GPUs & Macs › NVIDIA Tesla P40

NVIDIA Tesla P40: local LLM speed

Llama 2 7B (Q4_0) generates 59.4 tokens/s on the NVIDIA Tesla P40 (Vulkan), and processes prompts at 488 tokens/s.

Measured results

ModelGenerate
tok/s		Prompt
tok/s		QuantBackendTestSource
Llama 2 7B59.4488Q4_0Vulkan FA offtg128llama.cpp #10879 details
Llama 2 7B59.2484Q4_0Vulkan FA ontg128llama.cpp #10879 details
Llama 2 7B54.71,007Q4_0CUDA FA offtg128llama.cpp #15013 details
Llama 2 7B53.71,080Q4_0CUDA FA ontg128llama.cpp #15013 details
Llama 3.2 1B Instruct1052,662Q4_K_Mllamafile LocalScore avgLocalScore details
Qwen2.5 14B Instruct15.4329Q4_K_Mllamafile LocalScore avgLocalScore details

What fits on the NVIDIA Tesla P40

Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 59.4 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.

ModelBest quant that fitsFileExpected tok/s
Qwen3.8-27BUD-Q6_K22.0 GB10.3 estimate
gemma-4-12B-itQ8_012.7 GB17.9 estimate
Qwen3.5-9BQ8_09.5 GB23.8 estimate
Qwen3.5-4BQ8_04.5 GB50.6 estimate
Qwen3.6-35B-A3BUD-Q4_K_S20.9 GB63.4 estimate
gemma-4-26B-A4B-itUD-Q5_K_M21.2 GB34.9 estimate
MiMo-V2.6-Distill-Qwen-9BQ8_09.5 GB23.8 estimate
gemma-4-E4B-itQ8_08.2 GB27.7 estimate
Muse-Glimmer-30BUD-Q5_K_M19.2 GB11.8 estimate
Qwen3-4BQ8_04.3 GB53.0 estimate
gemma-4-31B-itQ5_K_M21.7 GB10.5 estimate
gemma-4-26B-A4B-it-qat-q4_0-unquantizedUD-Q4_K_XL14.2 GB51.8 estimate
gemma-4-12B-it-qat-q4_0-unquantizedUD-Q4_K_XL6.7 GB33.8 estimate
gemma-3-1b-itQ8_01.1 GB189 estimate

Open in the picker

Similar hardware

NVIDIA A30 · NVIDIA Tesla M40 · NVIDIA L4 · NVIDIA A10 · NVIDIA Tesla V100 32GB · NVIDIA Tesla P100 · NVIDIA Tesla T4 · NVIDIA Tesla V100 16GB