Memory rules, not the processor
This is the most repeated purchasing mistake: pick the fastest processor and leave memory at the minimum. But a point cloud loads entirely into RAM, and when it does not fit the system starts using the disk as memory. At that point no processor helps: the machine is not slow, it crawls.
The practical priority order for this work is memory first, a GPU with enough VRAM second, NVMe storage third and the processor last. Exactly the reverse of how a sales counter presents it.
What actually matters
- Memory. Sized to your typical cloud, with headroom to mesh and compare without closing everything else.
- A GPU with generous VRAM. What limits you when viewing large meshes is nearly always VRAM, not raw card power. A modest GPU with enough memory outperforms a powerful one running tight.
- NVMe storage. Scanning projects are heavy and get opened and saved constantly. A slow disk shows up in every operation of the day.
- Processor. It matters, but less than it costs. Many meshing operations use fewer cores than you would think.
- Display and colour calibration. Secondary for scanning, relevant if you also review deviation maps, where the colour is the data.
Laptop or desktop
If you will scan in the field or at customer sites you need a laptop, and the trade-off has to be accepted: for the same budget it delivers less and runs hotter, and performance drops exactly when it has been working a while. If scanning always happens in the same building, a tower gives considerably more for the same money. Many plants end up with a laptop to capture and a tower to process, which is usually the sensible combination.
What we do not recommend
A gaming machine. They are optimised for something else — high frame rates on light scenes — and typically pair modest system memory with an expensive GPU. For point clouds that is exactly the wrong split.
How it gets sized
Not from a catalogue. Tell us the size of the parts you will scan, at what resolution and with what software; that gives the expected cloud size and from there the configuration. If you already own a machine, upgrading memory and storage is often enough instead of buying a new one — and we will tell you that too.