GPU model | Polaris 22 | GP100 |
---|---|---|
GPU variant | VegaM MGL XL | GP100-890-A1 |
Architecture | GCN 4.0 | Pascal |
Foundry | GlobalFoundries | TSMC |
Manufacturing process | 14 nm | 16 nm |
Number of transistors | 5,000 million | 15,300 million |
Die size | 208 mm² | 610 mm² |
Release date | Apr 24th, 2018 | — |
---|---|---|
Generation | Vega (Vega M) | — |
Production | Active | — |
Interface | IGP | — |
Base clock | 931 MHz | 1328 MHz |
---|---|---|
Boost clock | 1011 MHz | 1480 MHz |
Memory clock | 700 MHz 1400 Mbps effective | 715 MHz 1430 Mbps effective |
Memory size | 4 GB | 16 GB |
---|---|---|
Memory type | HBM2 | HBM2 |
Memory bus width | 1024 bit | 4096 bit |
Bandwidth | 179.2 GB/s | 732.2 GB/s |
Shading units | 1280 | 3584 |
---|---|---|
TMUs | 80 | 224 |
ROPs | 32 | 96 |
Compute units | 20 | — |
Cache L1 | 16 KB (per CU) | 24 KB (per SM) |
Cache L2 | 1024 KB | 4 MB |
SM count | — | 56 |
Pixel rate | 32.35 GPixel/s | 142.1 GPixel/s |
---|---|---|
Texture rate | 80.88 GTexel/s | 331.5 GTexel/s |
FP16 (half) performance | 2.588 TFLOPS (1:1) | 21.22 TFLOPS (2:1) |
FP32 (float) performance | 2.588 TFLOPS | 10.61 TFLOPS |
FP64 (double) performance | 161.8 GFLOPS (1:16) | 5.304 TFLOPS (1:2) |
Slot width | IGP | — |
---|---|---|
TDP | 65 W | 300 W |
Outputs | No outputs | No outputs |
Board number | D136 | — |
Suggested PSU | — | 700 W |
Power connectors | — | None |
DirectX | 12 (12_0) | 12 (12_1) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.1 | 3.0 |
Vulkan | 1.2 | 1.2 |
Shader model | 6.4 | 6.4 |
CUDA | — | 6.0 |
Release date | — | Apr 5th, 2016 |
---|---|---|
Generation | — | Tesla |
Production | — | End-of-life |
Interface | — | PCIe 3.0 x16 |