Product Compare
NVIDIA H100 PCIe 80 GB vs AMD Radeon Instinct MI300A
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Oct 2022 | Dec 2023 |
| Generation | Server Hopper | Radeon Instinct |
| Type | AI GPU | Desktop |
| Bus Interface | PCIe 5.0 x16 | PCIe 5.0 x16 |
| Clock Speeds | ||
| Base Clock | 1095 MHz | 1000 MHz |
| Boost Clock | 1755 MHz | 2100 MHz |
| Memory Clock | 1593 MHz | 2525 MHz |
| Memory | ||
| Memory Size | 80GB | 192GB |
| Memory Type | HBM2e | HBM3 |
| Memory Bus | 5120bit | 8192bit |
| Bandwidth | 2.04TB/s | 10.3TB/s |
| Render Config | ||
| SM Count | 114 | |
| Shading Units | 14592 | 19456 |
| TMUs | 456 | 1216 |
| ROPs | 24 | 0 |
| Tensor Cores | 456 | 1216 |
| L1 Cache | 256 KB (per SM) | 16 KB (per CU) |
| L2 Cache | 50 MB | 16 MB |
| Compute Units | 304 | |
| L3 Cache | 256 MB | |
| Theoretical Performance | ||
| Pixel Rate | 42.12 GPixel/s | 0 MPixel/s |
| Texture Rate | 800.3 GTexel/s | 2554 GTexel/s |
| FP16 (half) | 204.9 TFLOPS | 653.7 TFLOPS |
| FP32 (float) | 51.22 TFLOPS | 81.72 TFLOPS |
| FP64 (double) | 25.61 TFLOPS | 81.72 TFLOPS |
| Graphics Processor | ||
| GPU Name | GH100 | Aqua Vanjaram |
| Architecture | Hopper | CDNA 3.0 |
| Foundry | TSMC | TSMC |
| Process Size | 5 nm | 5 nm |
| Transistors | 80 billion | 153 billion |
| Die Size | 814 mm² | 1017 mm² |
| Board Design | ||
| TDP | 350W | 750W |
| Suggested PSU | 750 W | 1150 W |
| Outputs | No outputs | No outputs |
| Power Connectors | 1x 16-pin | None |
| Graphics Features | ||
| DirectX | N/A | N/A |
| OpenGL | N/A | N/A |
| OpenCL | 3.0 | 3.0 |
| Vulkan | N/A | N/A |
| CUDA | 9.0 | |
| Shader Model | N/A | N/A |