Graphics processors
GeForce, Radeon, Mali, Adreno and the graphics chips of consoles and arcades. 176 chips on EveryChip so far.
A graphics processing unit (GPU) is a chip built to draw images: it works out the colour of millions of pixels many times a second. Because each pixel can be worked on at the same time as the others, a GPU has thousands of small arithmetic units running in parallel. Nvidia marketed its GeForce 256 of 1999 as the first GPU, and since the late 2000s GPUs have also been used for science and AI.
How a GPU differs from a CPU
A CPU has a few powerful cores tuned to finish one task quickly. A GPU has many simpler cores grouped into blocks, which Nvidia calls streaming multiprocessors and AMD calls compute units. They run the same instruction on many pieces of data at once. GPUs also have very fast memory, such as GDDR or HBM, because they must feed all those units at the same time.
From games to general computing
Early graphics chips had fixed stages for transforming shapes and filling in pixels. From the early 2000s these became programmable shaders, small programs written by game developers. In 2006 Nvidia's Tesla architecture merged them into one unified design, and its CUDA platform let programmers use the GPU for any parallel work. That opened the way to GPU supercomputers and, after 2012, to deep learning.
Famous GPU chips
Graphics Core NextAMD's GPU architecture from 2012Further reading
🇺🇸 United StatesRadeon HD 7000 series
AMD graphics processor family using 28 nm process and GCN architectureGPU · GPU architecture · 2012
🇺🇸 United StatesRadeon RX 400 series
AMD graphics processors using Polaris chips and GCN 4 architectureGPU · GPU architecture · 2016
🇺🇸 United StatesRadeon RX Vega
AMD graphics processors using the fifth generation GCN architectureGPU · GPU architecture · 2017
🇺🇸 United StatesIntel Data Center GPU Max
Data centre GPU for HPC and AI workloadsGPU · GPU architecture
🇺🇸 United States and 1 moreAMD Instinct MI100
CDNA architecture GPU for high-performance computing and AI workloads.GPU · GPU architecture · 2020
🇺🇸 United StatesAMD Instinct MI250X
Dual-GPU accelerator for supercomputersGPU · GPU architecture
🇨🇳 ChinaBiren BR100
AI GPU for data centre compute workloadsGPU · GPU architecture
🇺🇸 United StatesA100
Data centre GPU for AI and high-performance computing.GPU · GPU architecture · 2020
🇺🇸 United StatesIntel 2700G
Mobile graphics co-processor for XScale PXA27xGPU · GPU architecture · 2004
Photo credits on this page
The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.
- GeForce 256: Hyins, Public domain, source
- Mali: Open Grid Scheduler / Grid Engine, CC0, source
- Intel 2700G: Pinchies, CC BY-SA 4.0, source
- Nvidia GRID: Phiarc, CC BY-SA 4.0, source
- Tesla: Hyins, Public domain, source
- RIVA 128: Raimond Spekking, CC BY-SA 4.0, source
- Graphics Core Next: AMD, Public domain, source
- RDNA: Fritzchens Fritz, CC0, source
- Radeon HD 7000 series: Liam McSherry, CC BY-SA 4.0, source
- RIVA TNT: Raimond Spekking, CC BY-SA 4.0, source
- SiS 6326: Henry2106, CC BY-SA 4.0, source
- GeForce 100 series: Sgroey, CC BY-SA 4.0, source
- GeForce 800M series: PantheraLeo1359531, CC BY 4.0, source
- Hopper: 极客湾Geekerwan, CC BY 3.0, source
- Blackwell: 极客湾Geekerwan, CC BY 3.0, source
- Latte: Takimata (edited by: Tokyoship ), CC BY-SA 3.0, source