NVIDIA / Hopper
H100 SXM
| Provider / plan | Price / GPU-hour | Rental type | Terms & configuration | Last checked | Provider pricing link |
|---|---|---|---|---|---|
| Vast.aiOn-demand offer 49864346 | $1.9896Lowest flexible rate | On-demandNo reservation | 1 GPU minimumGermany, DEView detailsObserved marketplace offer, not a provider-wide list rate. Total hourly quote for one GPU with the API default disk allocation; bandwidth and additional storage can cost extra. Up to 5 cheapest matching offers per catalog GPU; availability can change.
|
2026-10-03Auto-collected | View pricing ↗ |
| Vast.aiOn-demand offer 51448486 | $2.6696 | On-demandNo reservation | 1 GPU minimumNew Jersey, USView detailsObserved marketplace offer, not a provider-wide list rate. Total hourly quote for one GPU with the API default disk allocation; bandwidth and additional storage can cost extra. Up to 5 cheapest matching offers per catalog GPU; availability can change.
|
2026-10-03Auto-collected | View pricing ↗ |
| Vast.aiOn-demand offer 48704103 | $3.1407 | On-demandNo reservation | 1 GPU minimumCzechia, CZView detailsObserved marketplace offer, not a provider-wide list rate. Total hourly quote for one GPU with the API default disk allocation; bandwidth and additional storage can cost extra. Up to 5 cheapest matching offers per catalog GPU; availability can change.
|
2026-10-03Auto-collected | View pricing ↗ |
| RunpodSecure Cloud Pods | $3.49 | On-demandNo reservation | 1 GPU minimumSee providerView detailsSecure Cloud Pods published GPU-hour rate. CPU/RAM allocation varies; storage/networking extra. Not live capacity.
|
2026-10-03Auto-collected | View pricing ↗ |
| ModalGPU Tasks | $3.9492GPU only · CPU/RAM extra | ServerlessActive GPU time | 1 GPU minimumSee providerView detailsGPU time only; CPU, RAM, storage, region and non-preemptible surcharges are extra.
|
2026-10-03Auto-collected | View pricing ↗ |
Prices are dated listings, not live availability. Compare minimum GPU counts and commitments. Storage, networking and taxes may add charges; serverless rates cover GPU time only.
H100 SXM is an 80 GB Hopper accelerator for training and inference in HGX-style servers. It combines HBM3 memory with high-bandwidth NVLink. It is especially relevant when several GPUs must exchange tensors frequently.
CHOOSING THIS GPU
Is it right for your workload?
80 GB SXM version. Do not substitute H100 PCIe or H100 NVL peak specifications; memory, power limits and interconnect differ.
Best suited to
- Distributed training on connected GPU nodes
- FP8-enabled inference with a compatible engine
- Fine-tuning models that need more memory than workstation GPUs
Strengths
- 80 GB HBM3 and 3.35 TB/s bandwidth
- 900 GB/s NVLink hardware capability
- Hopper Transformer Engine
Things to consider
- 80 GB still requires careful sizing for large models
- SXM is a server module, not a standalone PCIe card
- Sparse peak figures require supported sparsity
HARDWARE DETAILS
Inside the GPU
- Manufacturer
- NVIDIA
- Architecture
- Hopper
- VRAM (GB)
- 80
- Memory type
- HBM3
- Form factor
- SXM5
- Memory bandwidth (GB/s)
- 3350
- Maximum board power (W)
- 700
- CUDA cores
- 16896
- Tensor cores
- 528 / 4th generation
- Ray tracing cores
- Not verified for this edition
- CUDA capability
- 9.0
- Memory ECC
- Not verified for this edition
- Host interface
- PCIe Gen5
- GPU interconnect
- 4th generation; 900 GB/s aggregate bidirectional
- Hardware partitioning (MIG)
- Up to 7 hardware instances; provider must enable it
- Cooling
- Server platform dependent
Hardware capabilities are not a guarantee of access in a cloud instance. Check the exact edition, GPU allocation and server topology in the rental offer.
MODEL MEMORY
Plan your model size
70B 8-bit weights alone are about 70 GB; a full runtime needs additional memory. Compare quantized single-GPU serving with supported multi-GPU execution. Full fine-tuning requires substantially more than inference weights.
| Example model class | Weight format | Raw weights ≈ | Minimum GPUs for weights only |
|---|---|---|---|
| Llama 3.1 8B ↗ | 4-bit | 4 GB | 1Runtime needs more memory |
| Llama 3.1 8B ↗ | FP16 | 16 GB | 1Runtime needs more memory |
| Qwen2.5 14B ↗ | 4-bit | 7 GB | 1Runtime needs more memory |
| Qwen2.5 32B ↗ | 4-bit | 16 GB | 1Runtime needs more memory |
| Llama 3.1 70B ↗ | 4-bit | 35 GB | 1Runtime needs more memory |
| Llama 3.1 70B ↗ | 8-bit | 70 GB | 1Runtime needs more memory |
| Llama 3.1 70B ↗ | FP16 | 140 GB | 2Runtime needs more memory |
Arithmetic lower bound: rounded parameter count × bits per weight ÷ 8, in decimal GB. Excludes quantization metadata, KV cache, activations, CUDA workspaces and training states. Actual model sizes differ from their rounded names. No context length, batch size or concurrency is guaranteed. GPU counts assume supported model sharding and can be higher in practice. Quantized checkpoints and compatible kernels are required for 4-bit/8-bit execution.
SOFTWARE & PERFORMANCE
Before you deploy
Software compatibility
Use an NVIDIA CUDA-enabled framework/container compatible with this GPU and the host driver. Check PyTorch build and inference-engine kernel requirements before deployment. vLLM documents NVIDIA compute capability 7.5+ as a baseline; support for each model and quantization method still needs checking.
Measured benchmarks
A reproducible, comparable application benchmark has not yet been recorded for this edition. Peak TFLOPS above are not measured tokens per second or image-generation speed.
Compare tests using the same model, precision, GPU count, engine version, input/output length and batch or concurrency. A provider’s server configuration can change the result.
COMPARE YOUR OPTIONS
Alternatives to consider
BUDGET YOUR RUN
GPU cost estimator
GPU compute only: hourly rate × billable hours × GPU count. CPU, RAM, storage, egress, taxes, billing increments and minimum terms may add charges. Serverless hours mean active GPU task time; spot capacity can be interrupted. Reserved commitments and quote-only offers are excluded from this simple hourly estimator.