GPU model | Polaris 22 | GP107 |
---|---|---|
GPU variant | VegaM XL (D715A714) | — |
Architecture | GCN 4.0 | Pascal |
Foundry | GlobalFoundries | Samsung |
Manufacturing process | 14 nm | 14 nm |
Number of transistors | 5,000 million | 3,300 million |
Die size | 208 mm² | 132 mm² |
Release date | Feb 1st, 2018 | — |
---|---|---|
Generation | Vega (Vega M) | — |
Production | Active | — |
Interface | IGP | — |
Base clock | 931 MHz | 1266 MHz |
---|---|---|
Boost clock | 1011 MHz | 1354 MHz |
Memory clock | 700 MHz 1400 Mbps effective | 1252 MHz 5 Gbps effective |
Memory size | 4 GB | 2 GB |
---|---|---|
Memory type | HBM2 | GDDR5 |
Memory bus width | 1024 bit | 128 bit |
Bandwidth | 179.2 GB/s | 80.13 GB/s |
Shading units | 1280 | 512 |
---|---|---|
TMUs | 80 | 32 |
ROPs | 32 | 16 |
Compute units | 20 | — |
Cache L1 | 16 KB (per CU) | 48 KB (per SM) |
Cache L2 | 1024 KB | 1024 KB |
SM count | — | 4 |
Pixel rate | 32.35 GPixel/s | 21.66 GPixel/s |
---|---|---|
Texture rate | 80.88 GTexel/s | 43.33 GTexel/s |
FP16 (half) performance | 2.588 TFLOPS (1:1) | 21.66 GFLOPS (1:64) |
FP32 (float) performance | 2.588 TFLOPS | 1,386 GFLOPS |
FP64 (double) performance | 161.8 GFLOPS (1:16) | 43.33 GFLOPS (1:32) |
Slot width | IGP | Single-slot |
---|---|---|
TDP | 65 W | 40 W |
Outputs | No outputs | 4x mini-DisplayPort |
Board number | D136 | — |
Length | — | 145 mm 5.7 inches |
Width | — | 69 mm 2.7 inches |
Suggested PSU | — | 200 W |
Power connectors | — | None |
DirectX | 12 (12_0) | 12 (12_1) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.1 | 3.0 |
Vulkan | 1.2 | 1.2 |
Shader model | 6.4 | 6.4 |
CUDA | — | 6.1 |
Release date | — | Feb 1st, 2018 |
---|---|---|
Generation | — | Quadro |
Production | — | End-of-life |
Interface | — | PCIe 3.0 x16 |