Llama 2 7B on the NVIDIA Tesla P40
Best measured: 59.4 tokens/s generation and 488 tokens/s prompt processing (Q4_0, Vulkan, FA off).
Every run
| Quant | Backend | Generate tok/s | Prompt tok/s | Test | Build / by / notes | Source |
|---|---|---|---|---|---|---|
| Q4_0 | Vulkan FA off | 59.4 | 488 | tg128 | N/A | llama.cpp Vulkan scoreboard (#10879) |
| Q4_0 | Vulkan FA on | 59.2 | 484 | tg128 | N/A | llama.cpp Vulkan scoreboard (#10879) |
| Q4_0 | CUDA FA off | 54.7 | 1,007 | tg128 | c76b420 @m18coppola | llama.cpp CUDA scoreboard (#15013) |
| Q4_0 | CUDA FA on | 53.7 | 1,080 | tg128 | c76b420 @m18coppola | llama.cpp CUDA scoreboard (#15013) |
tg128 = generating 128 tokens; pp512 = reading a 512-token prompt (llama-bench). LocalScore averages several prompt and output lengths, so its numbers run lower than tg128.
Rent one by the hour
Rent a GPU on RunPodAffiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.
All results for the NVIDIA Tesla P40 · Llama 2 7B on other hardware