Product Compare
NVIDIA CMP 50HX vs NVIDIA B200
| GPU | GPU | |
|---|---|---|
| Graphics Card | ||
| Release Date | Jun 2021 | Jan 2024 |
| Generation | Mining GPUs | Server Blackwell |
| Type | Desktop | Desktop |
| Bus Interface | PCIe 1.0 x4 | PCIe 5.0 x16 |
| Clock Speeds | ||
| Base Clock | 1350 MHz | 700 MHz |
| Boost Clock | 1545 MHz | 1965 MHz |
| Memory Clock | 1750 MHz | 2000 MHz |
| Memory | ||
| Memory Size | 10GB | 90GB |
| Memory Type | GDDR6 | HBM3e |
| Memory Bus | 320bit | 4096bit |
| Bandwidth | 560.0GB/s | 4.10TB/s |
| Render Config | ||
| SM Count | 56 | 148 |
| Shading Units | 3584 | 18944 |
| TMUs | 192 | 592 |
| ROPs | 80 | 24 |
| Tensor Cores | 448 | 592 |
| RT Cores | 56 | |
| L1 Cache | 64 KB (per SM) | 256 KB (per SM) |
| L2 Cache | 5 MB | 50 MB |
| Theoretical Performance | ||
| Pixel Rate | 123.6 GPixel/s | 47.16 GPixel/s |
| Texture Rate | 296.6 GTexel/s | 1163.3 GTexel/s |
| FP16 (half) | 22.15 TFLOPS | 1191.2 TFLOPS |
| FP32 (float) | 11.07 TFLOPS | 74.45 TFLOPS |
| FP64 (double) | 346.1 GFLOPS | 37.22 TFLOPS |
| Graphics Processor | ||
| GPU Name | TU102 | GB100 |
| GPU Variant | TU102-100-A1 | |
| Architecture | Turing | Blackwell |
| Foundry | TSMC | TSMC |
| Process Size | 12 nm | 5 nm |
| Transistors | 18.6 billion | 104 billion |
| Die Size | 754 mm² | Unknown |
| Board Design | ||
| TDP | 250W | 1000W |
| Suggested PSU | 600 W | 1400 W |
| Outputs | No outputs | No outputs |
| Power Connectors | 2x 8-pin | |
| 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 | 7.5 | 10.0 |
| Shader Model | 6.8 | N/A |