GPU model | Ellesmere | GA102 |
---|---|---|
GPU variant | Ellesmere XTA (216-0886220) | GA102-300-A1 |
Architecture | GCN 4.0 | Ampere |
Foundry | GlobalFoundries | TSMC |
Manufacturing process | 14 nm | 7 nm |
Number of transistors | 5,700 million | 40,000 million |
Die size | 232 mm² | 627 mm² |
Release date | Jun 5th, 2017 | — |
---|---|---|
Generation | Radeon Pro Mac (500 Series) | — |
Production | Active | — |
Interface | PCIe 3.0 x16 | — |
Base clock | 1100 MHz | 1410 MHz |
---|---|---|
Boost clock | 1200 MHz | 1740 MHz |
Memory clock | 1695 MHz 6.8 Gbps effective | 2375 MHz 19000 MHz effective |
Memory size | 8 GB | 12 GB |
---|---|---|
Memory type | GDDR5 | GDDR6X |
Memory bus width | 256 bit | 384 bit |
Bandwidth | 217.0 GB/s | 912.0 GB/s |
Shading units | 2304 | 5120 |
---|---|---|
TMUs | 144 | 320 |
ROPs | 32 | 96 |
Compute units | 36 | — |
Cache L1 | 16 KB (per CU) | 192 KB (per SM) |
Cache L2 | 2 MB | 6 MB |
SM count | — | 80 |
Tensor cores | — | 640 |
RT cores | — | 160 |
Pixel rate | 38.40 GPixel/s | 167.0 GPixel/s |
---|---|---|
Texture rate | 172.8 GTexel/s | 556.8 GTexel/s |
FP16 (half) performance | 5.530 TFLOPS (1:1) | 35.64 TFLOPS (2:1) |
FP32 (float) performance | 5.530 TFLOPS | 17.82 TFLOPS |
FP64 (double) performance | 345.6 GFLOPS (1:16) | 556.8 GFLOPS (1:32) |
Slot width | IGP | Dual-slot |
---|---|---|
TDP | 185 W | 320 W |
Outputs | No outputs | 1x HDMI3x DisplayPort1x USB Type-C |
Power connectors | None | 1x 12-pin |
Length | — | 267 mm 10.5 inches |
Suggested PSU | — | 700 W |
Board number | — | PG132 SKU 20 |
DirectX | 12 (12_0) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.1 | 2.0 |
Vulkan | 1.2 | 1.2.140 |
Shader model | 6.4 | 6.5 |
CUDA | — | 8.5 |
Release date | — | 2020 |
---|---|---|
Generation | — | GeForce 30 |
Production | — | Unreleased |
Launch price | — | 1,399 USD |
Interface | — | PCIe 4.0 x16 |