GPUs & Macs › NVIDIA Tesla M40
NVIDIA Tesla M40: local LLM speed
Llama 2 7B (Q4_0) generates 41.7 tokens/s on the NVIDIA Tesla M40 (Vulkan), and processes prompts at 93 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 41.7 | 93 | Q4_0 | Vulkan FA on | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 40.0 | 290 | Q4_0 | CUDA FA on | tg128 | llama.cpp #15013 details |
| Llama 2 7B | 39.4 | 92 | Q4_0 | Vulkan FA off | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 38.0 | 283 | Q4_0 | CUDA FA off | tg128 | llama.cpp #15013 details |
What fits on the NVIDIA Tesla M40
Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 41.7 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | UD-Q6_K | 22.0 GB | 7.2 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 12.6 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 16.7 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 35.5 estimate |
| Qwen3.6-35B-A3B | UD-Q4_K_S | 20.9 GB | 44.5 estimate |
| gemma-4-26B-A4B-it | UD-Q5_K_M | 21.2 GB | 24.5 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 16.7 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 19.4 estimate |
| Muse-Glimmer-30B | UD-Q5_K_M | 19.2 GB | 8.3 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 37.2 estimate |
| gemma-4-31B-it | Q5_K_M | 21.7 GB | 7.4 estimate |
| gemma-4-26B-A4B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 14.2 GB | 36.3 estimate |
| gemma-4-12B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 6.7 GB | 23.7 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 133 estimate |
Similar hardware
NVIDIA A30 · NVIDIA Tesla P40 · NVIDIA L4 · NVIDIA A10 · NVIDIA Tesla V100 32GB · NVIDIA Tesla P100 · NVIDIA Tesla T4 · NVIDIA Tesla V100 16GB