NVIDIA H100 80GB vs NVIDIA H100 NVL for local LLMs
6 reports on the NVIDIA H100 80GB and 0 on the NVIDIA H100 NVL
Which is faster for local LLMs?
Nobody has reported the same model at the same quant on both cards yet, so there's no measured head-to-head. Memory bandwidth, which caps how fast a card can write, is 16% higher on the NVIDIA H100 NVL (3,350 GB/s on the NVIDIA H100 80GB, 3,900 on the NVIDIA H100 NVL). The estimates below compare them on the models people run most.
NVIDIA H100 80GB
- Memory
- 80GB
- Memory bandwidth
- 3,350 GB/s
- FP16 compute
- 989 TFLOPS
- Reports
- 6
NVIDIA H100 NVL
- Memory
- 94GB
- Memory bandwidth
- 3,900 GB/s
- FP16 compute
- 835.5 TFLOPS
- Reports
- 0
Estimated by the calculator
The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape.
| Model | NVIDIA H100 80GB | NVIDIA H100 NVL |
|---|---|---|
| Qwen3.8 27B | 34.1 t/sFP16 | 39.7 t/sFP16 |
| Qwen3.8 125B · 6B active | 202.0 t/sQ4_K_S | 217.3 t/sQ4_K_M |
| DeepSeek V4 Flash 284B · 13B active | 9.6 t/sQ2_K, experts in RAM | 9.6 t/sQ2_K, experts in RAM |
| Qwen3.6 35B · 3B active | 245.7 t/sQ8_0 | 123.0 t/sFP16 |
| Qwen3.6 27B | 34.1 t/sFP16 | 39.7 t/sFP16 |
| Gemma 4 26B · 4B active | 87.8 t/sFP16 | 107.3 t/sFP16 |
Change the context, RAM or model in the calculator: NVIDIA H100 80GB · NVIDIA H100 NVL
Every report on each card, with its source: NVIDIA H100 80GB · NVIDIA H100 NVL