GPU model | GA102 | TU102 |
---|---|---|
GPU variant | GA102-350-A1 | TU102-875-A1 |
Architecture | Ampere | Turing |
Foundry | Samsung | TSMC |
Manufacturing process | 8 nm | 12 nm |
Number of transistors | 28,300 million | 18,600 million |
Die size | 628 mm² | 754 mm² |
Release date | Unknown | Aug 13th, 2018 |
---|---|---|
gpu.details.availability | 2022 | — |
Generation | GeForce 30 | Quadro |
Predecessor | GeForce 20 | — |
Production | Unreleased | Active |
Launch price | 1,699 USD | 6,299 USD |
Interface | PCIe 4.0 x16 | PCIe 3.0 x16 |
Reviews | 55 in our database | — |
Base clock | 1560 MHz | 1440 MHz |
---|---|---|
Boost clock | 1860 MHz | 1770 MHz |
Memory clock | 1325 MHz 21.2 Gbps effective | 1750 MHz 14 Gbps effective |
Memory size | 24 GB | 24 GB |
---|---|---|
Memory type | GDDR6X | GDDR6 |
Memory bus width | 384 bit | 384 bit |
Bandwidth | 1,018 GB/s | 672.0 GB/s |
Shading units | 10752 | 4608 |
---|---|---|
TMUs | 336 | 288 |
ROPs | 112 | 96 |
SM count | 84 | 72 |
Tensor cores | 336 | 576 |
RT cores | 84 | 72 |
Cache L1 | 128 KB (per SM) | 64 KB (per SM) |
Cache L2 | 6 MB | 6 MB |
Pixel rate | 208.3 GPixel/s | 169.9 GPixel/s |
---|---|---|
Texture rate | 625.0 GTexel/s | 509.8 GTexel/s |
FP16 (half) performance | 40.00 TFLOPS (1:1) | 32.62 TFLOPS (2:1) |
FP32 (float) performance | 40.00 TFLOPS | 16.31 TFLOPS |
FP64 (double) performance | 625.0 GFLOPS (1:64) | 509.8 GFLOPS (1:32) |
Slot width | Triple-slot | Dual-slot |
---|---|---|
Length | 336 mm 13.2 inches | 267 mm 10.5 inches |
Width | 140 mm 5.5 inches | 111 mm 4.4 inches |
Height | 61 mm 2.4 inches | — |
TDP | 450 W | 260 W |
Suggested PSU | 850 W | 600 W |
Outputs | 1x HDMI3x DisplayPort | 4x DisplayPort1x USB Type-C |
Power connectors | 1x 12-pin | 1x 6-pin + 1x 8-pin |
Board number | PG132 SKU | PG150 SKU 510 |
DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
---|---|---|
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 |