H100 NVL vs H100 SXM
Side-by-side comparison of the NVIDIA H100 NVL and the NVIDIA H100 SXM for AI inference workloads.
Specifications
| Spec | H100 NVL | H100 SXM |
|---|---|---|
| Generation | hopper | hopper |
| Memory Type | HBM3 | HBM3 |
| VRAM | 94 GB | 80 GB |
| Memory Bandwidth | 3,938 GB/s | 3,350 GB/s |
| BF16 TFLOPS | 835 | 990 |
| FP16 TFLOPS | 835 | 990 |
| FP8 TFLOPS | 1,671 | 1,979 |
| INT8 TOPS | 1,671 | 1,979 |
| TDP | 400 W | 700 W |
| Interconnect | nvlink | nvlink |
| NVLink Bandwidth | 600 GB/s | 900 GB/s |
| Max GPUs per Node | 8 | 8 |
| PCIe Gen | Gen 5 | Gen 5 |
| CUDA Compute Capability | 9 | 9 |
Pricing
H100 NVL
| Provider | On-Demand | Reserved | Spot |
|---|---|---|---|
| coreweave | $4.10/hr | $3.09/hr | - |
| aws | $5.60/hr | $4.20/hr | - |
H100 SXM
| Provider | On-Demand | Reserved | Spot |
|---|---|---|---|
| runpod | $4.18/hr | - | $3.29/hr |
| lambda | $2.49/hr | $1.89/hr | - |
| coreweave | $3.79/hr | $2.57/hr | - |
| aws | $5.12/hr | $3.59/hr | - |
| gcp | $4.85/hr | $3.40/hr | - |
| azure | $4.98/hr | $3.49/hr | - |
| vast ai | $3.40/hr | - | $2.50/hr |
| tensordock | $3.29/hr | - | $2.49/hr |
| fluidstack | $2.85/hr | - | $2.10/hr |
Cheapest available rate: H100 NVL at $3.09/hr vs H100 SXM at $1.89/hr — H100 SXM is +63% cheaper
Efficiency Metrics
TFLOPS / Watt
2.1
H100 NVL
1.4
H100 SXM
BF16
VRAM / Dollar
30.4
H100 NVL
42.3
H100 SXM
GB/$/hr
Bandwidth / Watt
9.8
H100 NVL
4.8
H100 SXM
GB/s/W
Models (FP16, 1 GPU)
185.0
H100 NVL
182.0
H100 SXM
Model Compatibility (FP16, Single GPU)
Only on H100 NVL (3)
- Falcon 40B
- VILA 1.5 40B
- Phi 3.5 MoE
Both (182)
- Yi 1.5 34B
- Yi 1.5 9B
- Yi Coder 9B
- GTE Qwen2 7B
- Marco O1
- Qwen 1.5 MoE A2.7B
- Qwen 2 Audio 7B
- Qwen 2.5 14B
- Qwen 2.5 32B
- Qwen 2.5 3B
- Qwen 2.5 Coder 32B
- OLMo 2 13B
- OLMo 2 7B
- Amazon Nova Lite
- OpenELM 3B
- BGE Large EN v1.5
- BGE M3
- Baichuan 2 13B
- OctoCoder 15B
- StarCoder2 15B
- +162 more
Only on H100 SXM (0)
None
Summary
The H100 NVL (hopper generation) offers 94GB of HBM3 with 835 BF16 TFLOPS and 3,938 GB/s memory bandwidth at 400W TDP.
The H100 SXM (hopper generation) offers 80GB of HBM3 with 990 BF16 TFLOPS and 3,350 GB/s memory bandwidth at 700W TDP.
The H100 NVL has +18% more VRAM, allowing it to run larger models without multi-GPU setups.
From a cost perspective, the H100 SXM is more affordable at $1.89/hr vs $3.09/hr for the H100 NVL.