Intel Optane Persistent Memory 200

Headlines · 1

1. 3d
   Homelab Runs Kimi K3 on CPU With Second-Gen Optane

   A Reddit homelab builder reportedly ran Moonshot AI’s 2.8-trillion-parameter Kimi K3 on a 40-core third-generation Intel Xeon Scalable CPU using discontinued second-generation Optane PMem 200 modules, without a GPU for inference. Before thread-count or NUMA-affinity tuning, the system generated 0.85 tokens per second with 1-bit quantization and 0.75 tokens per second with 3-bit quantization. The builder said the Optane modules became extremely hot during sustained inference and advised treating active DIMM cooling as mandatory.

   Tom’s Hardware · 🔥 0

Experience and discussion from the community

Share my Intel Optane Persistent Memory 200 experienceAsk about Intel Optane Persistent Memory 200