Modern GPU Architecture
Total Page:16
File Type:pdf, Size:1020Kb
Modern GPU Architecture Msc Software Engineering student of Tartu University Ismayil Tahmazov [email protected] Agenda • What is a GPU? • GPU evolution!! • GPU costs • GPU pipeline • Fermi architecture • Kepler architecture • Pascal architecture • Turing architecture • AMD Radeon R9 • AMD vs Nvidia What is it a GPU? A graphics processing unit is a specialized electronic circuit designed to rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display device. photo reference Wikipedia What is it a GPU? GPU Evolution image:reference Evolution of Nvidia GeForce 1999 - 2018 Modern GPU First GPU Nvidia GeForce 256 GeForce GTX 1050 CUDA CUDA is a parallel computing platform and programming model developed by NVIDIA for general computing on graphical processing units (GPUs). With CUDA, developers are able to dramatically speed up computing applications by harnessing the power of GPUs. text source photo reference Graphics Processing image:refence source Nvidia GeForce 8880 GPU pipeline Why GPUs so fast ? Why GPUs so fast ? simplified Nvidia GPU architecture MP = Multi Processor SM = Shared Memory SFU = Special Functions Unit IU = Instruction Unit SP = Streaming processor reference Nvidia Fermi GPU Architecture Nvidia Fermi GPU Architecture photo Nvidia Kepler GPU Architecture reference Specifications reference for table Pascal architecture reference GP100 (Tesla P100) image Pascal Streaming Multiprocessor Compute Capabilities: GK110 vs GM200 vs GP100 Cross-section Illustrating GP100 adjacent HBM2 stacks Cross-section Photomicrograph of a P100 HBM2 stack and GP100 GPU image reference image refence Nvidia Turing GPU Architecture reference Nvidia Turing TU102 GPU reference NVIDIA Pascal GP102 vs Turing TU102 reference Turing Streaming Multiprocessor (SM) reference New shared memory architecture reference Turing Shading Performance reference Costs !!! Really? image image Conclusion We have several GPU architectures from different companies. Which is the best option? This depends on your PC and budget!!! For the last 10 years GPUs graphically evaluated and we now see this magic. References: https://www.notebookcheck.net/Nvidia-Pascal-Architecture-Overview.165493.0.html https://wccftech.com/amd-radeon-r9-290-hawaii-gpu-block-diagram-pictured-detailed/ https://en.wikipedia.org/wiki/Pascal_(microarchitecture) https://images.nvidia.com/content/pdf/tesla/whitepaper/pascal-architecture-whitepaper.pdf http://s08.idav.ucdavis.edu/luebke-nvidia-gpu-architecture.pdf https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/technologies/turing-architecture/NVIDIA-Turin g-Architecture-Whitepaper.pdf https://developer.nvidia.com/cuda-zone https://www.youtube.com/watch?v=lGefnd7Fmmo&t=471s https://www.youtube.com/watch?v=cvcAjgMUPUA https://en.wikipedia.org/wiki/Fermi_(microarchitecture) https://www.nvidia.com/content/PDF/kepler/NVIDIA-kepler-GK110-architecture-whitepaper.pdf https://www.youtube.com/watch?v=1ZMlJza5vwU .