Cyclop OS · Benchmarks
No figure until it is measured properly
This page holds real measurements only. Our first kernel comparison ran on one machine with background load and three runs: not enough to publish. Here is the method the published results will follow.
Results published with Patch 3
The rules
A number without its conditions says nothing. Each result will carry all of them.
- The full configuration next to every result: processor, graphics card, driver version, memory, resolution and in-game settings.
- The same machine for both sides of a comparison, nothing else changed between the two.
- At least three runs per test, on a quiet machine. We aim for five.
- Games: the average frame rate and the 1% low, recorded with MangoHud.
- The raw results published with the summary, so anyone can redo the sums.
- Where a test exists in the Phoronix Test Suite, we use it, so the result can be compared on OpenBenchmarking.org.
Kernel: Ubuntu against Cyclop
One installed system, two kernels picked in the boot menu, the same script on both.
- Scheduler
- hackbench, perf sched pipe, futex
- Compute
- zstd compression, openssl, a kernel compile
- Storage
- sequential and random disk reads and writes
- Start-up
- boot time to the login screen
A result that varies more between runs than between kernels will be published as "no measurable difference", not rounded in our favour.
Games
Titles with a built-in benchmark or a repeatable scene.
- Counter-Strike 2
- Cyberpunk 2077
- Shadow of the Tomb Raider
- Hitman 3
- Baldur's Gate 3
- Elden Ring
A game is compared on the same machine against its previous system, with the same settings. A game that does not run is written down as such.
Results
With Patch 3
The first published series will cover the kernel comparison above, with its raw data. Until then this page shows no figure, and nobody at Cyclop quotes one.
The kernel behind the tests
What linux-cyclop changes, where its source is and how to rebuild it.