Product Compare
NVIDIA A40 PCIe vs NVIDIA H100 CNX
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Oct 2020 | Mar 2023 |
| Generation | Tesla | Tesla Hopper |
| Type | Desktop | AI GPU |
| Bus Interface | PCIe 4.0 x16 | PCIe 5.0 x16 |
| Clock Speeds | ||
| Base Clock | 1305 MHz | 690 MHz |
| Boost Clock | 1740 MHz | 1845 MHz |
| Memory Clock | 1812 MHz | 1593 MHz |
| Memory | ||
| Memory Size | 48GB | 80GB |
| Memory Type | GDDR6 | HBM2e |
| Memory Bus | 384bit | 5120bit |
| Bandwidth | 695.8GB/s | 2039GB/s |
| Render Config | ||
| SM Count | 84 | 114 |
| Shading Units | 10752 | 14592 |
| TMUs | 336 | 456 |
| ROPs | 112 | 24 |
| Tensor Cores | 336 | 456 |
| RT Cores | 84 | |
| L1 Cache | 128 KB (per SM) | 256 KB (per SM) |
| L2 Cache | 6 MB | 50 MB |
| Theoretical Performance | ||
| Pixel Rate | 194.9 GPixel/s | 44.28 GPixel/s |
| Texture Rate | 584.6 GTexel/s | 841.3 GTexel/s |
| FP16 (half) | 37.42 TFLOPS | 215.4 TFLOPS |
| FP32 (float) | 37.42 TFLOPS | 53.84 TFLOPS |
| FP64 (double) | 584.6 GFLOPS | 26.92 TFLOPS |
| Graphics Processor | ||
| GPU Name | GA102 | GH100 |
| Architecture | Ampere | Hopper |
| Foundry | Samsung | TSMC |
| Process Size | 8 nm | 4 nm |
| Transistors | 28.3 billion | 80 billion |
| Die Size | 628 mm² | 814 mm² |
| Board Design | ||
| TDP | 300W | 350W |
| Suggested PSU | 700 W | 750 W |
| Outputs | 3x DisplayPort 1.4a | No outputs |
| Power Connectors | 8-pin EPS | 8-pin EPS |
| Graphics Features | ||
| DirectX | 12 Ultimate (12_2) | N/A |
| OpenGL | 4.6 | N/A |
| OpenCL | 3.0 | 3.0 |
| Vulkan | 1.3 | N/A |
| CUDA | 8.6 | 9.0 |
| Shader Model | 6.6 | N/A |