GPU model | GA104 | TU117 |
---|---|---|
GPU variant | GA104-200-A1 | — |
Architecture | Ampere | Turing |
Foundry | Samsung | TSMC |
Manufacturing process | 8 nm | 12 nm |
Number of transistors | 17,400 million | 4,700 million |
Die size | 392 mm² | 200 mm² |
Release date | Dec 1st, 2020 | May 6th, 2021 |
---|---|---|
gpu.details.availability | Dec 2nd, 2020 | — |
Generation | GeForce 30 | Quadro |
Predecessor | GeForce 20 | — |
Production | Active | Active |
Launch price | 399 USD | — |
Interface | PCIe 4.0 x16 | PCIe 3.0 x16 |
Reviews | 64 in our database | — |
Base clock | 1410 MHz | 1065 MHz |
---|---|---|
Boost clock | 1665 MHz | 1395 MHz |
Memory clock | 1750 MHz 14 Gbps effective | 1250 MHz 10 Gbps effective |
Memory size | 8 GB | 4 GB |
---|---|---|
Memory type | GDDR6 | GDDR6 |
Memory bus width | 256 bit | 128 bit |
Bandwidth | 448.0 GB/s | 160.0 GB/s |
Shading units | 4864 | 896 |
---|---|---|
TMUs | 152 | 56 |
ROPs | 80 | 32 |
SM count | 38 | 14 |
Tensor cores | 152 | — |
RT cores | 38 | — |
Cache L1 | 128 KB (per SM) | 64 KB (per SM) |
Cache L2 | 4 MB | 1024 KB |
Pixel rate | 133.2 GPixel/s | 44.64 GPixel/s |
---|---|---|
Texture rate | 253.1 GTexel/s | 78.12 GTexel/s |
FP16 (half) performance | 16.20 TFLOPS (1:1) | 5.000 TFLOPS (2:1) |
FP32 (float) performance | 16.20 TFLOPS | 2.500 TFLOPS |
FP64 (double) performance | 253.1 GFLOPS (1:64) | 78.12 GFLOPS (1:32) |
Slot width | Dual-slot | Single-slot |
---|---|---|
Length | 242 mm 9.5 inches | — |
Width | 112 mm 4.4 inches | — |
TDP | 200 W | 50 W |
Suggested PSU | 550 W | 250 W |
Outputs | 1x HDMI3x DisplayPort | 4x mini-DisplayPort |
Power connectors | 1x 12-pin | None |
Board number | PG142 SKU 20 | — |
DirectX | 12 Ultimate (12_2) | 12 (12_1) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 3.0 | 3.0 |
Vulkan | 1.2 | 1.2 |
CUDA | 8.6 | 7.5 |
Shader model | 6.6 | 6.6 |