GPU model | GA100 | GP104 |
---|---|---|
Architecture | Ampere | Pascal |
Foundry | TSMC | TSMC |
Manufacturing process | 7 nm | 16 nm |
Number of transistors | 54,200 million | 7,200 million |
Die size | 826 mm² | 314 mm² |
GPU variant | — | GP104-850-A1 |
Release date | Apr 12th, 2021 | Feb 6th, 2017 |
---|---|---|
gpu.details.availability | 2021 | — |
Generation | Tesla | Quadro |
Production | Active | End-of-life |
Interface | PCIe 4.0 x16 | PCIe 3.0 x16 |
Launch price | — | 815 USD |
Base clock | 930 MHz | 1202 MHz |
---|---|---|
Boost clock | 1440 MHz | 1480 MHz |
Memory clock | 1215 MHz 2.4 Gbps effective | 1901 MHz 7.6 Gbps effective |
Memory size | 24 GB | 8 GB |
---|---|---|
Memory type | HBM2 | GDDR5 |
Memory bus width | 3072 bit | 256 bit |
Bandwidth | 933.1 GB/s | 243.3 GB/s |
Shading units | 3584 | 1792 |
---|---|---|
TMUs | 224 | 112 |
ROPs | 96 | 64 |
SM count | 56 | 14 |
Tensor cores | 224 | — |
Cache L1 | 192 KB (per SM) | 48 KB (per SM) |
Cache L2 | 24 MB | 2 MB |
Pixel rate | 138.2 GPixel/s | 94.72 GPixel/s |
---|---|---|
Texture rate | 322.6 GTexel/s | 165.8 GTexel/s |
FP16 (half) performance | 10.32 TFLOPS (1:1) | 82.88 GFLOPS (1:64) |
FP32 (float) performance | 10.32 TFLOPS | 5.304 TFLOPS |
FP64 (double) performance | 5.161 TFLOPS (1:2) | 165.8 GFLOPS (1:32) |
Slot width | Dual-slot | Single-slot |
---|---|---|
Length | 267 mm 10.5 inches | 241 mm 9.5 inches |
Width | 112 mm 4.4 inches | 111 mm 4.4 inches |
TDP | 165 W | 105 W |
Suggested PSU | 450 W | 300 W |
Outputs | No outputs | 4x DisplayPort |
Power connectors | 8-pin EPS | 1x 6-pin |
Board number | P1001 SKU 205 | PG410 SKU 501 |
DirectX | — | 12 (12_1) |
---|---|---|
OpenGL | — | 4.6 |
OpenCL | 3.0 | 3.0 |
Vulkan | — | 1.2 |
CUDA | 8.0 | 6.1 |
Shader model | — | 6.4 |