GPU model | GP108 | TU102 |
---|---|---|
GPU variant | GP108-300-A1 | TU102-875-A1 |
Architecture | Pascal | Turing |
Foundry | Samsung | TSMC |
Manufacturing process | 14 nm | 12 nm |
Number of transistors | 1,800 million | 18,600 million |
Die size | 74 mm² | 754 mm² |
Release date | May 17th, 2017 | Unknown |
---|---|---|
Generation | GeForce 10 | GRID |
Predecessor | GeForce 900 | — |
Successor | GeForce 20 | — |
Production | End-of-life | Active |
Launch price | 79 USD | — |
Interface | PCIe 3.0 x4 | PCIe 3.0 x16 |
Reviews | 6 in our database | — |
Base clock | 1228 MHz | 1065 MHz |
---|---|---|
Boost clock | 1468 MHz | 1395 MHz |
Memory clock | 1502 MHz 6 Gbps effective | 1750 MHz 14 Gbps effective |
Memory size | 2 GB | 8 GB |
---|---|---|
Memory type | GDDR5 | GDDR6 |
Memory bus width | 64 bit | 384 bit |
Bandwidth | 48.06 GB/s | 672.0 GB/s |
Shading units | 384 | 4608 |
---|---|---|
TMUs | 24 | 288 |
ROPs | 16 | 96 |
SM count | 3 | 72 |
Cache L1 | 48 KB (per SM) | 64 KB (per SM) |
Cache L2 | 512 KB | 6 MB |
Tensor cores | — | 576 |
RT cores | — | 72 |
Pixel rate | 23.49 GPixel/s | 133.9 GPixel/s |
---|---|---|
Texture rate | 35.23 GTexel/s | 401.8 GTexel/s |
FP16 (half) performance | 17.62 GFLOPS (1:64) | 25.71 TFLOPS (2:1) |
FP32 (float) performance | 1,127 GFLOPS | 12.86 TFLOPS |
FP64 (double) performance | 35.23 GFLOPS (1:32) | 401.8 GFLOPS (1:32) |
Slot width | Single-slot | Dual-slot |
---|---|---|
Length | 145 mm 5.7 inches | 267 mm 10.5 inches |
Width | 69 mm 2.7 inches | — |
Height | 18 mm 0.7 inches | — |
TDP | 30 W | 260 W |
Suggested PSU | 200 W | 600 W |
Outputs | 1x DVI1x HDMI | No outputs |
Power connectors | None | 1x 6-pin + 1x 8-pin |
Board number | PG110 SKU 0 | PG150 SKU 220 |
DirectX | 12 (12_1) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 3.0 | 3.0 |
Vulkan | 1.2 | 1.2 |
CUDA | 6.1 | 7.5 |
Shader model | 6.4 | 6.6 |