Product Compare
NVIDIA P102 100 vs NVIDIA H100 CNX
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Feb 2018 | Mar 2023 |
| Generation | Mining GPUs | Tesla Hopper |
| Type | Desktop | AI GPU |
| Bus Interface | PCIe 3.0 x4 | PCIe 5.0 x16 |
| Clock Speeds | ||
| Base Clock | 1582 MHz | 690 MHz |
| Boost Clock | 1683 MHz | 1845 MHz |
| Memory Clock | 1376 MHz | 1593 MHz |
| Memory | ||
| Memory Size | 5GB | 80GB |
| Memory Type | GDDR5X | HBM2e |
| Memory Bus | 320bit | 5120bit |
| Bandwidth | 440.3GB/s | 2039GB/s |
| Render Config | ||
| SM Count | 25 | 114 |
| Shading Units | 3200 | 14592 |
| TMUs | 200 | 456 |
| ROPs | 80 | 24 |
| L1 Cache | 48 KB (per SM) | 256 KB (per SM) |
| L2 Cache | 0 MB | 50 MB |
| Tensor Cores | 456 | |
| Theoretical Performance | ||
| Pixel Rate | 134.6 GPixel/s | 44.28 GPixel/s |
| Texture Rate | 336.6 GTexel/s | 841.3 GTexel/s |
| FP16 (half) | 168.3 GFLOPS | 215.4 TFLOPS |
| FP32 (float) | 10.77 TFLOPS | 53.84 TFLOPS |
| FP64 (double) | 336.6 GFLOPS | 26.92 TFLOPS |
| Graphics Processor | ||
| GPU Name | GP102 | GH100 |
| GPU Variant | GP102-100-A1 | |
| Architecture | Pascal | Hopper |
| Foundry | TSMC | TSMC |
| Process Size | 16 nm | 4 nm |
| Transistors | 11.8 billion | 80 billion |
| Die Size | 471 mm² | 814 mm² |
| Board Design | ||
| TDP | 250W | 350W |
| Suggested PSU | 600 W | 750 W |
| Outputs | No outputs | No outputs |
| Power Connectors | 2x 8-pin | 8-pin EPS |
| Graphics Features | ||
| DirectX | 12 (12_1) | N/A |
| OpenGL | 4.6 | N/A |
| OpenCL | 3.0 | 3.0 |
| Vulkan | 1.3 | N/A |
| CUDA | 6.1 | 9.0 |
| Shader Model | 6.4 | N/A |