GPU model | Baffin | TU102 |
---|---|---|
GPU variant | Baffin LE | TU102-875-A1 |
Architecture | GCN 4.0 | Turing |
Foundry | GlobalFoundries | TSMC |
Manufacturing process | 14 nm | 12 nm |
Number of transistors | 3,000 million | 18,600 million |
Die size | 123 mm² | 754 mm² |
Release date | Oct 13th, 2017 | Unknown |
---|---|---|
Generation | Polaris | GRID |
Production | Active | Active |
Interface | PCIe 3.0 x8 | PCIe 3.0 x16 |
Predecessor | Arctic Islands | — |
Successor | Vega | — |
Base clock | 1019 MHz | 1065 MHz |
---|---|---|
Boost clock | 1071 MHz | 1395 MHz |
Memory clock | 1500 MHz 6 Gbps effective | 1750 MHz 14 Gbps effective |
Memory size | 2 GB | 2 GB |
---|---|---|
Memory type | GDDR5 | GDDR6 |
Memory bus width | 128 bit | 384 bit |
Bandwidth | 96.00 GB/s | 672.0 GB/s |
Shading units | 640 | 4608 |
---|---|---|
TMUs | 40 | 288 |
ROPs | 16 | 96 |
Compute units | 10 | — |
Cache L1 | 16 KB (per CU) | 64 KB (per SM) |
Cache L2 | 1024 KB | 6 MB |
SM count | — | 72 |
Tensor cores | — | 576 |
RT cores | — | 72 |
Pixel rate | 17.14 GPixel/s | 133.9 GPixel/s |
---|---|---|
Texture rate | 42.84 GTexel/s | 401.8 GTexel/s |
FP16 (half) performance | 1,371 GFLOPS (1:1) | 25.71 TFLOPS (2:1) |
FP32 (float) performance | 1,371 GFLOPS | 12.86 TFLOPS |
FP64 (double) performance | 85.68 GFLOPS (1:16) | 401.8 GFLOPS (1:32) |
Slot width | Dual-slot | Dual-slot |
---|---|---|
Length | 145 mm 5.7 inches | 267 mm 10.5 inches |
TDP | 60 W | 260 W |
Suggested PSU | 250 W | 600 W |
Outputs | 1x DVI1x HDMI1x DisplayPort | No outputs |
Power connectors | None | 1x 6-pin + 1x 8-pin |
Board number | D090-21 | PG150 SKU 215 |
DirectX | 12 (12_0) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.1 | 3.0 |
Vulkan | 1.2 | 1.2 |
Shader model | 6.4 | 6.6 |
CUDA | — | 7.5 |