DigestOps
Search⌘KLog inCreate account
Archived signal · Frontier Models
Chesny@chesnyx

3x Nvidia RTX 5080 inference speed on DeepSeek R1 with AMD Ryzen AI Max+ 395

  • AMD Ryzen AI Max+ 395 chip shares 128 GB memory between CPU and GPU, giving GPU 110 GB for model inference.
  • The chip outperformed Nvidia RTX 5080 by over 3x in DeepSeek R1 inference in a live demo.
  • The system runs a 235-billion-parameter model locally on a $1,499 mini PC without cloud or data center.
how to reproduce · Purchase AMD Ryzen AI Max+ 395 based system (e.g., mini PC) with 128 GB unified memory. Install Linux to allocate 110 GB to GPU. Load DeepSeek R1 model and run inference.
♥ 1.5K239.3K views↗ Open the original
Share
Where this lives now

This signal is part of the DigestOps evidence archive. The living answers built on evidence like this are at Frontier Models; agents can query it as JSON at /api/v1/cards/117 or over MCP.

3x Nvidia RTX 5080 inference speed on DeepSeek R1 with AMD Ryzen AI Max+ 395 · DigestOps