OPEN_SOURCE ↗
REDDIT · REDDIT// 21d agoINFRASTRUCTURE
GMKtec EVO-X2 128GB suits local RAG
A Reddit user asks whether the GMKtec EVO-X2 with 128GB of RAM is a sensible home box for local RAG, research, CAD, and CFD. The appeal is AMD’s Ryzen AI Max+ 395 platform, which gives this mini PC unusually large unified memory for AI experiments, but the compact form factor still means you should think carefully about sustained loads and upgrade flexibility.
// ANALYSIS
Hot take: if your main goal is local RAG and model tinkering, the EVO-X2 is one of the few mini PCs that actually looks workstation-grade on paper; if CAD/CFD is the real priority, a bigger tower will still age better on cooling, noise, and flexibility.
- –AMD says Ryzen AI Max+ 395 brings 16 Zen 5 cores, a 50+ TOPS NPU, 40 RDNA 3.5 CUs, up to 128GB unified memory, and up to 96GB configurable VRAM, which is exactly the kind of memory pool local LLM work likes.
- –GMKtec’s own page lists 64GB, 96GB, and 128GB configurations, so the 128GB model is clearly the one aimed at heavier AI and creator workloads.
- –PCWorld notes that the 128GB config can load local models above 100B parameters more easily, which is the strongest argument for buying this over a more conventional mini PC.
- –Tom’s Hardware says the 128GB setup can expose 96GB to the GPU, calls it a standout for local AI development, and also notes that CPU/GPU/RAM are soldered, so you are locking in your memory choice.
- –The main caution is practical: this is premium hardware in a small chassis, so long CFD runs or other sustained all-core workloads may be better served by a larger workstation with more thermal headroom.
// TAGS
gmktec-evo-x2ragllminferencegpuself-hosted
DISCOVERED
21d ago
2026-03-21
PUBLISHED
21d ago
2026-03-21
RELEVANCE
7/ 10
AUTHOR
Icy_Annual_9954