Llama 2 7B › AMD Instinct MI60
Llama 2 7B on the AMD Instinct MI60
Best measured: 91.5 tokens/s generation and 1,289 tokens/s prompt processing (Q4_0, ROCm, FA off).
Every run
| Quant | Backend | Generate tok/s | Prompt tok/s | Test | Build / by / notes | Source |
|---|---|---|---|---|---|---|
| Q4_0 | ROCm FA off | 91.5 | 1,289 | tg128 | 504af20 @Said-Akbar | llama.cpp ROCm scoreboard (#15021) |
| Q4_0 | Vulkan FA off | 78.2 | 369 | tg128 | 504af20 | llama.cpp Vulkan scoreboard (#10879) |
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 AMD Instinct MI60 · Llama 2 7B on other hardware