Product Compare
NVIDIA H100 PCIe 96 GB vs NVIDIA RTX 4000 SFF Ada Generation
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Mar 2023 | Nov 2018 |
| Generation | Tesla Hopper | Radeon Instinct |
| Type | AI GPU | Professional |
| Bus Interface | PCIe 5.0 x16 | PCIe 4.0 x16 |
| Clock Speeds | ||
| Base Clock | 1665 MHz | 1200 MHz |
| Boost Clock | 1837 MHz | 1746 MHz |
| Memory Clock | 1313 MHz | 1000 MHz |
| Memory | ||
| Memory Size | 96GB | 16GB |
| Memory Type | HBM3 | HBM2 |
| Memory Bus | 5120bit | 4096bit |
| Bandwidth | 1681GB/s | 1.02TB/s |
| Render Config | ||
| SM Count | 132 | 48 |
| Shading Units | 16896 | 3840 |
| TMUs | 528 | 240 |
| ROPs | 24 | 64 |
| Tensor Cores | 528 | 192 |
| L1 Cache | 256 KB (per SM) | 16 KB (per CU) |
| L2 Cache | 50 MB | 4 MB |
| Compute Units | 60 | |
| RT Cores | 48 | |
| Theoretical Performance | ||
| Pixel Rate | 44.09 GPixel/s | 111.7 GPixel/s |
| Texture Rate | 969.9 GTexel/s | 419.0 GTexel/s |
| FP16 (half) | 248.3 TFLOPS | 26.82 TFLOPS |
| FP32 (float) | 62.08 TFLOPS | 13.41 TFLOPS |
| FP64 (double) | 31.04 TFLOPS | 6.705 TFLOPS |
| Graphics Processor | ||
| GPU Name | GH100 | Vega 20 |
| Architecture | Hopper | GCN 5.1 |
| Foundry | TSMC | TSMC |
| Process Size | 4 nm | 7 nm |
| Transistors | 80 billion | 13.23 billion |
| Die Size | 814 mm² | 331 mm² |
| GPU Variant | Vega 20 GLXT | |
| Board Design | ||
| TDP | 700W | 300W |
| Suggested PSU | 1100 W | 700 W |
| Outputs | No outputs | 1x mini-DisplayPort 1.4a |
| Power Connectors | 8-pin EPS | 2x 8-pin |
| Graphics Features | ||
| DirectX | N/A | 12 (12_1) |
| OpenGL | N/A | 4.6 |
| OpenCL | 3.0 | 2.1 |
| Vulkan | N/A | 1.3 |
| CUDA | 9.0 | 8.9 |
| Shader Model | N/A | 6.7 |