GPU model | GP100 | TU102 |
---|---|---|
Architecture | Pascal | Turing |
Foundry | TSMC | TSMC |
Manufacturing process | 16 nm | 12 nm |
Number of transistors | 15,300 million | 18,600 million |
Die size | 610 mm² | 754 mm² |
GPU variant | — | TU102-875-A1 |
Release date | Oct 1st, 2016 | Aug 13th, 2018 |
---|---|---|
Generation | Quadro | Quadro |
Production | End-of-life | Active |
Interface | PCIe 3.0 x16 | PCIe 3.0 x16 |
Launch price | — | 6,299 USD |
Base clock | 1304 MHz | 1305 MHz |
---|---|---|
Boost clock | 1442 MHz | 1560 MHz |
Memory clock | 715 MHz 1430 Mbps effective | 1750 MHz 14 Gbps effective |
Memory size | 16 GB | 24 GB |
---|---|---|
Memory type | HBM2 | GDDR6 |
Memory bus width | 4096 bit | 384 bit |
Bandwidth | 732.2 GB/s | 672.0 GB/s |
Shading units | 3584 | 4608 |
---|---|---|
TMUs | 224 | 288 |
ROPs | 96 | 96 |
SM count | 56 | 72 |
Cache L1 | 24 KB (per SM) | 64 KB (per SM) |
Cache L2 | 4 MB | 6 MB |
Tensor cores | — | 576 |
RT cores | — | 72 |
Pixel rate | 138.4 GPixel/s | 149.8 GPixel/s |
---|---|---|
Texture rate | 323.0 GTexel/s | 449.3 GTexel/s |
FP16 (half) performance | 20.67 TFLOPS (2:1) | 28.75 TFLOPS (2:1) |
FP32 (float) performance | 10.34 TFLOPS | 14.38 TFLOPS |
FP64 (double) performance | 5.168 TFLOPS (1:2) | 449.3 GFLOPS (1:32) |
Slot width | Dual-slot | Dual-slot |
---|---|---|
Length | 267 mm 10.5 inches | 267 mm 10.5 inches |
Width | 111 mm 4.4 inches | 111 mm 4.4 inches |
TDP | 235 W | 260 W |
Suggested PSU | 550 W | 600 W |
Outputs | 1x DVI4x DisplayPort | No outputs |
Power connectors | 1x 8-pin | 1x 6-pin + 1x 8-pin |
Board number | — | PG150 SKU 510 |
DirectX | 12 (12_1) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 3.0 | 3.0 |
Vulkan | 1.2 | 1.2 |
CUDA | 6.0 | 7.5 |
Shader model | 6.4 | 6.6 |