odroid@odroid:~/llama.cpp$ taskset -c 0-7 ./build/bin/llama-bench -m qwen0.5b.gguf -p 512 -n 128 -t 5,6,7 | model | size | params | backend | threads | test | t/s | | ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: | | qwen2 1B Q4_K - Medium | 462.96 MiB | 630.17 M | CPU | 5 | pp512 | 16.65 ± 0.04 | | qwen2 1B Q4_K - Medium | 462.96 MiB | 630.17 M | CPU | 5 | tg128 | 2.67 ± 0.02 | | qwen2 1B Q4_K - Medium | 462.96 MiB | 630.17 M | CPU | 6 | pp512 | 17.47 ± 0.04 | | qwen2 1B Q4_K - Medium | 462.96 MiB | 630.17 M | CPU | 6 | tg128 | 2.82 ± 0.02 | | qwen2 1B Q4_K - Medium | 462.96 MiB | 630.17 M | CPU | 7 | pp512 | 18.54 ± 0.04 | | qwen2 1B Q4_K - Medium | 462.96 MiB | 630.17 M | CPU | 7 | tg128 | 3.46 ± 0.02 | build: 925e11799 (10642)