GPU model | GA100 | GM20B |
---|---|---|
Architecture | Ampere | Maxwell 2.0 |
Foundry | TSMC | TSMC |
Manufacturing process | 7 nm | 20 nm |
Number of transistors | 54,200 million | 2,000 million |
Die size | 826 mm² | 118 mm² |
GPU variant | — | TM660M-A2 |
Release date | Jun 28th, 2021 | — |
---|---|---|
Generation | Tesla | — |
Production | Active | — |
Interface | PCIe 4.0 x16 | — |
Base clock | 1065 MHz | 640 MHz |
---|---|---|
Boost clock | 1410 MHz | 921 MHz |
Memory clock | 1593 MHz 3.2 Gbps effective | System Shared |
Memory size | 80 GB | System Shared |
---|---|---|
Memory type | HBM2e | System Shared |
Memory bus width | 5120 bit | System Shared |
Bandwidth | 2,039 GB/s | System Dependent |
Shading units | 6912 | 128 |
---|---|---|
TMUs | 432 | 16 |
ROPs | 160 | 16 |
SM count | 108 | — |
Tensor cores | 432 | — |
Cache L1 | 192 KB (per SM) | 48 KB (per SMM) |
Cache L2 | 80 MB | 256 KB |
SMM count | — | 1 |
Pixel rate | 225.6 GPixel/s | 14.74 GPixel/s |
---|---|---|
Texture rate | 609.1 GTexel/s | 14.74 GTexel/s |
FP16 (half) performance | 77.97 TFLOPS (4:1) | 471.6 GFLOPS (2:1) |
FP32 (float) performance | 19.49 TFLOPS | 235.8 GFLOPS |
FP64 (double) performance | 9.746 TFLOPS (1:2) | 7.368 GFLOPS (1:32) |
Slot width | Dual-slot | IGP |
---|---|---|
Length | 267 mm 10.5 inches | 70 mm 2.8 inches |
TDP | 250 W | 10 W |
Suggested PSU | 600 W | — |
Outputs | No outputs | No outputs |
Power connectors | 8-pin EPS | — |
Board number | P1001 SKU 200 | — |
Width | — | 45 mm 1.8 inches |
DirectX | — | 12 (12_1) |
---|---|---|
OpenGL | — | 4.6 |
OpenCL | 3.0 | 1.2 |
Vulkan | — | 1.1 |
CUDA | 8.0 | 5.3 |
Shader model | — | 6.4 |
Release date | — | Mar 2019 |
---|---|---|
Generation | — | Tegra |
Production | — | End-of-life |
Interface | — | IGP |