Moving My VCF 9 Lab to Six Minisforum MS-A2 PCs
Explore a move to six Minisforum MS-A2 PCs for a VCF homelab, with separate management and workload hosts, memory choices and rack layout.
Historical hardware snapshot
This post records November 2025. For the later seven-system MS-A2 design and revised network, see the September 2026 homelab architecture. Use the VCF lab guide index to choose a deployment walkthrough.
Overview
I moved my VCF 9 homelab to six Minisforum MS-A2 PCs as a deliberate redesign of the lab. My previous hardware could support VCF 9 when paired with NVMe storage, so this was not a compatibility-driven replacement or a tradeoff between an unsupported platform and a supported one. I wanted to build around the compact MS-A2 platform while following William Lam’s published VCF 9 work with Minisforum systems. This update records the models, host roles, memory choices and rack layout I planned for the change.
Hardware choices and roles
In November 2025, I purchased the following systems for the lab redesign:
| (Qty 3) MS-A2 AMD Ryzen 9 7940HX with Radeon Graphics | Purpose: VCF 9 management domain |
| (Qty 3) MS-A2 AMD Ryzen 9 9955HX with Radeon Graphics | Purpose: VCF9 workload domain, Kubernetes and possible Private AI |
Memory and deployment plans
The design follows William Lam’s VCF on Minisforum MS-A2 guidance and his VCF 9 in a Box build for the most part. I am only putting 96GB of RAM in my workload hosts, due mostly to the increased cost of DDR5-5600 SODIMM memory. ESXi memory tiering also gives me additional flexibility for this lab allocation.
Rack layout and progress
At the time of this November 2025 update, the MS-A2 7940HX systems were installed in the rack and running VCF 9. The MS-A2 9955HX systems had not yet arrived.
Here is the new layout, been using my 3d printer a ton lately to print out mounts for this setup.
This post records the initial six-system transition. The later September 2026 architecture update documents four management hosts, three GPU-equipped workload hosts and the revised network.