GPU model | Ellesmere | GA103 |
---|---|---|
GPU variant | Polaris 10 XT GL | GA103 |
Architecture | GCN 4.0 | Ampere |
Foundry | GlobalFoundries | Samsung |
Manufacturing process | 14 nm | 8 nm |
Number of transistors | 5,700 million | unknown |
Die size | 232 mm² | unknown |
Release date | Nov 10th, 2016 | — |
---|---|---|
Generation | Radeon Pro | — |
Production | Active | — |
Launch price | 799 USD | — |
Interface | PCIe 3.0 x16 | — |
Base clock | 1188 MHz | 585 MHz |
---|---|---|
Boost clock | 1243 MHz | 1125 MHz |
Memory clock | 1750 MHz 7 Gbps effective | 1500 MHz 12 Gbps effective |
Memory size | 8 GB | 16 GB |
---|---|---|
Memory type | GDDR5 | GDDR6 |
Memory bus width | 256 bit | 256 bit |
Bandwidth | 224.0 GB/s | 384.0 GB/s |
Shading units | 2304 | 7680 |
---|---|---|
TMUs | 144 | 240 |
ROPs | 32 | 96 |
Compute units | 36 | — |
Cache L1 | 16 KB (per CU) | 128 KB (per SM) |
Cache L2 | 2 MB | 4 MB |
SM count | — | 60 |
Tensor cores | — | 240 |
RT cores | — | 60 |
Pixel rate | 39.78 GPixel/s | 108.0 GPixel/s |
---|---|---|
Texture rate | 179.0 GTexel/s | 270.0 GTexel/s |
FP16 (half) performance | 5.728 TFLOPS (1:1) | 17.28 TFLOPS (1:1) |
FP32 (float) performance | 5.728 TFLOPS | 17.28 TFLOPS |
FP64 (double) performance | 358.0 GFLOPS (1:16) | 270.0 GFLOPS (1:64) |
Slot width | Single-slot | — |
---|---|---|
Length | 241 mm 9.5 inches | — |
Width | 112 mm 4.4 inches | — |
TDP | 130 W | 95 W |
Suggested PSU | 300 W | — |
Outputs | 4x DisplayPort | No outputs |
Power connectors | 1x 6-pin | None |
Board number | C954 | — |
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 | 6.5 |
CUDA | — | 8.6 |
Release date | — | 2022 |
---|---|---|
gpu.details.availability | — | — |
Generation | — | GeForce 30 Mobile |
Predecessor | — | GeForce 20 Mobile |
Production | — | Unreleased |
Interface | — | PCIe 4.0 x16 |