Скачать или смотреть DeepSeek R1 on Intel Arc B580: More Performance with llama.cpp and Overclocking

Disclaimer:
This video is intended for general informational and educational purposes only. It does not constitute professional advice, nor does it create any professional relationship. For personalized guidance, consult a qualified professional. The creator of this video disclaims all liability for any actions taken based on its content.

In this video, I’ll demonstrate how I transitioned from running DeepSeek R1 on Intel’s Arc B580 with Ollama to using llama.cpp—along with a bit of overclocking—for even better performance, boosting tokens per second from roughly 28.9 to 32.5. I’ll walk through the llama.cpp setup process, run the tests, and explain why I believe the easy accessibility and cost-effectiveness of running DeepSeek R1 locally on a Intel Arc B580 is truly transformative. Finally, I’ll share a brief message for Intel. I hope you enjoy the content!

Timecodes:
0:00:00 Intro
0:00:54 Before We Begin
0:01:09 Test System
0:01:40 Setting Up llama.cpp
0:02:44 Preparing the Model
0:03:13 Running the Test Prompt
0:04:50 Overclocking
0:06:15 Result Highlights
0:07:09 DeepSeek R1: A Game Changer?
0:09:46 A Message to Intel
0:10:25 Thank You

Links:

Running DeepSeek R1 on Intel Arc B580: My Setup and Performance Experience
   • Running DeepSeek R1 on Intel Arc B580: My ...  

https://github.com/intel/ipex-llm/blo...

https://github.com/deepseek-ai/DeepSe...