Intel's Clear Linux Is Working On Steam Support

[arjan_intel]

I doubt't that too This takes advantage of specific intructions found in HSW+ hardware only.



Yeah it is that simple or It is much more simplier than maintain for diversive hardware... you know you just apply patch and don't care about else ("works for me, but not for you as i don't care about you" sounds already easier isn't it) it is much harder to do same on generic distros, where it is required of course to also care about older

Clear Linux goes back to westmere (2010) and runs on the current generations Atom as well... (so 10+ generations at least across the two product lines)
you seem to imply it's HSW+ only and that's outright not correct.... so it's not just "use AVX" it's "detect AVX when available and then use it", which even more generic distributions can do just as well

[dungeon]

OK, so it is generic partialy... but i mostly think of that perf adventage in mind, as those certainly goes from new instructions found on newer hardware.

Sorry but i am from Debian community where generic has different meaning, our generic means we even support Pentium PRO i think

But yeah i expect generic distros will get AVX if it is done in generic detect way of course.

[AsuMagic]

OK, so it is generic partialy... but i mostly think of that perf adventage in mind, as those certainly goes from new instructions found on newer hardware.

Sorry but i am from Debian community where generic has different meaning, our generic means we even support Pentium PRO i think

But yeah i expect generic distros will get AVX if it is done in generic detect way of course.

x86-64 per definition supports SSE2 though.

[andre30correia]

Excellent: the first linux operating system optimized for gaming. I'm waiting for native opengl games.

for gaming with intel igpu lol

[dungeon]

x86-64 per definition supports SSE2 though.

Yep, but that does not mean all apps uses it... that only describe requirement and make sense since all 64bit x86 CPUs support instruction.

Even if you use 32bit OS, some apps which has SSE2 or whatever else instruction code paths will take advantage of it... of course if your CPU supports it.

[andre30correia]

Speculative mode on/

Could be that Valve is asking then to make way to a SteamOS move from Debian to Clear?

i believe more in make steamos in ubuntu, clear linux is intel only

[arjan_intel]

OK, so it is generic partialy... but i mostly think of that perf adventage in mind, as those certainly goes from new instructions found on newer hardware.

Sorry but i am from Debian community where generic has different meaning, our generic means we even support Pentium PRO i think

But yeah i expect generic distros will get AVX if it is done in generic detect way of course.

I remember running Debian on a Pentium Pro.... 17 years ago It had I think 32Mb or 64Mb of ram.. huge for the time. It was also built like a tank :-)

But if you leave, say, 10% performance on the table for the maybe 100 people who run a current debian on 64Mb PPRO still ...... I respect Debian a lot for what they do, but that's a tradeoff something like Clear Linux makes differently.

And with runtime checks that's also a point largely moot ... pretty much all the clear linux using AVX stuff is runtime detection level

[slacka]

Yeah it is that simple

If it's just a few compiler options, then how come I haven't seen any benchmarks from Gentoo showing that it performs equally well? Every time I've seen Gentoo put up against traditional distros like Debian and RHEL, it never shows the kind of performance gains we're seeing with Clear. So there's something special about these new CPUs or more to it. I won't buy that it's just the CPU until I see some real hard numbers.

[dungeon]

If it's just a few compiler options, then how come I haven't seen any benchmarks from Gentoo showing that it performs equally well? Every time I've seen Gentoo put up against traditional distros like Debian and RHEL, it never shows the kind of performance gains we're seeing with Clear. So there's something special about these new CPUs or more to it. I won't buy that it's just the CPU until I see some real hard numbers.

I didn't check what it is, but there is a lot of crap by default on other distros too. Recent gaming benchmarks of Clear Linux, show only advantage on Xonotic low settings... but that is weirdest benchmark when even AMD with opensource driver nearly beat nvidia driver

And why it is weird, well i found something with sound is wrong (it uses plain alsa, not openal or so) so take a lot of CPU time when you run benchmark first time but not second where Michael actually uses only first as representative, second point - it does not setup resolutions properly particulary lower than native if benchmark is run from commandline, third - benchmark rendering is borked in 0.8.1 thus Michael uses 0.8. And now that actually also benchmarking my network via curl since it wanna to be updated during benchmark, forth - that is enough really

They don't care, i don't care... and if you wanna benchmark Xonotic Linux against Windows, perf is entirely different probably unrelated to GPU and drivers again as there is not different features only different unrelated issues

Refering to this article:

http://www.phoronix.com/scan.php?pag...ro-Distro-2016

You see, since i don't believe Xonotic low tell me what example to use there to prove advantage... i also found GPUtest does not go to fullscreen even when you wanna fullscreen on some window managers, which of course leads to bit faster results, but to me if that happen it is invalid so making moot point again

[slacka]

Recent gaming benchmarks of Clear Linux, show only advantage on Xonotic low setting

Low settings are exactly where you'd expect to see Clear's CPU optimizations shine. Not in high resolutions/settings where you become GPU bound.

And why it is weird, well i found something with sound is wrong (it uses plain alsa, not openal or so) so take a lot of CPU time when you run benchmark first time but not second where Michael actually uses only first as representative, second point

Sounds like you may have found an error in this methodology. You should let him know if you turn anything up.

I'm not saying you're right or wrong about this only being a compiler optimization. but I'm going to reserve judgement until I see some benchmarks(this is what this site is about about isn't it?). Historically, Gentoo has always been one of the fastest Linux distros. But I've never seen numbers like we're seeing with Clear.