GPU model | RV670 | GA100 |
---|---|---|
Architecture | TeraScale | Ampere |
Foundry | TSMC | TSMC |
Manufacturing process | 55 nm | 7 nm |
Number of transistors | 666 million | 54,200 million |
Die size | 192 mm² | 826 mm² |
Release date | Nov 8th, 2007 | May 14th, 2020 |
---|---|---|
Generation | FireStream | GRID |
Production | End-of-life | Active |
Interface | PCIe 2.0 x16 | PCIe 4.0 x16 |
GPU clock | 800 MHz | — |
---|---|---|
Memory clock | 800 MHz 1600 Mbps effective | 1215 MHz 2.4 Gbps effective |
Base clock | — | 900 MHz |
Boost clock | — | 1005 MHz |
Memory size | 2 GB | 48 GB |
---|---|---|
Memory type | GDDR3 | HBM2E |
Memory bus width | 256 bit | 6144 bit |
Bandwidth | 51.20 GB/s | 1,866 GB/s |
Shading units | 320 | 6912 |
---|---|---|
TMUs | 16 | 432 |
ROPs | 16 | 192 |
Compute units | 4 | — |
Cache L2 | 256 KB | 48 MB |
SM count | — | 108 |
Tensor cores | — | 432 |
Cache L1 | — | 192 KB (per SM) |
Pixel rate | 12.80 GPixel/s | 193.0 GPixel/s |
---|---|---|
Texture rate | 12.80 GTexel/s | 434.2 GTexel/s |
FP32 (float) performance | 512.0 GFLOPS | 13.89 TFLOPS |
FP64 (double) performance | 102.4 GFLOPS (1:5) | 6.947 TFLOPS (1:2) |
FP16 (half) performance | — | 55.57 TFLOPS (4:1) |
Slot width | Dual-slot | IGP |
---|---|---|
Length | 241 mm 9.5 inches | — |
Width | 111 mm 4.4 inches | — |
TDP | 105 W | 400 W |
Suggested PSU | 300 W | 800 W |
Outputs | 2x DVI1x S-Video | No outputs |
Power connectors | 1x 6-pin | None |
DirectX | 10.1 (10_1) | — |
---|---|---|
OpenGL | 3.3 (full) 4.0 (partial) | — |
OpenCL | — | 3.0 |
Vulkan | — | — |
Shader model | 4.1 | — |
CUDA | — | 8.0 |