AI Workstations

Every other build here starts from a blank sheet. This one starts from an inventory — find the one component actually stopping you, and check whether it is worth relieving.

Same model, same prompt, same seed, same GPU — and still a different answer. Why that happens, when it matters, and the machine that stops it.

This site tells you to ignore gaming rankings. If you also game, you cannot — the two workloads want almost disjoint parts of the same spec sheet, and every choice has to survive both.

You cannot future-proof compute, only the chassis around it. A machine bought over a year or two — what to over-specify at stage one, and why the graphics card comes last.

Most organisations that say air-gapped need no data egress instead. For the ones that genuinely need isolation: provisioning, verification, patch cadence and audit — where the hardware is the easy part.

Training inverts almost everything this site says about buying hardware — and most people should rent instead. The memory arithmetic, the inversions, and the machine, specified.

Two used 24 GB cards is the cheapest route to 48 GB of CUDA memory — and the point where a home AI machine becomes an engineering problem. Bifurcation, NVLink, circuit capacity and thermal stacking, specified.

A component specification for a machine that runs 32B-class models well — with the reasoning behind every part, and an explicit checklist of what you must verify before ordering.