post

AI Without GPUs: Using Your Existing CPU Resources to Run AI Workloads

This is a talk Keith Bradley and I gave at VMware Explore 2024 Las Vegas called AI Without GPUs: Using Your Existing CPU Resources to Run AI Workloads. Keith Bradley is the Vice President of IT and Security at Nature Fresh Farms. Nature Fresh Farms uses AI to control every aspect of their agricultural operations and they’re using CPUs to process those AI workloads.

Graphics processing units (GPUs) are expensive, hard to acquire and extremely powerful, but there are many AI/ML applications that can run just fine without GPUs. This session covers how to use your existing central processing unit (CPU) resources to run AI workloads, what you can do, what you shouldn’t do and what types of problems you can solve without using any GPUs at all.

Covered in this talk:

  • Nature Fresh Farms use case
  • The AI/ML software stack
  • Introduction to Intel Xeon 4th Gen CPUs w/ AMX
  • Requirements for AMX on vSphere 8
  • Getting started with OneAPI and OpenVINO
  • Demo OpenVino for Computer Vision
  • Demo of LLM optimized for AMX
  • Demo tuned LLM on AMX
  • Takeaways

I hope you enjoy the talk.