Prices are shown in the selected currency using configured rounding rules. Rental quotes are confirmed against the approved price version.
Display currencies follow configured commercial price rules, not live exchange rates. The accepted quote or invoice states the payable currency.
Compare GPUs
Compare two to four GPU options listed in the FusionPointAI catalogue using consistent price, availability and specification data.
| GPU |
NVIDIA H100 PCIe 80GB |
|---|---|
| Monthly rental | USD 1,579/month |
| Hourly equivalent | ~ USD 2.16/hour |
| VRAM | 80 GB HBM2e |
| Architecture | Hopper |
| Bandwidth | 2,000 GB/s |
| Form factor | PCIe dual-slot accelerator card |
| Interface | PCIe Gen5 x16 |
| TDP | 350 W |
| FP64 | 26 TFLOPS dense |
| FP64 Tensor | 51 TFLOPS dense |
| FP32 | 51 TFLOPS dense |
| TF32 | 378 TFLOPS dense / 756 TFLOPS sparse |
| BF16 | 756.5 TFLOPS dense / 1513 TFLOPS sparse |
| FP16 | 756.5 TFLOPS dense / 1513 TFLOPS sparse |
| FP8 | 1513 TFLOPS dense / 3026 TFLOPS sparse |
| INT8 | 1513 TOPS dense / 3026 TOPS sparse |
| MIG | Up to 7 MIGs @ 10 GB |
| NVLink | 600 GB/s NVLink Bridge |
| Availability | Limited availability |
| Frameworks | PyTorch, TensorFlow, JAX, Hugging Face, vLLM, TensorRT, ONNX Runtime, CUDA |
| Model compatibility guidance | Recommended for quantised inference; review context and concurrency |
| Action | Request rental quote |