# Builder runs Kimi K3 on CPU with discontinued Optane memory

_Published Sunday, November 8, 2026 at 10:27 AM EST · AI, Infrastructure · Latest · Tier 2 — Notable_

![Open white PC case with a large CPU cooler and two black blower fans aimed at a row of memory modules — Primary](https://cdn.mos.cms.futurecdn.net/Wpdi2avEo8ZBZMremLVgaU-2496-80.jpg)

A home server builder ran Moonshot AI's 2.8-trillion-parameter Kimi K3 model without a GPU, using discontinued Intel Optane PMem 200 memory and a 40-core Xeon processor, Tom's Hardware reported. The builder measured 0.85 tokens per second with 1-bit quantization, which reduces the precision used to store model parameters.

The system exposes Optane as persistent memory that software accesses directly. Even the smallest community quantization of Kimi K3 requires at least 610GB of RAM. The builder used dual 12W blower fans to cool the modules and warned that sustained inference could endanger neighboring DRAM. The performance measurements preceded thread-count and NUMA-affinity tuning.

## Sources

- [Tom's Hardware](https://www.tomshardware.com/tech-industry/artificial-intelligence/homelabber-uses-second-gen-optane-memory-sticks-to-run-a-2-8-trillion-parameter-ai-model-on-cpu-alone-warns-that-optanes-extreme-heat-can-burn-out-nearby-memory-dram-used-blower-fans-to-prevent-melting-ram)

---
Canonical: https://techandbusiness.org/newswire/DQItEFae41HQ7a6K4EYzpB
Published: 2026-11-08T15:27:22.516Z
Story chronology: 2026-11-08T12:20:00.000Z
Retrieved: 2026-11-08T18:31:36.023Z
Publisher: Tech & Business (techandbusiness.org)