GPU model | Polaris 20 | TU106 |
---|---|---|
GPU variant | Polaris 20 XL (215-0910052) | TU106-125-A1 |
Architecture | GCN 4.0 | Turing |
Foundry | GlobalFoundries | TSMC |
Manufacturing process | 14 nm | 12 nm |
Number of transistors | 5,700 million | 10,800 million |
Die size | 232 mm² | 445 mm² |
Release date | Oct 15th, 2018 | Jun 18th, 2020 |
---|---|---|
Generation | Polaris | GeForce 16 |
Production | Active | Active |
Interface | PCIe 3.0 x16 | PCIe 3.0 x16 |
Reviews | 52 in our database | — |
Predecessor | Arctic Islands | GeForce 10 |
Successor | Vega | GeForce 20 |
Base clock | 1168 MHz | 1410 MHz |
---|---|---|
Boost clock | 1284 MHz | 1590 MHz |
Memory clock | 1750 MHz 7 Gbps effective | 1500 MHz 12 Gbps effective |
Memory size | 4 GB | 4 GB |
---|---|---|
Memory type | GDDR5 | GDDR6 |
Memory bus width | 256 bit | 128 bit |
Bandwidth | 224.0 GB/s | 192.0 GB/s |
Shading units | 2048 | 896 |
---|---|---|
TMUs | 128 | 56 |
ROPs | 32 | 32 |
Compute units | 32 | — |
Cache L1 | 16 KB (per CU) | 64 KB (per SM) |
Cache L2 | 2 MB | 1024 KB |
SM count | — | 14 |
Tensor cores | — | 112 |
RT cores | — | 14 |
Pixel rate | 41.09 GPixel/s | 50.88 GPixel/s |
---|---|---|
Texture rate | 164.4 GTexel/s | 89.04 GTexel/s |
FP16 (half) performance | 5.259 TFLOPS (1:1) | 5.699 TFLOPS (2:1) |
FP32 (float) performance | 5.259 TFLOPS | 2.849 TFLOPS |
FP64 (double) performance | 328.7 GFLOPS (1:16) | 89.04 GFLOPS (1:32) |
Slot width | Dual-slot | Dual-slot |
---|---|---|
Length | 241 mm 9.5 inches | 229 mm 9 inches |
TDP | 150 W | 90 W |
Suggested PSU | 450 W | 250 W |
Outputs | 1x DVI1x HDMI3x DisplayPort | 1x DVI1x HDMI1x DisplayPort |
Power connectors | 1x 8-pin | 1x 6-pin |
Board number | C940, D000 | — |
Width | — | 111 mm 4.4 inches |
Height | — | 35 mm 1.4 inches |
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 |