Explainer · 08

What's inside the machine?

A private AI machine looks like a normal computer. What makes it good at AI is how four parts are chosen and balanced — and the most important one isn't the processor.

The four parts that matter

  1. Graphics card (GPU)

    This does the thinking. Its memory decides how large a model fits, and larger models are generally smarter. It's the biggest factor in what the machine can do, and usually the biggest part of the price.

  2. System memory

    Holds your documents and working data and gives the machine room for bigger jobs. Plenty of it keeps everything responsive.

  3. Storage

    Fast solid-state drives for the models — which are large files — plus every conversation and document. All of it stays on the machine.

  4. Cooling

    AI work produces heat. Good cooling keeps the machine quiet enough to live in an office rather than a server room.

What matters less than you'd expect

  • The processor. It matters, but for running AI it's rarely the bottleneck. Spending extra here usually buys very little.
  • Gamer branding. Lighting and premium brand names add cost without adding capability.
  • The most expensive GPU. If your work fits comfortably on a smaller card, the top card is money you won't get back in results.

How the tiers differ

Our Desktop, Workstation and Server builds differ mostly in graphics memory and how many people they're sized to serve — from one person's daily thinking up to a household or small team. Exact parts are specified after the consult, around the work you actually do.