CASE STUDIES
Measurements from real installs. The failures are included because a configuration that only ever reports success is not documentation. See the known issues for the general list.
8C/16T · 16 GiB · NVMe
The machine the figures on this site are measured on. Runs the full configuration with no exceptions: no swap, no compositor, software rendering throughout.
Both OOM events were a browser with too many tabs open, and both were recoverable because the failure was immediate and obvious. This is the intended behaviour, not a defect.
2C/4T · 8 GiB · SATA SSD
A 2014 laptop that was unusable under a conventional desktop environment. The interesting result is not that it got faster, but which parts of the configuration had to be relaxed.
Zero swap failed here. With 8 GiB and a browser the machine hit OOM daily, so a swap file was restored. The display changes held up fine, which suggests the two are more separable than assumed.
6C/12T · 32 GiB · no display
The configuration with the entire display path removed rather than minimised. Included because it is the honest control case: most of what makes Kane's Arch distinctive does not apply.
Worth noting that the gap between this and a stock headless Arch install is small. Most of the measurable difference on the other machines comes from the display decisions, not the package trimming.
Under 8 GiB of RAM, the swap policy is not recommended — see the FAQ.