Benchmarks Of Clear Linux's Native Kernel Against Current/Mainline/Preempt-RT Kernels

Given the recent discussions over the default performance of the Linux scheduler, the Liquorix patches to the Linux kernel, and other recent forum discussions over different kernel configurations and flavors, here are some reference benchmarks looking at the performance of some of the kernel options available to Clear Linux users.

Besides the default "native" kernel option for bare metal Clear Linux installs, there are also current/mainline options that more closely follow the upstream state compared to the default Clear kernel configuration that generally carries a number of patches (their current kernel patches listed here). There is also a preempt-rt kernel option for those wanting a real-time kernel. More details on the various kernel flavors can be found via the ClearLinux.org bundle search.

Though at the moment the current native/current/mainline/preempt kernel options aren't all on the same series... The native kernel is on the latest Linux 5.4.6 kernel, current/mainline still tracking Linux 5.3, and the preempt-rt flavor on Linux 5.2. Due to that, it makes comparing these different kernels more difficult. As such, this is just serving as a quick one-page article for those wanting to dig into all the numbers and for alerting users to these different Clear kernel flavors should you have been unaware.

I had run various benchmarks using the Intel Core i9 10980XE Cascadelake-X system for reference.

The results varied a fair amount, but again at this time a mix of kernel versions were also at play, thus dampening the article.

The standard "native" kernel was the fastest of the tested kernels about half the time.

But the current/mainline kernels were the fastest when taking the geometric mean of all the benchmarks carried out, but some of this difference may be due to those kernel versions still tracking 5.3 rather than 5.4. Nevertheless, when all the kernels are closer to the same kernel version, I'll run some more tests. In the meantime those interested can see the dozens of benchmark results via this OpenBenchmarking.org result file.