Llama 2 7B on the NVIDIA RTX A5000
Best measured: 140 tokens/s generation and 4,071 tokens/s prompt processing (Q4_0, Vulkan, FA on).
Every run
| Quant | Backend | Generate tok/s | Prompt tok/s | Test | Build / by / notes | Source |
|---|---|---|---|---|---|---|
| Q4_0 | Vulkan FA on | 140 | 4,071 | tg128 | 4ae88d0 coopmat2 | llama.cpp Vulkan scoreboard (#10879) |
| Q4_0 | Vulkan FA off | 140 | 3,642 | tg128 | 4ae88d0 coopmat2 | llama.cpp Vulkan scoreboard (#10879) |
| Q4_0 | CUDA FA on | 136 | 4,552 | tg128 | e5155e6 @Hedede | llama.cpp CUDA scoreboard (#15013) |
| Q4_0 | CUDA FA off | 130 | 4,028 | tg128 | e5155e6 @Hedede | 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.
No card yet?
Try it on a rented GPU (RunPod)Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.
All results for the NVIDIA RTX A5000 · Llama 2 7B on other hardware