GPU model | Ellesmere | GA104 |
---|---|---|
GPU variant | Ellesmere PRO | GA104-775-A1 |
Architecture | GCN 4.0 | Ampere |
Foundry | GlobalFoundries | Samsung |
Manufacturing process | 14 nm | 8 nm |
Number of transistors | 5,700 million | 17,400 million |
Die size | 232 mm² | 392 mm² |
Release date | Aug 4th, 2016 | Jan 12th, 2021 |
---|---|---|
Generation | Mobility Radeon (RX M400) | GeForce 30 Mobile |
Production | Active | Active |
Interface | MXM-B (3.0) | PCIe 4.0 x16 |
Predecessor | Crystal System | GeForce 20 Mobile |
gpu.details.availability | — | Feb 2nd, 2021 |
Base clock | 926 MHz | 780 MHz |
---|---|---|
Boost clock | 1074 MHz | 1245 MHz |
Memory clock | 1750 MHz 7 Gbps effective | 1500 MHz 12 Gbps effective |
Memory size | 8 GB | 8 GB |
---|---|---|
Memory type | GDDR5 | GDDR6 |
Memory bus width | 256 bit | 256 bit |
Bandwidth | 224.0 GB/s | 384.0 GB/s |
Shading units | 2048 | 6144 |
---|---|---|
TMUs | 128 | 192 |
ROPs | 32 | 96 |
Compute units | 32 | — |
Cache L1 | 16 KB (per CU) | 128 KB (per SM) |
Cache L2 | 2 MB | 4 MB |
SM count | — | 48 |
Tensor cores | — | 192 |
RT cores | — | 48 |
Pixel rate | 34.37 GPixel/s | 119.5 GPixel/s |
---|---|---|
Texture rate | 137.5 GTexel/s | 239.0 GTexel/s |
FP16 (half) performance | 4.399 TFLOPS (1:1) | 15.30 TFLOPS (1:1) |
FP32 (float) performance | 4.399 TFLOPS | 15.30 TFLOPS |
FP64 (double) performance | 274.9 GFLOPS (1:16) | 239.0 GFLOPS (1:64) |
Slot width | MXM Module | — |
---|---|---|
TDP | 85 W | 80 W |
Outputs | No outputs | No outputs |
Power connectors | None | None |
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 | — | 8.6 |