Notes On 30hz

The Effects of Adaptive Synchronization on Performance
and Experience in Gameplay
BENJAMIN WATSON, North Carolina State University, USA

RACHIT SHRIVASTAVA, Microsoft, USA
AJINKYA GAVANE, North Carolina State University, USA
Adaptive synchronization (Async) is a recent display technology that lowers frame time mean and variation
by altering communication between display and GPU. While conventional graphics systems introduce delay
to match irregular GPU frame rates to regular display refresh rates, Async systems instead match display
refresh rates to to irregular frame rates. Both NVIDIA and AMD claim that Async displays improve experience
for enthusiast gamers. In a first, controlled examination of these claims with non-professionals playing a
popular first-person shooter, we find that Async improves gaming performance, can improve experience, and
has intriguing relationships to game familiarity and years of gameplay.
CCS Concepts: • Human-centered computing → Displays and imagers; Empirical studies in
HCI ; • Applied computing → Computer games.
Additional Key Words and Phrases: refresh rate, frame rate, latency, computer games, user experience
ACM Reference Format:
Benjamin Watson, Rachit Shrivastava, and Ajinkya Gavane. 2019. The Effects of Adaptive Synchronization on
Performance and Experience in Gameplay. Proc. ACM Comput. Graph. Interact. Tech. 2, 1, Article 5 (May 2019),
13 pages. https://doi.org/10.1145/3320286
1 ADAPTIVE SYNC’S PROMISE

For decades, computer displays with fixed refresh rates have set the pace of real-time graphics
display. And for decades, this made good sense, with frame generation happening more slowly
than display refresh. However, as graphics hardware (GPUs) improved, displays became a signifi-
cant throttle on graphics performance. GPU and collaborating display manufacturers therefore 5
introduced adaptive synchronization (ASync), a technology that allows GPUs to adaptively synchro-
nize display rates to frame rates, lowering frame time mean and variation. NVIDIA’s proprietary
implementation is called G-SYNC, while FreeSync is AMD’s competing and open standard.
ASync promises improved gaming experience for the broader gaming market. NVIDIA claims
that G-SYNC display systems are “go-to equipment for enthusiast gamers,” giving them “more
of what [they] want in a gaming experience.” AMD similarly urges gamers to “maximize [their]
gaming experience” with FreeSync.
This research is a first experimental examination of the effects of ASync on the experience
of dedicated (but not professional) gamers. We learned that in at least one popular first-person
Authors’ addresses: Benjamin Watson, North Carolina State University, Dept. Computer Science, Campus Box 8206, Raleigh,
NC, 27695-8206, USA, bwatson@ncsu.edu; Rachit Shrivastava, Microsoft, Redmond, WA, USA, rashr@microsoft.com;
Ajinkya Gavane, North Carolina State University, Dept. Computer Science, Campus Box 8206, Raleigh, NC, 27695-8206,
USA, asgavane@ncsu.edu.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the
full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires
prior specific permission and/or a fee. Request permissions from permissions@acm.org.
© 2019 Copyright held by the owner/author(s). Publication rights licensed to ACM.
2577-6193/2019/5-ART5 $15.00
https://doi.org/10.1145/3320286
Proc. ACM Comput. Graph. Interact. Tech., Vol. 2, No. 1, Article 5. Publication date: May 2019.
5:2 Watson, Shrivastava and Gavane
shooter (FPS), ASync can improve gaming performance, and may also benefit experience. We also
found that ASync has intriguing relationships to game familiarity and years of gameplay that merit
additional investigation. Further research should study these relationships, as well as ASync’s
effects in systems with higher frame rates.
2 TEMPORAL GAMING TECHNOLOGY

In the early days of computer graphics, refresh rates were much higher than frame rates, and their
synchronization artifacts were negligible. But as graphics hardware improved, frame and refresh
rates converged, making artifacts appearing as “tears” in imagery much more common. Double
buffering eliminates tearing at the cost of increased latency, introducing a back buffer that receives
frame updates while the front buffer is displayed. At the vertical synchronization (Vsync) signal
between display refreshes, if the frame update is complete, the buffers are swapped. If the GPU
finished the back frame before Vsync, it waits, introducing delay. If the GPU did not finish, the
front frame displays again, causing still more delay.
To eliminate tearing without introducing latency, new ASync technologies such as G-SYNC and
FreeSync allow displays to adaptively match refresh rates to frame rates, even as frame rates vary.
With such synchronization, GPUs need not wait, and frames need not repeat. Frame time mean
and variation decrease. G-SYNC is supported on all current NVIDIA GPUs, and FreeSync on all
current AMD GPUs. Monitors with proprietary G-SYNC are currently available for as little as $350,
while open FreeSync monitors are sold for $200. Quite recently NVIDIA announced that it would
support some FreeSync monitors, so we expect these prices to converge.
Although most dedicated gamers would love to have the highest-end gaming technology, the
majority do not. About one-third of current console games limit frame rates to a 30Hz maximum,
the remainder to 60Hz [Brown and Locke 2018] [IGN 2016]. The PC and console markets are
roughly equal in size [Papadopoulos 2017], and though data are limited, one recent survey showed
well over half of enthusiast PC gamers using displays refreshing at 60Hz or less [Kampman 2017].
Actual console and PC frame rates are still lower due to Vsync delays.
3 TIME, PERFORMANCE AND EXPERIENCE

There is a long tradition of research on the effects of temporal sampling and delay on human
performance and experience. Claypool and Claypool [Claypool and Claypool 2006] as well as
Chen and Thropp [Chen and Thropp 2007] have surveyed this work well, finding that temporal
improvements benefit human performance and less clearly, experience. In this section we review
more recent additions to this body of work, with particular attention to gaming and experience.
There has been interesting work examining the impact of delay in recently developed touch
interfaces, such as phones, tablets and mousepads. Janzen and Teather [Janzen and Teather 2014]
studied temporal effects on laptop touchpads, finding that 30Hz frame rates reduced performance
when users tried to click on moving targets. Jota et al. [Jota et al. 2013] varied delay in tapping and
dragging tasks on tablets, finding that while latencies of just a few milliseconds are perceivable,
limiting delay to 20ms or less is adequate to avoid impacting performance. Deber et al. [Deber et al.
2015] followed up Jota et al.’s work by focusing on perceptibility, finding that latencies of 11ms or
more can be perceived when users are dragging, and 69ms or more when users are tapping.
Another line of research has examined the effect of networked latency in games on performance
and experience. Hohlfeld et al. [Hohlfeld et al. 2016] added delay as novices played Minecraft,
finding that delays of up to 1s did not impact experience as measured by a standard questionnaire.
Lee and Chang [Lee and Chang 2015] varied the “tick” rate in Counter-Strike (the rate at which
simulated game time was sampled), along with the use predictive latency compensation and v-sync
(the traditional alternative to G-SYNC, which requires GPUs to wait for the display refresh). They
The Effects of Adaptive Synchronization on Performance and Experience in Gameplay 5:3
learned that gaming performance is improved by higher tick rates and latency compensation, but
not affected by v-sync. The same authors [Lee and Chang 2018] later presented an advanced lag
compensation technique that improved gaming performance further.
Quite recently, visual perception researchers Poth et al. [Poth et al. 2018] described how G-SYNC
could be used to gain finer control of temporal resolution in display, allowing them to increase
frame time in sub-millisecond increments above a display’s frame time minimum. Even such minor
increases in frame time improved letter recognition by viewers. In sum, prior work provides strong
evidence that temporal improvements like those achieved by ASync should help gamers perform
better, across many different tasks. The recent work on networking delays in gaming are particularly
relevant in this respect. At the same time, the evidence for matching improvements in experience is
limited, largely because gaming experience is a relatively new research concern. We are not aware
of any prior research that directly examines the performance and experience effects of ASync.
In sum, prior work provides strong evidence that temporal improvements like those achieved
by ASync should help gamers perform better, across many different tasks. The recent work on
networking delays in gaming are particularly relevant in this respect. At the same time, the
evidence for matching improvements in experience is limited, largely because gaming experience
is a relatively new research concern. We are not aware of any prior research that directly examines
the performance and experience effects of ASync.
4 EXPERIMENT
ASync improves temporal sampling and reduces delay. Previous research shows enhancements
like these boost human performance, and may also result in better experience. We anticipated that
ASync would have similar effects, and tested that expectation in an experiment. In the process, we
also hoped to learn something about how such improvements interact with learning and expertise,
which prior work does not significantly address.
4.1 Participants
We had 27 participants in our experiment. All were students in undergraduate or graduate Computer
Science courses, and were recruited with a message stating that prior experience with FPS games
was “preferred.” They were compensated with extra credit in their course.
4.2 Design
We used a 3-factor mixed design (2GS × 2GS1 × 3YG). We describe each of these independent
variables and our dependent measures in more detail below. Figure 1 shows how our independent
variables and dependent measures relate to our experimental procedure.
4.2.1 Independent Variables. To introduce ASync, we used NVIDIA’s G-SYNC. As a proprietary

technology, it is more uniformly implemented than AMD’s FreeSync. All 27 participants played
Battlefield 4 (B4) both with G-SYNC on, and with G-SYNC off (and Vsync on). Thus our G-SYNC
(GS) variable was within subjects.
14 randomly chosen participants played with G-SYNC first (GS1 true), the other 13 played with
G-SYNC second (GS1 false). This made our G-SYNC first variable between subjects, and slightly
unbalanced.
Before any experimental gameplay, participants indicated whether they had been playing com-
puter games for less than a year, less than five years, or more. We did not screen to equalize
participation across these three levels of our years of gaming (YG) variable, making it not only
between subjects, but also unbalanced (Table 1).
Table 1. Participants vs. Independent Variables
years of gaming
<1 <5 > 5 total
true 2 5 7 14
G-SYNC first
false 3 4 6 13
total 5 9 13 27
Fig. 1. Our experimental procedure, and how it relates to independent variables and dependent measures.
Bold abbreviations are variables, other abbreviations are measures.
4.2.2 Dependent Measures. As participants began our experiment, we informed them that they
would be helping us test a new display technology, G-SYNC. After they completed gameplay, we
asked participants “Which monitor do you think is equipped with G-SYNC, the improved display?”
Participants replied by identifying the display they used in their first session of gameplay, or in their
second. Our display identification (DI ) was the proportion of participants who correctly identified
the GSYNC on session.
As our only measure of gaming performance (GP), we used B4’s own in-game single-player
scoring system. However, we used several experiential measures.
Before any gameplay, and after each gameplay session, we asked participants to assess their
emotional state using the self-assessment manikin (SAM) [Bradley and Lang 1994], which employs
pictorial stimuli to aid description of emotion in a three-dimensional model originally proposed by
Mehrabian and Russell [Mehrabian and Russell 1974]. In their model, the pleasure dimension (SAM-
P) corresponds to semantic differential adjective pairs such as “unhappy-happy,” “annoyed-pleased,”
and “unsatisfied-satisfied;” the arousal dimension (SAM-A) to pairs such as “relaxed-stimulated,”
“calm-excited,” and “sluggish-frenzied;” and the dominance dimension (SAM-D) to pairs such as
“controlled-controlling,” “influenced-influential,” and “cared for-in control.” By substituting three
pictorial questions for Mehrabian and Russell’s 18 textual questions, SAM simplifies collection
of emotional self-assessments, and avoids many linguistic and cultural challenges. According to
Perse [Perse 2010], the SAM measure’s validity has been demonstrated repeatedly, and it has used
successfully by researchers in several fields, especially with visual audiovisual stimuli such as
movies, TV and film. To assess the emotional change brought about by each gameplay session
relative to prior emotional state, we used the difference between succeeding SAM assessments as
our measure, resulting in a between - before SAM difference, and an after - between difference.
We also asked participants to estimate the number of minutes they had just played after each
gaming session. We then compared this number to the number of minutes they had actually played
to create our subjective duration assessment (SDA) measure. Czerwinski et al. [Czerwinski et al.
2001] suggested this implicit measure of usability, arguing that it reduced the positive bias prevalent
in explicit usability survey responses. When tasks felt difficult, users consistently overestimated the
task time. When tasks seemed easier, users underestimated task time. Bederson [Bederson 2004]
suggested further that SDA might be related to flow [Nakamura and Csikszentmihalyi 2014], a
state of full involvement in one’s activity, often characterized by a subjective acceleration of time.
4.3 Equipment
Because we sought to verify the existence of any effect of ASync on human performance and expe-
rience, we configured our equipment to maximize its improvement of frame times, and (hopefully)
experience. When frame times are very high, ASync’s improved times will still be high, and poor
experience should not improve much. When frame times are very low, ASync’s improved times
will still be low, and good experience should not improve much. We therefore targeted a modest
hardware platform with moderate frame rates of approximately 30Hz, where we expected ASync’s
improvements to have maximal experiential impact. As our review of temporal gaming technology
above makes clear, such modest systems are at the lower end of today’s broader market, but are
still widespread.
4.3.1 Hardware and Software. Games are specifically targeted in both NVIDIA’s marketing of
G-SYNC [NVIDIA 2018a] and AMD’s marketing of FreeSync [AMD 2018]. They make a natural
platform for our experiment, since they challenge both the people that play them, and the technology
that displays them. The game Battlefield 4 is a popular FPS specifically supported by NVIDIA
[NVIDIA 2018b]. All this makes NVIDIA’s G-SYNC quite likely to be effective with B4, so we chose
it as the software platform for our experiment.
Our PC used a 64-bit Intel Core 2 6600 CPU with 2.40GHz and 4GB RAM, and ran Windows 8.
In it we installed an unexceptional GeForce GTX 750 G-SYNC-enabled graphics card, which we
coupled to an Acer XB240H-A G-SYNC monitor. The GTX 750 is comparable to the GPU used in
the XBox One and the Playstation 4. Players used keyboard and mouse to interact with B4.
4.3.2 Configuration. NVIDIA, Acer, Windows and B4 provide many parameters that affect frame
times along with other display characteristics, which made configuring our experimental system
difficult. In making our choices, we continued to seek to maximize the effects of G-SYNC, but also
to find a configuration that was not atypical among gamers, and to ensure that gameplay would
not be taxing for participants. We collected data about G-SYNC’s effects on our system using the
well-known Fraps tool [Fraps 2013], which measures individual frame times. We did not use Fraps
to record video, minimizing its effect on system performance. We varied four parameters as we
collected frame time data: B4’s frame rate target (60Hz or 144Hz), its LOD (low, medium, high or
ultra), the Acer display’s refresh rate (120Hz or 144Hz), and NVIDIA’s G-SYNC (on or off). We
limited B4’s frame rate target to 60Hz or above, since 60Hz is a common minimum rate used in
gaming. Note that B4 does not strictly enforce this target at all, as the discussion below will make
clear. We used only the higher display refresh rates, because low refresh rates bring gamers few if
any compensating benefits. We also ensured that, as do actual frame rates, target frame rates never
exceeded refresh rates.
In examining our data, we sought parameters that maximized the improvements realized with
G-SYNC as measured by frame time mean, standard deviation, and number of outliers (over two
standard deviations from mean, which gamers experience as “jitter”). In our first round of testing
(included in supplementary data), we set B4’s target frame rate to 60Hz, while display refresh rate
Table 2. Frame Times on Experimental Equipment
G-SYNC On G-SYNC Off

Frame Rate Mean Std Dev Number Mean Std Dev Number
Target (Hz) LOD (ms) (ms) Outliers (ms) (ms) Outliers
60 Low 32.8 12.9 147 31.5 16.7 65
60 Medium 32.1 7.7 151 36.6 13.1 170
60 High 35.2 8.3 126 37.3 13.7 59
60 Ultra 45.8 16.3 62 48.1 12.9 108
144 Low 31.7 15.6 69 30.9 14.4 120
144 Medium 32.1 9.3 113 33.0 13.2 159
144 High 34.7 14.9 113 34.0 9.8 126
144 Ultra 42.0 12.9 111 38.8 12.3 87
varied between 120Hz and 144Hz. Across all B4 LODs, the 144Hz refresh rate maximized G-SYNC’s
effects. In our second round of testing (Table 2), we held refresh rate at 144Hz while varying
B4’s target frame rate between between 60Hz and 144Hz. G-SYNC generally reduced frame time
mean, standard deviation, and outliers — but there were many exceptions. We chose the following
configuration for our experiment: 60Hz as B4’s frame rate target, 144Hz as the display refresh rate,
and B4’s medium LOD. We would manipulate G-SYNC during the experiment itself.
4.4 Procedure and Task

Figure 1 sketches our procedure. When participants arrived, they were randomly assigned to either
the GS1 true or the GS1 false group. They then moved through seven experimental stages, which
we discuss below. Participants took an average of 50 minutes to complete the entire experiment.
4.4.1 Introduction. At the beginning of our experiment, we thanked participants for their help,
and made it clear to them that they could pause or halt their work at any time. We explained that
the experiment sought to determine the benefits of NVIDIA G-SYNC in gaming, and that they
would use G-SYNC in one of their gaming sessions, though they would not know which until
the experiment was completed. We noted that we would often ask them for their feelings and
thoughts about gameplay, and that we needed their honest responses — those responses were
anonymous, there were no “wrong” answers, and they would not hurt our feelings. We then gave
them our consent form, which summarized experimental procedure and detailed the risks, benefits,
confidentiality of and compensation for performing the experiment. All participants gave their
consent and proceeded with the experiment.
4.4.2 Before Gameplay Measures and Practice Session. We then surveyed participants to determine
which years of gaming group they belonged to, and assessed their emotional state using the SAM.
This pre-gameplay SAM data allowed us to account for individual differences in participant emotion
that did not result from gameplay.
Most participants did not know B4 well. To bring them to a basic level of familiarity with the game
and allow them to engage with the B4’s content rather than its interface, we gave participants time
to practice the game and learn how to move, use weapons, collaborate with non-player teammates,
and understand the map. We encouraged questions and provided guidance. For this practice session,
we used B4’s “Shanghai” mission in the “easy” mode. Participants started at the mission’s beginning,
and halted when they felt comfortable with the interface, or when 10 minutes had passed. Should
they die in the game, B4 would let them restart from an earlier checkpoint. Most participants halted
practice early. If participants were members of the GS1 true group, G-SYNC was on during practice,
otherwise G-SYNC was off.
4.4.3 First Gameplay Session and Between Gameplay Measures. Participants then began their first
recorded session of gameplay. As in practice, if they were members of the GS1 true group, G-SYNC
was on, otherwise G-SYNC was off. However, to distinguish this session from practice gameplay,
participants played a different mission called “Baku” in the “easy” mode, starting inside a building
after the non-interactive cutscene, and finishing when they exited the building and entered an
open field. To reach the field, players must find their teammates, acquire weapons, and win a few
firefights with enemy soldiers. Should they die, players must restart from an earlier checkpoint.
All players were able to reach the field. Participants took an average of 11 minutes, 45 seconds to
finish this gameplay session. Participants were not told how long they had been playing during the
session.
We then recorded the participants’ scores from B4 as our GP measure. Participants also filled out
a survey that for our SDA measure asked them to estimate how long they been playing in the first
session, and used the SAM to assess how the first session affected them emotionally.
4.4.4 Second Gameplay Session and After Gameplay Measures. Participants then started their
second recorded session of gameplay, which was more difficult than the first. Now, if they were
members of the GS1 true group, G-SYNC was off, otherwise it was on. Participants continued the
“Baku” mission in “easy” mode from where they left it in the first session, attempting to cross an
open field and reach a large drainage pipe. To so so, players must engage a larger number of enemy
soldiers and vehicles, restarting from an earlier checkpoint if they died. Most participants were
able to reach the pipe, but we halted the sessions of three players after 15 minutes. Including these,
participants completed this session in an average of 11 minutes, 51 seconds. Again participants
were not told how long they had played during the session.
After this final gameplay session, we once more recorded participants’ B4 scores for our GP
measure. Participants again estimated the length of time they had been playing in this session for
our SDA measure, and assessed their emotional state using the SAM. For our DI measure, we also
asked participants to tell us which gameplay session they thought made use of G-SYNC. Only now,
after completing the entire experiment including the DI measure, did we tell participants which
session did in fact use G-SYNC.
4.5 Hypotheses
We introduce our experimental expectations by first discussing G-SYNC’s perceptibility, then its
effect on performance, followed by its impact on experience.
4.5.1 Perceptibility. With our display identification measure, we ask participants to compare two
temporally differing display conditions using visual and motor sensitivities. The peak frequency at
which humans can see temporal flicker is about 60Hz [Watson 1986], and just noticeable differences
between temporal frequencies above 30Hz are in the single digits (3-6Hz) [Watson and Ahumada
2011]. As we reviewed above, users can sense delays of as little as 9ms [Deber et al. 2015] in motor
tasks. In testing of our experimental equipment, the difference between G-SYNC on and G-SYNC off
conditions was only about 5Hz (centered around 30Hz) and 4.5ms on average. Moreover, conditions
were far from optimal, with brightness and frame rates varying widely, and judgments relying on
human memory. We therefore expected that according to the DI measure, G-SYNC would:
Be imperceptible. Participants would not be able to reliably identify the G-SYNC on session.
Table 3. ANOVA of Gaming Performance
effects distribution p value

GS F(1,21) = 4.43 = 0.048
YG F(2,21) = 3.69 = 0.042
GS × GS1 F(1,21) = 5.61 = 0.028
4.5.2 Performance. As reviewed above, prior research provides solid evidence that better temporal
sampling and reduced delay improve human performance. We therefore expected that according to
the gaming performance measure, G-SYNC would:
Improve performance. Participants would be awarded more points by Battlefield 4 with G-SYNC
on than with G-SYNC off.
4.5.3 Experience. Prior research also provides evidence that better temporal display characteristics
improve experience, though it is less conclusive. We therefore expected that according to the explicit
Self Assessment Manikin measure and implicit subjective duration assessment measure, G-SYNC
would:
Improve experience. Participants would report feeling happier, more in control, and (perhaps)
more excited. They would also underestimate the passage of time, indicating they feel
appropriately challenged and more deeply engaged.
While we were confident that feelings of happiness and control were hallmarks of good experi-
ence, we were less certain that feeling excited, stimulated or frenzied were — particularly when the
game was new to most participants.
4.6 Results
In our review of experimental results, we again begin with perceptibility, continue with performance,
and close with experience. We only show significant main effects and two-way interactions.
4.6.1 Perceptibility. Participants correctly identified the G-SYNC display 56% of the time. This was
not significantly different from chance (50%) (χ 2 (1,27) = 0.15, p = .70). G-SYNC could not be reliably
identified by our participants, even though they knew they would make use of it.
4.6.2 Performance. Figure 2’s means show how G-SYNC interacts with G-SYNC first to affect
gaming performance. We analyzed these using a 3-factor mixed analysis of variance (ANOVA), with
G-SYNC within subjects, while G-SYNC first and years of gaming were between subjects. Table 3
shows all significant analysis results. G-SYNC significantly improved gaming performance, with
participants scoring nearly 10% more when G-SYNC was on (mean µ = 2342, standard error of
mean SEM = 139) than when it was off (µ = 2143, SEM = 167). Years of gaming also significantly
affected performance, with experience capable of raising scores by almost 100% (< 1: µ = 1618, SEM
= 203; < 5: µ = 2081, SEM = 125; > 5: µ = 2595, SEM = 164).
The effect of G-SYNC depended significantly on whether it was used in the first or the second
session (G-SYNC first). When participants began playing with G-SYNC on, they performed well
(µ = 2417, SEM = 166), and continued to perform well even when G-SYNC was turned off (µ = 2413,
SEM = 193). When participants started playing with G-SYNC off, they performed quite poorly
(µ = 1866, SEM = 263), but when G-SYNC was later turned on, their performance improved
dramatically (µ = 2273, SEM = 231).
Fig. 2. G-SYNC × G-SYNC first effects on gam- Fig. 3. G-SYNC × G-SYNC first effects on SAM plea-
ing performance, with 95% confidence inter- sure, with 95% confidence intervals.
vals.
Table 4. ANOVA of Gaming Experience
measure effects distribution p value

SAM-P GS1 F(1,21) = 4.41 = 0.048
SAM-P GS × GS1 F(1,21) = 9.64 = 0.0054
SAM-D GS F(1,21) = 3.00 = 0.098
SAM-D GS × YG F(2,21) = 4.14 = 0.031
SDA GS1 F(1,21) = 3.94 = 0.06
4.6.3 Experience. Figure 3’s means show how G-SYNC interacts with G-SYNC first to affect the
SAM’s pleasure dimension. We analyzed the effects of our three independent variables on this
measure using a 3-factor mixed ANOVA, much as we did our gaming performance measure. Table 4
shows significant analysis results. Over the entire experiment, participants experienced a slight
reduction of pleasure (µ = 0.28, SEM = 0.13), but those who started gameplay with G-SYNC on lost
significantly more pleasure (µ = 0.43, SEM = 0.17) than those who started with G-SYNC off (µ = .12,
SEM = .20). This G-SYNC first effect depended significantly on G-SYNC (Figure 3). All participants
experienced less pleasure after the first gameplay session, but while GS1 false participants reclaimed
most of their pleasure after the second gameplay session with G-SYNC on, the experienced pleasure
of GS1 true participants did not change after their second gameplay session with G-SYNC off.
Figure 4’s means show how G-SYNC interacts with years of gaming to affect SAM’s dominance
dimension. We again sought effects using a 3-factor mixed ANOVA. Table 4 shows significant and
marginally significant analysis results. G-SYNC had a marginally significant effect, with participants
experiencing a slight gain of control with G-SYNC on (µ = 0.19, SEM = 0.20), and a very slight
loss of control with G-SYNC off (µ = −0.07, SEM = 0.15). This experience of control depended
significantly on years of gaming, with participants with less than 1 and less than 5 years of gaming
experiencing a drop in control when G-SYNC was off, and an increase of control when G-SYNC
was on. In contrast, participants with more than 5 years of gaming felt no change of control when
G-SYNC was off, and a decline of control when G-SYNC was on.
There were no significant effects on SAM’s arousal dimension, and only one marginally significant
effect on subjective duration assessment. When participants used G-SYNC first, they experienced a
Fig. 4. G-SYNC × years of gaming effects on SAM dominance, with 95% confidence intervals.
very slight time acceleration (µ = −0.36minutes, SEM = 0.85). However, when participants did not
begin the experiment with G-SYNC, they experienced a slowing of time (µ = 3.46minutes, SEM =
1.37), indicating subjective difficulty and a reduced sense of flow.
5 DISCUSSION
Our discussion first spans our experimental hypotheses, and then moves on to implications for
researchers, gamers and technologists.
5.1 Hypotheses
Our hypotheses focused on perceptibility, performance and experience.
5.1.1 Perceptibility. We hypothesized that ASync would be imperceptible, and this proved true in
our experiment. This has important implications for our experiential measures, since ASync imper-
cepthttps://www.overleaf.com/project/5c1a0be6485aff621e3ffd18ibility eliminates the possibility
of conscious bias reacting to any “improved” display. It is interesting to note that when explicitly
asked to identify the G-SYNC display, participants could not accurately do it, but when explicitly
asked for experiential self-assessments, participant responses were reliably related to G-SYNC.
5.1.2 Performance. We also hypothesized that ASync would improve performance, and indeed it
increased gaming performance in Battlefield 4 by 10%. Interestingly, we also found that ASync’s
effect on performance was “sticky”: those who began gameplay with G-SYNC quickly achieved a
high level of performance, and when they continued without it were able to maintain that improved
performance. In contrast, the performance of those who began gameplay without G-SYNC suffered
in comparison (roughly 33% worse), though when G-SYNC was eventually turned on, they achieved
similar performance to those who began with G-SYNC.
One possible explanation of ASync’s “stickiness” is an interaction with learning. In the first
gaming session, participant performance was lower than it would be later. However, participants
using G-SYNC first benefitted from a compensating improvement in performance. In the second
gaming session, those participants lost those benefits, but learning had improved their performance
— leaving their resulting performance unchanged. Participants using G-SYNC second now received
its performance benefits, and were able to achieve the same level of performance as those using it
first. It remains to be seen what might happen if participant learning continued with G-SYNC.
5.1.3 Experience. Our final hypothesis was that ASync would improve experience, at least as by
the SAM’s pleasure and dominance dimensions, and by subjective duration assessment. Our results
here were mixed, and more complex than we anticipated.
ASync in fact reduced self-reported pleasure slightly. Moreover, while all participants felt less
pleasure after the first gaming session, only those who played that session without G-SYNC
then experienced an increase in pleasure. We speculate that learning dominated and reduced the
experience of pleasure in the first session of gameplay, whether or not G-SYNC was on. In the
second session, learning’s effect on pleasure no longer dominated, giving those using G-SYNC in
that session more pleasure than those without G-SYNC. Thus those who played without G-SYNC in
the first session had worse performance then, but experienced more pleasure in the second session.
Those who played with G-SYNC in the first session has better performance then, but experienced
less pleasure in the second session.
ASync produced a marginally reliable, moderate improvement in self-reported dominance.
ASync’s effect on dominance gained clarity when viewed in relation to the years of gameplay
participants reported. The most experienced players actually felt less in control when G-SYNC was
on, and no change in control when it was off. Other players, as we anticipated, felt less in control
when G-SYNC was off, and more in control when it was on. We speculate that seasoned players were
better able to assess their own level of mastery, and so were better able to decouple the experience
of control from the performance benefits of G-SYNC. When G-SYNC was off, it hid performance
failures, and did not affect experienced control. When G-SYNC was on, it exposed these failures, and
reduced their experienced control. Less experienced players confounded G-SYNC’s performance
benefits with their own mastery.
We hypothesized that ASync might increase experienced excitement, but were uncertain, par-
ticularly while players were still at this early stage of gameplay. We found no effect of G-SYNC
on arousal. We also expected that ASync would reduce subjective difficulty and increase flow as
measured by subjective duration assessment. In fact, G-SYNC had no reliable overall impact, though
when used in the first gameplay session, it produced a marginally reliable improvement. In implicit
experience as measured by SDA, G-SYNC may have compensated for the learning difficulties of
the first gameplay session, even as it did not significantly alter explicitly reported pleasure in that
session.
5.2 Implications
We begin discussing the implications of our research with a review of its limitations. We then move
on to its implications for gamers and technologists.
5.2.1 Limitations. Our results should not be over-generalized. Our experiment explored ASync
effects in a small range of modest frame rates, with only one game genre, using participants
somewhat uniform in age and gaming familiarity. In this first experiment, we sought to maximize
and confirm ASync’s impacts by using a modest graphics system with moderate 30Hz frame rates.
Though our system is similar to many platforms still in broad use, we could future-proof and
improve the external validity of our work by examining systems with 60Hz and perhaps 120Hz
frame rates. B4 is an FPS representing one of the most popular game genres. However, there are
other genres that are just as popular, including action, sports and role-playing games. Our players
were uniformly young and though dedicated, not professional or very experienced with B4. We
hope to continue this work both with more experienced gamers quite familiar with the game they
are playing, and with a broader range of gamers.
5.2.2 Gaming implications. Gamers will be glad to have evidence that ASync can improve gaming
performance, though it remains to be seen if this holds true at higher frame rates, as play continues
for extended periods, as gamer expertise reaches peak levels, and ASync is used with games beside
B4. We also found evidence that ASync help particpants learn new games, suggesting that ASync
might help players find a good experience more quickly.
5.2.3 Technological implications. The fact that participants could not reliably identify the ASync
display poses a marketing challenge for NVIDIA and AMD: it might be difficult to sell a product
that does not look like an obvious improvement. ASync manufacturers might instead emphasize
and reproduce the performance benefits we found in our research — or perhaps seek to make ASync
more perceivable, without reducing its performance benefits. Prior research indicates that content
containing significant motion, or latency-sensitive tasks such as the tracking that FPS gamers must
perform to hit moving targets, should make ASync’s benefits more obvious.
6 CONCLUSIONS AND FUTURE WORK

Our work explored the benefits to gaming performance and experience realized by adaptive
synchronization. We found reliable performance benefits, coupled with complex and intriguing
impacts on player experience.
This experiment is an early examination of a complex set of questions, and merits extensive
follow-up. In particular, we plan and encourage others to:
• replicate this study with higher frame rates. While 30Hz gaming systems are still common,
they are becoming less so. Moreover, we expect ASync’s beneficial impact on frame time to
decline as overall system performance improves.
• investigate ASync’s “stickiness”, which enabled players to maintain gaming performance
even after ASync was turned off. One possibility is a between subjects experimental design,
in which participants worked only with ASync, or only without it.
• generalize these results by studying ASync’s impacts on other gaming genres and platforms,
with more experienced gamers, without v-sync, and in more tightly controlled experimental
settings.
Researchers might also employ emerging measures of experience, including biophysical measures
and the implicit association test. It may in fact prove that ASync benefits other applications than
gaming, and should be more widely adopted in the computing world.
ACKNOWLEDGMENTS
Our thanks to I3D’s anonymous reviewers, who provided valuable perspective and suggestions.
Josef Spjut and Joowhan Kim of NVIDIA Research also helped improve the paper, and provided
detailed information about G-SYNC. Early work on this project was performed by undergraduates
supported by an NSF REU Site grant.
REFERENCES
AMD. 2018. Radeon FreeSync Technology. (2018). Retrieved December 19, 2018 from https://www.amd.com/en/technologies/
free-sync
Benjamin Bederson. 2004. Interfaces for staying in the flow. Ubiquity 2004, September (2004), 1–1.
Margaret M Bradley and Peter J Lang. 1994. Measuring emotion: the self-assessment manikin and the semantic differential.
Journal of behavior therapy and experimental psychiatry 25, 1 (1994), 49–59.
J Brown and J Locke. 2018. Here’s every Enhanced game for the PS4 Pro. (2018). Retrieved November 27, 2018 from
https://www.androidcentral.com/every-game-getting-enhanced-treatment-ps4-pro
Jessie YC Chen and Jennifer E Thropp. 2007. Review of low frame rate effects on human performance. IEEE Transactions on
Systems, Man, and Cybernetics-Part A: Systems and Humans 37, 6 (2007), 1063–1076.
Mark Claypool and Kajal Claypool. 2006. Latency and player actions in online games. Commun. ACM 49, 11 (2006), 40–45.
Mary Czerwinski, Eric Horvitz, and Edward Cutrell. 2001. Subjective duration assessment: An implicit probe for software
usability. In Proceedings of IHM-HCI 2001 conference, Vol. 2. 167–170.
Jonathan Deber, Ricardo Jota, Clifton Forlines, and Daniel Wigdor. 2015. How much faster is fast enough?: User perception
of latency & latency improvements in direct and indirect touch. In Proceedings of the 33rd Annual ACM Conference on
Human Factors in Computing Systems. ACM, 1827–1836.
Fraps. 2013. Fraps: real-time video capture & benchmarking. (2013). Retrieved September 17, 2018 from http://www.fraps.
com/
Oliver Hohlfeld, Hannes Fiedler, Enric Pujol, and Dennis Guse. 2016. Insensitivity to Network Delay: Minecraft Gaming
Experience of Casual Gamers. In Teletraffic Congress (ITC 28), 2016 28th International, Vol. 3. IEEE, 31–33.
IGN. 2016. PS4 vs. Xbox One Native Resolutions and Framerates. (2016). Retrieved November 27, 2018 from https:
//www.ign.com/wikis/xbox-one/PS4_vs._Xbox_One_Native_Resolutions_and_Framerates
Benjamin F Janzen and Robert J Teather. 2014. Is 60 FPS better than 30?: the impact of frame rate and latency on moving
target selection. In CHI’14 Extended Abstracts on Human Factors in Computing Systems. ACM, 1477–1482.
Ricardo Jota, Albert Ng, Paul Dietz, and Daniel Wigdor. 2013. How fast is fast enough?: a study of the effects of latency
in direct-touch pointing tasks. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM,
2291–2300.
J Kampman. 2017. Poll: What’s the resolution and refresh rate of your gaming monitor? (2017). Retrieved November 27,
2018 from https://techreport.com/news/31542/poll-what-the-resolution-and-refresh-rate-of-your-gaming-monitor
Steven WK Lee and Rocky KC Chang. 2018. Enhancing the experience of multiplayer shooter games via advanced lag
compensation. In Proceedings of the 9th ACM Multimedia Systems Conference. ACM, 284–293.
Wai-Kiu Lee and Rocky KC Chang. 2015. Evaluation of lag-related configurations in first-person shooter games. In
Proceedings of the 2015 International Workshop on Network and Systems Support for Games. IEEE Press, 11.
Albert Mehrabian and James A Russell. 1974. An approach to environmental psychology. the MIT Press.
Jeanne Nakamura and Mihaly Csikszentmihalyi. 2014. The concept of flow. In Flow and the foundations of positive psychology.
Springer, 239–263.
NVIDIA. 2018a. NVIDIA G-SYNC. (2018). Retrieved September 17, 2018 from https://www.geforce.com/hardware/
technology/g-sync
NVIDIA. 2018b. Supported Games: GeForce Experience. (2018). Retrieved September 17, 2018 from https://www.nvidia.
com/en-us/geforce/geforce-experience/games/
J Papadopoulos. 2017. PC games sales in 2017 are almost as big as all console sales combined. (2017). Retrieved December 18,
2018 from https://www.dsogaming.com/news/pc-games-sales-in-2017-are-almost-as-big-all-console-sales-combined/
Elizabeth M Perse. 2010. Self-assessment manikin. In Communication Research Measures II, A Sourcebook, Rebecca B Rubin,
Alan M Rubin, Elizabeth M Perse, Elizabeth Graham, and David Seibold (Eds.). Routledge, 354–359.
Christian H Poth, Rebecca M Foerster, Christian Behler, Ulrich Schwanecke, Werner X Schneider, and Mario Botsch. 2018.
Ultrahigh temporal resolution of visual presentation using gaming monitors and G-Sync. Behavior research methods 50, 1
(2018), 26–38.
Andrew B Watson. 1986. Temporal sensitivity. In Handbook of perception and human performance, Kenneth R Boff, Lloyd
Kaufman, and James P Thomas (Eds.). Wiley, New York.
Andrew B Watson and Albert J Ahumada. 2011. 64.3: Flicker Visibility: A Perceptual Metric for Display Flicker. In SID
symposium digest of Technical Papers, Vol. 42. Wiley Online Library, 957–959.

Notes On 30hz

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Notes On 30hz

Uploaded by

Copyright:

Available Formats

The Effects of Adaptive Synchronization on Performance

and Experience in Gameplay

BENJAMIN WATSON, North Carolina State University, USA

1 ADAPTIVE SYNC’S PROMISE

2 TEMPORAL GAMING TECHNOLOGY

3 TIME, PERFORMANCE AND EXPERIENCE

4.2.1 Independent Variables. To introduce ASync, we used NVIDIA’s G-SYNC. As a proprietary

Table 1. Participants vs. Independent Variables

Table 2. Frame Times on Experimental Equipment

G-SYNC On G-SYNC Off

4.4 Procedure and Task

Table 3. ANOVA of Gaming Performance

effects distribution p value

Table 4. ANOVA of Gaming Experience

measure effects distribution p value

6 CONCLUSIONS AND FUTURE WORK

You might also like