Ubuntu 24.04 LTS & Fedora 40 Continue To Trail Intel's Linux Performance Optimizations

[phoronix]

Phoronix: Ubuntu 24.04 LTS & Fedora 40 Continue To Trail Intel's Linux Performance Optimizations

While Canonical has been investing more into the performance of Ubuntu Linux and engaged some new performance improvements in Ubuntu 24.04 LTS, it's still not the fastest Linux distribution out there on x86_64 hardware. Similarly, the recently released Fedora Workstation 40 features the brand new GCC 14 compiler and other leading-edge open-source software packages, but there's still more performance left on the table as shown by Intel. Here are some fresh benchmarks looking at how Ubuntu 24.04 LTS and Fedora Workstation 40 are competing with Intel's in-house Clear Linux distribution that offers aggressive x86_64 Linux performance defaults and the best possible out-of-the-box Linux performance on modern x86_64 hardware.

https://www.phoronix.com/review/intel-ubuntu2404-fedora40

[drbuckingham]

"Canonical continues working on their performance optimizations for Ubuntu 23.10", presumably should be "Ubuntu 24.10"

[sarmad]

Can someone remind me why Clear Linux is always more performant? I assume whatever Clear Linux is doing has some downsides preventing other distros from doing it.

[drbuckingham]

My understanding is that they (Intel) maintain a significant amount of patches specifically for improving performance that are not upstreamed. At least one other distro does make use of them (SerpentOS), but I don't know how many others do. It would be a non-trivial amount of work to keep a significant set of non-upstreamed patches integrated with a distro, especially with the testing required. I imagine that is why most distros don't make the effort.

[sarmad]

My understanding is that they (Intel) maintain a significant amount of patches specifically for improving performance that are not upstreamed. At least one other distro does make use of them (SerpentOS), but I don't know how many others do. It would be a non-trivial amount of work to keep a significant set of non-upstreamed patches integrated with a distro, especially with the testing required. I imagine that is why most distros don't make the effort.

Then the question is why aren't those patches upstreamed if they improve performance?

[swastika]

Then the question is why aren't those patches upstreamed if they improve performance?

Plenty of reasons possible for this including Intel didn't submit them upstream because they made some quick hacks that upstream won't accept, breaking compatibility with older hardware, regressions for non Intel hardware, conflict with other features that aid in debugging etc. Performance only on Intel is just one of the many aspects other common distros need to be concerned about.

[drbuckingham]

My guess that is it is primarily intended for Intel CPU performance on Intel's distro; however, as noted by Phoronix they also improve AMD CPU performance as well. I have no knowledge of those patches being applied to ARM CPU code, but I suspect that most of the patches won't work, or at least provide the performance boost as that on Intel & AMD CPUs due to the differences in architecture.

[zachw]

I'm sure many would say that the minor performance difference is too little to make the decision to change distros, but in the era of cloud where an image could be auto-deployed into the hundreds or thousands, those little difference do seem to add up.

Also, I hope we can add some FreeBSD 14 to these comparisons at some point in the future. I would find that particularly interesting even though it's arguable that FreeBSD matters way less in the container era.

[muncrief]

It would be helpful if you could start including Arch and CachyOS in your benchmarks Michael. They are pure leading edge Linux distros, with CachyOS providing many optimized packages, and it would be very interesting to see how they compare to the more processed Linux distributions with older software.

[drbuckingham]

8% lead over Ubuntu 24.04 is significant when deployed at scale. Granted Ubuntu closed half the gap, as 22.04 was trailing 14% behind Clear Linux, but most of that is likely the move to -o2 (or was it -o3?) compiler options along with some of their other more recent performance improvement efforts.