GPU model | Mars | Vega 10 |
---|---|---|
GPU variant | Mars PRO (216-0842000) | Vega 10 XL (215-0894216) |
Architecture | GCN 1.0 | GCN 5.0 |
Foundry | TSMC | GlobalFoundries |
Manufacturing process | 28 nm | 14 nm |
Number of transistors | 950 million | 12,500 million |
Die size | 77 mm² | 495 mm² |
Release date | Feb 26th, 2013 | Jun 1st, 2018 |
---|---|---|
Generation | Solar System (HD 8700M) | Mobility Radeon (Vega) |
Production | End-of-life | Active |
Interface | PCIe 3.0 x8 | PCIe 3.0 x16 |
Predecessor | London | Crystal System |
Successor | Crystal System | — |
Reviews | — | 23 in our database |
Base clock | 620 MHz | 1138 MHz |
---|---|---|
Boost clock | 775 MHz | 1301 MHz |
Memory clock | 1000 MHz 4 Gbps effective | 800 MHz 1600 Mbps effective |
Memory size | 1024 MB | 8 GB |
---|---|---|
Memory type | GDDR5 | HBM2 |
Memory bus width | 128 bit | 2048 bit |
Bandwidth | 64.00 GB/s | 409.6 GB/s |
Shading units | 384 | 3584 |
---|---|---|
TMUs | 24 | 224 |
ROPs | 8 | 64 |
Compute units | 6 | 56 |
Cache L1 | 16 KB (per CU) | 16 KB (per CU) |
Cache L2 | 256 KB | 4 MB |
Pixel rate | 6.200 GPixel/s | 83.26 GPixel/s |
---|---|---|
Texture rate | 18.60 GTexel/s | 291.4 GTexel/s |
FP32 (float) performance | 595.2 GFLOPS | 9.326 TFLOPS |
FP64 (double) performance | 37.20 GFLOPS (1:16) | 582.8 GFLOPS (1:16) |
FP16 (half) performance | — | 18.65 TFLOPS (2:1) |
TDP | unknown | 120 W |
---|---|---|
Outputs | No outputs | 1x HDMI3x DisplayPort |
Slot width | — | MXM Module |
Length | — | 105 mm 4.1 inches |
Power connectors | — | None |
DirectX | 12 (11_1) | 12 (12_1) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 1.2 | 2.1 |
Vulkan | 1.2 | 1.2 |
Shader model | 5.1 | 6.4 |