using both iGPU and CPU together

[fanlessfan]

Hello,

I can successfully running Ollama in iGPU of Core i7 13700K, but the perfomance is half of the CPU. Is it possible to use CPU and iGPU together so that we can get 1.5x performance than CPU only?

thx

[sgwhat]

Hi @fanlessfan , which model are you running, and can you provide the script you use to run ollama and the ollama serve log?

[fanlessfan]

It's not model related. I just followed the instructions on below link. It works and I can see the iGPU utilization is almost 100% while the model is running using intel_GPU_top and CPU is only one core busy.

https://github.com/intel-analytics/ipex-llm/blob/main/docs/mddocs/Quickstart/ollama_quickstart.md