Professional Documents
Culture Documents
ABSTRACT
GPU, the Graphics Processing Unit, is one of the most important types of
computer technology. GPU has impacted both the personal and business
computing sectors. GPU was initially designed for parallel processing, but it
became the root element of several other applications, such as rendering
graphics and videos, from where the name of this unit stuck. Nowadays, we
have been made to think that the most important purpose of GPU is in the
gaming industry. Still, as the capabilities of GPU are being harnessed further,
the applications such as creative production and implementations in Machine
Learning are coming to light.
Originally the GPU was designed for the accelerated rendering of 3D graphics
and video elements. But as the technology grew, it was made more flexible
and programmable to enhance the GPU performance. This enhancement
allowed graphics programmers to create more interesting visual effects and
realistic scenes with advanced lighting and shadowing techniques.
Since the rendering of 3D graphics required high computing resources, the
GPU was made much more capable than the regular processor. To exploit the
feature of HPC, AI developers used the GPU to perform the high-demanding
calculations to be done by the GPU. Eventually, GPU has become an
important part of implementations of various AI fields such as Deep
Learning(DL)
I. INTRODUCTION
GPUs used nowadays are very fast and efficient at manipulating digital graphics and
processing images. The general-purpose CPU (Central Processing Unit) is not as
capable of processing data parallelly as much as the GPU. Hence the GPUs can
process large blocks of data in parallel. Usually, a GPU is present on a video card or
embedded on the motherboard of a personal computer.
The earliest research related to GPU can be found in the 1970s. The term "GPU"
was referred to “graphics processor unit” which was defined as a programmable
processing unit that worked independently from the CPU and was responsible for
manipulating graphics on the output device. In 1994, Sony used the term GPU for
“Graphics Processing Unit” in reference to the PlayStation console's Toshiba-
designed Sony GPU.
The term GPU was popularized in 1999 when Nvidia marketed their GeForce 256 as
"the world's first GPU." It was presented as a "single-chip processor with integrated
transform, lighting, triangle setup/clipping, and rendering engines."
A rival company to the Nvidia, ATI Technologies coined the term "visual processing
unit" or VPU with the release of the Radeon 9700 in 2002.
In this report, different types of GPU and their components will be discussed in
detail.
From various Internet Sources and Wikipedia, there are the following components of
a Graphics Processing Unit:
Wikipedia resources define the memory controller as “A digital circuit that manages
the flow of data going to and from the computer's main memory.” This circuit can be
integrated into the main CPU chip (called an Integrated Memory Controller) or be a
separate chip itself. The most common form of a Memory Controller is used in the
Operating System as the Memory Management Unit (MMU) for Virtual Addressing.
The memory controller’s main purpose is to store the logic to read and write to the
Dynamic RAM (DRAM) and refresh the DRAM's memory. In the GPU, it has a similar
purpose, and it contains the VRAM (Video RAM) and the logic to control its content.
VRAM is the specialized memory that has the application of storing image data for
computer display. The sub-components of GMC:
1. VRAM -
(Video RAM - this ensures that the graphics display has even and smooth
execution.)
2. WRAM -
(Window RAM - better than VRAM and less expensive; it has features that
make it more efficient to read data for use in block fills and text drawing)
3. MDRAM -
(Multibank Dynamic RAM - High-performance RAM; Divides the total memory
components of 32 KB which can be accessed individually)
4. DDR -
(Double Data Rate Synchronous Dynamic Random-Access Memory or DDR-
SDRAM; This type of SDRAM can send and receive signals twice per clock
cycle.)
5. GDDR -
(Graphics Double Data Rate; This type of DDR is specialized in fast rendering
on GPU)
6. HBM -
(High Bandwidth Memory; Used in 3D-stacked DRAM. It is divided into
multiple banks of memory, vertically stacks memory chips to reduce commute
time in the processor)
The major function of this unit is to use programmable elements that can store data
for the GPU and process them using generic classes of algorithms. They are much
more efficient than general-purpose architectures.
C. VGA BIOS -
This is a type of software. VGA BIOS software is used to initialize a VGA (Video
Graphics Array) card. It is an extension of the system BIOS. It is loaded by the
system BIOS from an expansion ROM located on the VGA card into a specific
address in system memory.
Display Interface aims to fetch frame buffer data from system memory and transmit
them to the display.
Fig 4 - Display Interface
Image_Source=link
Wikipedia defines this unit as a “GPU Unit specialized for video encoding and
decoding.” This unit encodes and decodes video/image display data to and from
GMC and DIF.
G. Compression Unit -
This unit compresses data coming from the GCA to the GMC to improve the
performance of the GPU.
This unit’s purpose is to monitor, control, manage and regulate the power of the
complete GPU.
III. METHODOLOGIES
This case study is a part of a project on the subject of Computer Architecture (CA).
We chose this topic as the working and functioning of the Graphics Processing Unit,
which is so widely used, intrigued us.
We first studied the topic and understood the task, which was to explore different
types of GPU available and the components of a GPU.
We took the help of various internet resources mentioned in the references and took
the important points that needed to be emphasized.
We took some important definitions from various sources and mentioned them in the
document. There were excellent references for different components of a graphics
processing unit we used in the study. Many such resources were used, and their
sources were properly mentioned.
Finally, we compiled all the information and data that we collected and represented
them as best as possible.
VI. CONCLUSION -
In the conclusion of this case study, we can say that GPU technology has come very
far with time. Now modern GPU technology is going to play a vital role in future Deep
Learning Algorithms. The high-computing capability of the GPU makes it very distinct
from a general Central Processing Unit (CPU). Its capability to process data with the
parallel pipelining has made it a vital component in applications requiring high-
performance computing (HPC) such as gaming consoles, rendering 3D graphics,
creative development, etc.
We also saw each and every component of the GPU in detail and learnt their
purpose in the GPU’s working. This case study can help students or scholars who
want to have GPU's basic idea, different types of GPU, and its components.
VII. REFERENCES
1. https://www.intel.com/content/www/us/en/products/docs/processors/what-is-a-
gpu.html
2. https://en.wikipedia.org/wiki/Graphics_processing_unit
3. https://xoticpc.com/blogs/news/history-of-gpus#:~:text=Back%20in
%201999%2C%20NVIDIA%20popularized,card%20to%20rule%20them
%20all.
4. https://www.intel.com/content/www/us/en/products/docs/processors/what-is-a-
gpu.html#:~:text=GPUs%20come%20in%20two%20basic,to%20a%20PCI
%20Express%20slot
5. https://en.wikipedia.org/wiki/Graphics_card
6. https://en.wikipedia.org/wiki/Memory_controller
7. https://www.techtarget.com/searchstorage/definition/video-
RAM#:~:text=Multibank%20DRAM%20(MDRAM)%20is%20a,is%20accessed
%20at%20one%20time.
8. https://en.wikipedia.org/wiki/Video_RAM_(dual-ported_DRAM)
9. https://www.techtarget.com/searchstorage/definition/WRAM
10. https://en.wikipedia.org/wiki/GDDR_SDRAM
11. https://en.wikipedia.org/wiki/DDR_SDRAM
12. https://dl.acm.org/doi/10.1145/319120.319125
13. https://www.usenix.org/legacyurl/freevga-architecture-independent-video-
graphics-initialization-linuxbios1
14. https://en.wikipedia.org/wiki/Video_BIOS
15. https://www.sciencedirect.com/topics/computer-science/display-interface
16. https://en.wikipedia.org/wiki/Vision_processing_unit
17. https://en.wikipedia.org/wiki/Power_Management_Unit
18. https://www.google.com/url?sa=i&url=https%3A%2F
%2Fwww.researchgate.net%2Ffigure%2FComparison-between-CPU-and-
GPU-processing-time-of-various-blocks-for-the-JPEG-
XR_tbl2_241499253&psig=AOvVaw1apBpWsIzE50qCqpRcG1Sg&ust=16508
93919189000&source=images&cd=vfe&ved=0CAwQjRxqFwoTCKjx2qDprPc
CFQAAAAAdAAAAABAD
19. https://www.google.com/url?sa=i&url=https%3A%2F%2Fresolume.com
%2Fforum%2Fviewtopic.php%3Ft
%3D11301&psig=AOvVaw1jZvaQjmqKzHVt6_Gcd5Va&ust=1650894278688
000&source=images&cd=vfe&ved=0CAwQjRxqFwoTCLD62srqrPcCFQAAAA
AdAAAAABAK
20. https://www.google.com/url?sa=i&url=https%3A%2F
%2Fwww.researchgate.net%2Ffigure%2FFlowchart-of-the-differences-
between-utilizing-a-CPU-versus-a-GPU-in-converting-a-
single_fig2_49798091&psig=AOvVaw14F1hE_jxrLeKG_s1sCyrT&ust=16508
94758354000&source=images&cd=vfe&ved=0CAwQjRxqFwoTCOC6qbDsrP
cCFQAAAAAdAAAAABAS
21. https://www.google.com/url?sa=i&url=https%3A%2F
%2Fwww.researchgate.net%2Ffigure%2FProgram-flow-chart-of-SIMPLE-on-
GPU_fig2_317486084&psig=AOvVaw2DbwGqdMlg-
YLpF_iw7epS&ust=1650894805880000&source=images&cd=vfe&ved=2ahU
KEwiw3__E7Kz3AhUMZ2wGHYvyASUQjRx6BAgAEAs