OpenSUSE's Spectre Mitigation Approach Is One Of The Reasons For Its Slower Performance

[Weasel]

Whether or not certain distros "win" in benchmark speed tests is only one small point of relevance regarding whether or not to use that distro. It shouldn't be benchmarked because it's pointless - no one who uses it and wants performance is going to go with the default settings.

lol.

How about letting users make an informed decision about using a distro or not? If they don't care about performance, fine, then they ignore the benchmark. Sweeping it under the rug sounds like a case of being afraid some users will see how awful it is and reconsider.

You know what they say about being scared of facts.

[ermo]

I would argue that even if the OpenSuSE benchmarks are made on the "safe (but slow)" defaults, they do represent the out-of-box experience.

If anything, this implies that OpenSuSE is simply not targeted at the desktop per se.

Obviously, that does not mean that it can't be made to run decently as a desktop, but it seems to require skills that are normally reserved for the people who tinker with source-based distributions such as Gentoo and Exherbo (and Arch to some extent).

That said, one has to wonder why a (semi-)official desktop-kernel + assorted config tweaks isn't provided as per post #4? It seems a bit odd to me that people haven't pooled their resources in a common effort to address this particular need?

[DanL]

Michael should not benchmark openSUSE. It is not built for gaming speed and multimedia like Ubuntu, or built to win benchmark contests like Intel has done with Clear. It is the testing ground for SLES, which is used to run mainframes among other complex tasks. Anyone using openSUSE for performance will be making significant changes from the default settings to eek out performance gains.

Yeah, that seems to be the solution - that Michael should not waste his precious time benchmarking opensuse, because its defaults are not useful for desktop users, who will configure suse in their own way.

[andyprough]

I simply don't understand this mentality. "We know it won't win so don't show it in a bad light, corporate customers may see these benchmarks" is my assumption here.


It's too bad other distros can't handle complex tasks or run on mainframes ...

Whether or not certain distros "win" in benchmark speed tests is only one small point of relevance regarding whether or not to use that distro. It shouldn't be benchmarked because it's pointless - no one who uses it and wants performance is going to go with the default settings.

[Charlie68]

Michael should not benchmark openSUSE. It is not built for gaming speed and multimedia like Ubuntu, or built to win benchmark contests like Intel has done with Clear. It is the testing ground for SLES, which is used to run mainframes among other complex tasks. Anyone using openSUSE for performance will be making significant changes from the default settings to eek out performance gains.

This is not the problem, the point is another, not taking speed benchmarks as the only parameter, but this should be understood by users regardless. Speed is just one of the parameters, it's like when you do benchmarks on file systems, obviously a file system like EXT4 can be faster than other file systems that also have the task of handling things that don't exist in EXT4, but are just those functions that are missing in Ext4 that for some users are an important one. Also I can assure you that Kde Neon and Tumbleweed installed on the same PC, do not present substantial differences in terms of speed, so I believe it also depends on the hardware in use. So my advice is to leave the default settings, if you notice it's too slow, then you can work on the security pachs in question.

[xorbe]

should not benchmark openSUSE

I simply don't understand this mentality. "We know it won't win so don't show it in a bad light, corporate customers may see these benchmarks" is my assumption here.

used to run mainframes among other complex tasks

It's too bad other distros can't handle complex tasks or run on mainframes ...

[andyprough]

Yeah, back on topic. Enough of generalizing distros.

1. Are there any real world example of Skylake+ machines getting attacked with retpoline when IBRS would have saved them?
2. Why default to IBRS on all machines?
3. Should Michael force retpoline to be used when benchmarking opensuse?

Michael should not benchmark openSUSE. It is not built for gaming speed and multimedia like Ubuntu, or built to win benchmark contests like Intel has done with Clear. It is the testing ground for SLES, which is used to run mainframes among other complex tasks. Anyone using openSUSE for performance will be making significant changes from the default settings to eek out performance gains.

[Weasel]

I don't like how performance driven some people think and how also tests on Phoronix are tending a lot toward only performance.

Go use an in-order CPU then you won't need any mitigations and are 100% safe from speculative execution. Like an Intel Atom.

Who cares about performance, teh pseudo security is what matters, aye? Don't be a hypocrite now.

[DanL]

Yeah, back on topic. Enough of generalizing distros.

1. Are there any real world example of Skylake+ machines getting attacked with retpoline when IBRS would have saved them?
2. Why default to IBRS on all machines?
3. Should Michael force retpoline to be used when benchmarking opensuse?

Close to 40 comments and 90% of them are off-topic... That's really sad.

Welcome to Phoronix.

[xorbe]

The reddit link to this article and these factual findings got marked as editorial.