Performance tests

We should run the tests on this page periodically, at least once at the end of each release cycle, to guard against regressions.

Pay attention to the following when running performance tests:

• Make sure you're not doing anything resource-intensive in the background (e.g., downloading large files, copying files to an USB drive, performing OS updates or virus scans, running games or anything performance-intensive, including browsers with many tabs open, etc.)
• Run the tests at least three times and take the best result.
• If using a laptop, connect it to the power adapter to prevent it from entering "energy saver" mode.
• Display resolution (e.g., 1080p vs 4K display) has a large effect on performance, even if the rest of the system is identical.
• This should be fairly obvious, but use the same system if possible for regression testing, using the same desktop resolution.

Quake benchmark

This is our standard benchmark to guard against performance regressions. The Quake benchmark is the best tool for the job as it is one of the most demanding DOS programs, and it can scale to pretty much any CPU speed (even up to Pentium 4 speeds), so the bottleneck will always be the emulator. Always use the included configuration to ensure comparable results.

quake-benchmark.zip