Llama 2 7B › AMD Instinct MI50
Llama 2 7B on the AMD Instinct MI50
Best measured: 118 tokens/s generation and 1,127 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 | 118 | 1,127 | tg128 | 3af34b9 | llama.cpp Vulkan scoreboard (#10879) |
| Q4_0 | Vulkan FA off | 109 | 1,120 | tg128 | 3af34b9 | llama.cpp Vulkan scoreboard (#10879) |
| Q4_0 | ROCm FA on | 106 | 1,129 | tg128 | 97d5117 @wtarreau | llama.cpp ROCm scoreboard (#15021) |
| Q4_0 | ROCm FA off | 99.0 | 1,057 | tg128 | 97d5117 @wtarreau | llama.cpp ROCm scoreboard (#15021) |
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 MI50 · Llama 2 7B on other hardware