chapter 4

2000 News

4.1 Feb 22, 2000

9 out of 10 doctors say now would be a good time to prune the Unreal
tech page, and that sounds like a good idea to me. I’d like to start over
by talking about a bunch of things I’ve been meaning to cover, but never
found the time.

4.1.1 Common Questions I’m Asked

Q. I’m an aspiring game developer and I want to get into the business.
What’s the best way to be hired?
A. The single most important thing game companies will look for is past
experience: what cool stuff you have worked on previously. This might
sound like a Catch-22: ”You need experience to get a job, and you need
a job to get experience”. But nowadays, the mod communities around
leading games like Half-Life, Unreal Tournament, and Quake 3 Arena
are great proving grounds where aspiring game developers can work on
projects freely (mostly in their spare time), and build levels, mods, and
other game-play enhancements. In the past few years, the majority of
new talent Epic has hired have been people from the Unreal and Quake

Tim Sweeney Archive 104 News


• Steve Polge, creator of the ReaperBot mod for Quake.

• Brandon ”GreenMarine” Reinhart, early Unreal mod maker and writer
of cool UnrealScript programming tutorials.
• Jack ”mek” Porter, Unreal mod maker and PlanetUnreal news cor-
• Alan Willard, maker of cool Unreal and Quake levels.
• Shane Caudle, amazing artist and level designer who was building
his own Quake TC.

While past work on cool projects is the number one criteria for most
game developers, having a University degree is a major advantage. While
this isn’t a must at Epic, most larger game developers place more em-
phasis on having a degree. Besides that, college is a great opportunity
to learn useful stuff. My degree is in Mechanical Engineering (University
of Maryland) – by the time I was in college, I had been programming for
about 10 years and didn’t feel getting a computer science degree would
be challenging. So, I chose engineering, and that turned out to be a major
challenge, and incredibly valuable.
The math I learned there, including vector calculus and finite element
analysis, which are directly applicable to 3D games, is something I never
would have studied independently. Self-taught programmers pick up al-
gorithms just by looking at other code and reading the occasional book
(that’s how I learned to program). But differential equations are just not
the kind of thing you’re likely to rediscover on your own, though un-
derstanding them brings clarity to lots of real-world problems you’ll en-
Q. What do I need to get started making mods for popular games?
A. Just a copy of the game, the editing and compiling tools, and access
to the web sites that contain information for mod authors. For http:
//, you just need a copy of UT, which in-
cludes the editor (on CD#2) and all the

4.1. FEB 22, 2000

Tim Sweeney Archive 105 News

unrealscript.htm code for the entire game; visit the Unreal community
web sites (listed at the top of this page) for tons of pointers to Unreal in-
formation. For, you need the retail game
and the utilities; visit
to get started. For, get the game and
see for tips.
Q. I’m working on a project; do I need to license the Unreal Tournament
engine or sign any paperwork?
A. That depends on what you’re doing:

• If you are making a freely available mod, all you need is the retail
version of the game.
• If you are using the engine for non-commercial academic purposes,
all you need is the retail game. Lots of people have been using the
Unreal Tournament engine as part of university projects and theses.
• If you are a startup game developer and are prototyping a game in
the Unreal Tournament engine (using the tools available in the re-
tail version), you don’t need a license agreement while you’re proto-
typing the game. You do need a license agreement in place before
you advertise a game, sign a publishing agreement, or otherwise
profit from the game.

4.1.2 Programming Languages

Ever since my Developer Week article, ”http:

//”, I’ve been getting ques-
tions about whether we plan to ”do something about it” and define a
new programming language. The answer is, we’re already doing that to
a certain extent with,
which is a high-level, object-oriented, platform-independent language
with some cool extensions aimed at game code development.
As an engine developer, we’ll always be involved in language design to
a certain extent. In the future, the scripting language could evolve into a

4.1. FEB 22, 2000

Tim Sweeney Archive 106 News

simpler form, more true to the ”scripting” term: controlling sequences of

in-game events. Or, the scripting language could grow to subsume even
more of the engine code. We’ll be evaluating these options anew with
each major iteration of the engine, doing whatever is most appropriate
for the technology, recognizing the benefits and drawbacks of going with
well-known proven tech, versus creating newer, better tech.

4.1.3 Also...
• The cool soon-to-be-public site al-
ready has posted a
unreal_tourney.html on how to get started with Unreal Tourna-
ment under Linux.
This site is a great idea; the biggest barrier to widespread Linux
adoption right now is the difficulty of getting started, and now in
standard decentralized Linux fashion, sites and resources are start-
ing to pop up to address this.

• on http://www.chessmess.
com/ (saw this on


4.2 Feb 28, 2000

Check out the on

the state of mod development, by our own Brandon ”GreenMarine” Rein-
hart. This is the first feature in new http:// feature.
Upcoming articles in the series feature Jesse Taylor from the Infiltration
mod team for Unreal Tournament, and David ”crt” Wright of Rocket Arena
for Quake 3 and Unreal Tournament.

4.2. FEB 28, 2000

Tim Sweeney Archive 107 News

Now is an exciting time for mod development, as the best teams of enthu-
siasts are earning tremendous praise, opening up serious biz opportuni-
ties. This dynamic reminds me of the shareware game business around
1991 when Epic, Apogee/3D Realms and id Software were born. Back
then, we were just a handful of kids making games for our own enjoy-
ment, and releasing them online for others to enjoy – hoping that we
might be able to make some money to pay the bills.
Today all three companies are industry leaders.
Now, think about where today’s best mod authors will be in 9 years!
The advantage that enabled id, 3D Realms, and Epic to rise to the top
is something that today’s mod authors have too: we’re in the middle of
a thriving community, and that gives us an awesome feedback loop. If
we release something that sucks – whether it’s something big like Un-
real 1’s network code, or something subtle like weapon balancing – then
thousands of people will email us complaining until we fix it. Thousands
of smart people besides ourselves are expanding the game universe by
running web sites, making mods, building levels, and running servers,
and they let us know what they like and dislike. In this kind of pressure-
cooker environment, only good games and good ideas survive.

4.2. FEB 28, 2000