Intel Appears On The Verge Of Some Exciting Performance Optimizations For Linux Distros

[Eirikr1848]

Do you have a gist or something with your optimizations outside of the PKGBUILDs in your repo?

Or is everything done primarily via PKGBUILD?

Let's wait and see if the results back up the hype as we don't need another APO-like let down right now from Intel. Better software tuning is very much welcome, I compile a lot of packages myself and while it is quite a lot of time consuming work to squeeze everything out of my ageing hardware, but there are rewards to be gained, e.g. Cyberpunk saw an improvement from 71 fps to 86 fps (1440p/Ultra/6950XT/Xeon 2696 V3) just by tuning a couple of game dlls (OpenSSL, zlib-ng, pcre2, libexpat - next to a custom Proton-CachyOs, vkd3d-proton and a PGOed Mesa) for my hardware.

I wonder what they cooked up, a new technique that doesn't need extra profiling and is widely applicable (even with the Linux Kernel) while reaching BOLT-levels of improvements would be great. It would already be an improvement if we could get a repository with good PGO-profiles to download. There already was some work on BOLT that improved performance for stale PGO-profiles considerably, so that idea is not too far fetched.

[ms178]

Do you have a gist or something with your optimizations outside of the PKGBUILDs in your repo?

Or is everything done primarily via PKGBUILD?

It is done primarily via PKGBUILDs which I update manually, as I am not that well-versed to automate that process yet. Feel free to adjust them to your liking and do whatever you want with them. It is a personal collection that I use myself as I was fed up with broken/unoptimized packages on the AUR which I needed to fix every time on a new installation.

[geerge]

There's the blind Intel hatred that that others imply i am imagining.

Intel's hardware is "miserable"?

I guess that's why a $600 i9 14900K is capable of beating the new 32 core AMD Threadripoffs and tying the mew 64 core AMD Threadripoffs.

Don't get me wrong, AMD is great at employing the 1960's and 1970's American muscle car approach to performance, big iron block, lots of cubic inches and massive carburetors, an approach that Dodge continues to employ in a modified form with their Hemi Chargers and Challenger, and Chevy and Ford also use that approach, while Intel employs a more finesse approach, like a 991.1 Carrera 4S.

The both have their place, AMDs are for people that think they know CPUs, Intels are for people that do know CPUs.

That is not hate, you're being deliberately obtuse. intel's hardware is currently behind on multiple fronts, objectively. Power, efficiency, core count, density. Last I checked the strongest iGPU is from AMD but that might have changed (doubt it given RDNA3 is here in iGPU form now/soon). intel might be viable on desktop because power isn't a dealbreaker, but knowledgeable people should steer clear of intel in laptops because there it normally is a dealbreaker.

Would it be nice if intel was more competitive so I could buy an AMD CPU cheaper? Sure it would. But to be honest I'd rather they focus on GPU to have competition in the non-nvidia GPU space, because I'll buy whichever of intel/AMD has the best bang for buck efficient GPU design as long as the drivers are open and good.

[qarium]

Michael
Now you did it!
Haven't you realized by now that any article or post that does not praise AMD and/or praises Intel and/or Nvidia is met with suspicion and.or derision?
You need to rewrite this statement:
So that it reads:
Then finish with something along the lines of how you are looking forward to this initiative of Intel's failing miserably because they don't know what they are doing.
See, you have to keep the AMD faithful happy, you don't want then having a fit as they read this article.

everyone should know that sophisticles
has malicious intentions he spread links to fraudulent Intel benchmarks on infected webservers who do in fact lurge in their victims to send exploids the the visitors web browser to abuse Vulnerability to finally install a Trojan horse on the victims computer to spy and also install files who are against the law to cause legal trouble for the victim.

he and his deep state left-wing friends use Israel based Trojans from the company NSO-Group.

they use Vulnerabilities like
CVE-2023-6212

https://github.com/advisories/GHSA-4cv2-qh42-x2j4

"
CVE-2023-6212
Memory safety bugs present in Firefox 119, Firefox 115.4,...

Unreviewed Published Nov 21, 2023 to the GitHub Advisory Database • Updated Nov 23, 2023

Package
No package listed— Suggest a package

Affected versions
Unknown

Patched versions
Unknown

Description
Memory safety bugs present in Firefox 119, Firefox 115.4, and Thunderbird 115.4. Some of these bugs showed evidence of memory corruption and we presume that with enough effort some of these could have been exploited to run arbitrary code. This vulnerability affects Firefox < 120, Firefox < 115.5, and Thunderbird < 115.5.0.
References

https://nvd.nist.gov/vuln/detail/CVE-2023-6212

https://bugzilla.mozilla.org/buglist...%2C1829252%2C1 856072%2C1856091%2C1859030%2C1860943%2C1862782

https://www.mozilla.org/security/advisories/mfsa2023-49/

https://www.mozilla.org/security/advisories/mfsa2023-50/

https://www.mozilla.org/security/advisories/mfsa2023-52/

https://www.debian.org/security/2023/dsa-5561

Published by the National Vulnerability Database Nov 21, 2023

Published to the GitHub Advisory Database Nov 21, 2023

Last updated Nov 23, 2023

Severity
Unknown

Weaknesses
No CWEs

CVE ID
CVE-2023-6212

GHSA ID
GHSA-4cv2-qh42-x2j4

"

https://lists.debian.org/debian-lts-announce/2023/11/msg00017.html

https://nvd.nist.gov/vuln/detail/CVE-2023-6212

https://www.debian.org/security/2023/dsa-5561

https://www.mozilla.org/en-US/security/advisories/mfsa2023-49/

" CVE-2023-6205: Use-after-free in MessagePort::Entangled

Reporter Yangkang of 360 ATA Team Impact high Description

It was possible to cause the use of a MessagePort after it had already been freed, which could potentially have led to an exploitable crash"

[Kjell]

It is done primarily via PKGBUILDs which I update manually, as I am not that well-versed to automate that process yet. Feel free to adjust them to your liking and do whatever you want with them. It is a personal collection that I use myself as I was fed up with broken/unoptimized packages on the AUR which I needed to fix every time on a new installation.

Have you considered upstreaming your tweaks to a popular distro like CachyOS?

I believe centralized effort is better than individual for longevity

[ms178]

Have you considered upstreaming your tweaks to a popular distro like CachyOS?

I believe centralized effort is better than individual for longevity

Fully agreed! In fact, ptr1337 brought me to CachyOS and we are in contact for more than a year by now. It is great that a lot of great minds help out that project that are far more knowledgeable than me (I am just a hobbyist after all). It is also funny that it gets mentioned in this thread as optimizing LLVM with BOLT started it all. Unfortunately, there are limits that can be done for a distro that needs to serve a lot of different people's needs and wants to stay compatible with Arch upstream repos. There are also limits in maintaining extra patches and compatibility problems that come with it. There is also some breakage from time to time due to me experimenting that would not be acceptable for general usage.

But few ideas made it into some AUR upstream packages already (maybe my optimizations for proton-cachyos might be worth a try), but most of my contributions are bug reports while testing the ISO and packages that I feed directly to ptr1337 or upstream projects (mostly llvm and mesa).

[Kjell]

(...)

That's awesome to hear
Thanks for taking the time to reply!

[Volta]

There's the blind Intel hatred that that others imply i am imagining.

Intel's hardware is "miserable"?

I guess that's why a $600 i9 14900K is capable of beating the new 32 core AMD Threadripoffs and tying the mew 64 core AMD Threadripoffs.

Don't get me wrong, AMD is great at employing the 1960's and 1970's American muscle car approach to performance, big iron block, lots of cubic inches and massive carburetors, an approach that Dodge continues to employ in a modified form with their Hemi Chargers and Challenger, and Chevy and Ford also use that approach, while Intel employs a more finesse approach, like a 991.1 Carrera 4S.

The both have their place, AMDs are for people that think they know CPUs, Intels are for people that do know CPUs.

Looking at some benchmarks it seems AMD CPU's with more cores use less power than Intel?