Scribd
Upload
98 views14 pages

9 - CPU vs. GPU

Central Processing Units (CPUs) are optimized for sequential serial processing while Graphical Processing Units (GPUs) are optimized for parallel processing. GPUs have hundreds of smaller, more efficient cores than CPUs and are better suited to problems that can leverage data parallelism like graphics, physics simulations, and machine learning algorithms. Money drives innovation in computing hardware, so GPUs have advanced rapidly as their markets in computer graphics and gaming have grown enormously in scale.

Uploaded by

Nguyễn Phạm Thiên Bảo
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
98 views14 pages

9 - CPU vs. GPU

Central Processing Units (CPUs) are optimized for sequential serial processing while Graphical Processing Units (GPUs) are optimized for parallel processing. GPUs have hundreds of smaller, more efficient cores than CPUs and are better suited to problems that can leverage data parallelism like graphics, physics simulations, and machine learning algorithms. Money drives innovation in computing hardware, so GPUs have advanced rapidly as their markets in computer graphics and gaming have grown enormously in scale.

Uploaded by

Nguyễn Phạm Thiên Bảo
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

Central Processing Units (CPUs)

vs.
Graphical Processing Units (GPUs)
Edited by Xuan Ly NGUYEN THE

1
Money – root of all innovation

• Money drives computing, as much as technology.


• If there’s a big enough market, someone will develop the product.
• Need huge economies of scale to make chips cheaply, so very few
companies and competing products.
• To anticipate computing trends, look at market drivers, key applications.

2
CPUs
• chip size (amount of circuitry) continues to doubles every 18-24 months (Moore’s
Law)
• similar growth in other hardware aspects, but memory bandwidth struggles to
keep up
• safe to assume that this will continue for at least the next 10 years, driven by:
• multimedia applications (streaming video, HD)
• image / video processing
• “intelligent” software
• internet applications (Google, Bing, Facebook)

3
CPU Innovation

• CPU clock stuck at about 3GHz since 2006 due to high power consumption
(up to 130W per chip)
• chip circuitry still doubling every 18-24 months
⇒ more on-chip memory and MMU (memory management units)
⇒ specialized hardware (e.g. multimedia, encryption)
⇒ multi-core (multiple CPU’s on one chip)

4
Intel’s Tick – Tock Model

5
Intel’s Sandy Bridge Micro-architecture

6
Memory Hierarchy

7
Technical Challenges

• compiler to extract best performance, reordering instructions if necessary


• out-of-order CPU execution to avoid delays waiting for read/write or earlier
operations
• branch prediction to minimize delays due to conditional branching (loops, if-
then-else)
• memory hierarchy to deliver data to registers fast enough to feed the
processor

8
GPUs – the big development

• Economics is again the key:


• produced in vast numbers for computer graphics increasingly being used for
• computer games (e.g. Physics)
• video (e.g. HD video decoding)
• audio (e.g. MP3 encoding)
• multimedia (e.g. Adobe software)
• computational finance
• medical imaging
• computational science

9
GPUs Vendors
• 4 major vendors:
• NVIDIA:
• Gaming (Geforce)
• Design and Creation (Quadro)
• High Performance Computing – HPC (Tesla)
• AMD
• Gaming (Radeon)
• Design and Creation, HPC (FirePro)
• IBM
• co-developed Cell processor with Sony and Toshiba for Sony PlayStation, but now dropped
it for HPC
• Intel
• Integrated graphics (e.g. HD4000)
10
GPU and CPU Communication

Type of connection Between Bandwidth (GB/s)

Intel Front Side Bus CPU-RAM 2 to 12.8

Intel Memory Controller CPU-RAM 25.6 to 38.4

AGP CPU-GPU 2.1

PCIe 1.1/2.0/3.0 CPU-GPU 4/8/16

AMD VRAM bus AMD GPU-VRAM 320

NVIDIA VRAM bus NVIDIA GPU-VRAM 336

11
Intel Core i7 vs. NVIDIA GTX
CPU: Intel “Sandy Bridge” Core i7 GPU: NVIDIA GTX 580

 4 MIMD cores  512 cores, arranged as 16 units each with 32 SIMD cores
 few registers, multilevel caches  lots of registers, almost no cache
 30 GB/s bandwidth to main memory  5 GB/s bandwidth to host processor (PCIe x16 gen 2)
 190 GB/s bandwidth to graphics memory

12
Intel Core i7 vs. AMD R9

13
CPUs and GPUs Performance

14

You might also like