Product Compare
NVIDIA RTX PRO 4000 Blackwell vs NVIDIA H100 PCIe 80 GB
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Mar 2025 | Oct 2022 |
| Generation | Blackwell PRO W | Server Hopper |
| Type | Desktop | AI GPU |
| Bus Interface | PCIe 5.0 x16 | PCIe 5.0 x16 |
| Clock Speeds | ||
| Base Clock | 1230 MHz | 1095 MHz |
| Boost Clock | 2055 MHz | 1755 MHz |
| Memory Clock | 1750 MHz | 1593 MHz |
| Memory | ||
| Memory Size | 24GB | 80GB |
| Memory Type | GDDR7 | HBM2e |
| Memory Bus | 192bit | 5120bit |
| Bandwidth | 672.0GB/s | 2.04TB/s |
| Render Config | ||
| SM Count | 70 | 114 |
| Shading Units | 8960 | 14592 |
| TMUs | 280 | 456 |
| ROPs | 96 | 24 |
| Tensor Cores | 280 | 456 |
| RT Cores | 70 | |
| L1 Cache | 128 KB (per SM) | 256 KB (per SM) |
| L2 Cache | 48 MB | 50 MB |
| Theoretical Performance | ||
| Pixel Rate | 197.3 GPixel/s | 42.12 GPixel/s |
| Texture Rate | 575.4 GTexel/s | 800.3 GTexel/s |
| FP16 (half) | 36.83 TFLOPS | 204.9 TFLOPS |
| FP32 (float) | 36.83 TFLOPS | 51.22 TFLOPS |
| FP64 (double) | 575.4 GFLOPS | 25.61 TFLOPS |
| Graphics Processor | ||
| GPU Name | GB203 | GH100 |
| Architecture | Blackwell 2.0 | Hopper |
| Foundry | TSMC | TSMC |
| Process Size | 5 nm | 5 nm |
| Transistors | 45.6 billion | 80 billion |
| Die Size | 378 mm² | 814 mm² |
| Board Design | ||
| TDP | 140W | 350W |
| Suggested PSU | 300 W | 750 W |
| Outputs | 4x DisplayPort 2.1b | No outputs |
| Power Connectors | 1x 16-pin | 1x 16-pin |
| Graphics Features | ||
| DirectX | 12 Ultimate (12_2) | N/A |
| OpenGL | 4.6 | N/A |
| OpenCL | 3.0 | 3.0 |
| Vulkan | 1.4 | N/A |
| CUDA | 12.0 | 9.0 |
| Shader Model | 6.8 | N/A |