Product Compare
AMD Radeon Instinct MI50 vs NVIDIA H100 CNX
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Nov 2018 | Mar 2023 |
| Generation | Radeon Instinct | Tesla Hopper |
| Type | Professional | AI GPU |
| Bus Interface | PCIe 4.0 x16 | PCIe 5.0 x16 |
| Clock Speeds | ||
| Base Clock | 1200 MHz | 690 MHz |
| Boost Clock | 1746 MHz | 1845 MHz |
| Memory Clock | 1000 MHz | 1593 MHz |
| Memory | ||
| Memory Size | 16GB | 80GB |
| Memory Type | HBM2 | HBM2e |
| Memory Bus | 4096bit | 5120bit |
| Bandwidth | 1.02TB/s | 2039GB/s |
| Render Config | ||
| Compute Units | 60 | |
| Shading Units | 3840 | 14592 |
| TMUs | 240 | 456 |
| ROPs | 64 | 24 |
| RT Cores | 32 | |
| L1 Cache | 16 KB (per CU) | 256 KB (per SM) |
| L2 Cache | 4 MB | 50 MB |
| L3 Cache | 32 MB | |
| SM Count | 114 | |
| Tensor Cores | 456 | |
| Theoretical Performance | ||
| Pixel Rate | 111.7 GPixel/s | 44.28 GPixel/s |
| Texture Rate | 419.0 GTexel/s | 841.3 GTexel/s |
| FP16 (half) | 26.82 TFLOPS | 215.4 TFLOPS |
| FP32 (float) | 13.41 TFLOPS | 53.84 TFLOPS |
| FP64 (double) | 6.705 TFLOPS | 26.92 TFLOPS |
| Graphics Processor | ||
| GPU Name | Vega 20 | GH100 |
| GPU Variant | Vega 20 GLXT | |
| Architecture | GCN 5.1 | Hopper |
| Foundry | TSMC | TSMC |
| Process Size | 7 nm | 4 nm |
| Transistors | 13.23 billion | 80 billion |
| Die Size | 331 mm² | 814 mm² |
| Board Design | ||
| TDP | 300W | 350W |
| Suggested PSU | 700 W | 750 W |
| Outputs | 1x mini-DisplayPort 1.4a | No outputs |
| Power Connectors | 2x 8-pin | 8-pin EPS |
| Graphics Features | ||
| DirectX | 12 (12_1) | N/A |
| OpenGL | 4.6 | N/A |
| OpenCL | 2.1 | 3.0 |
| Vulkan | 1.3 | N/A |
| Shader Model | 6.7 | N/A |
| CUDA | 9.0 | |