Ryzen 7000 and 9000 see massive gains from Windows 11 24H2. Mostly.

[Aretak]

I see this as the narrative being pushed that windows was broke but whose to say that MS weren’t making further software optimisations to enhance existing chips rather than fixing something that is broke.

When AMD make such optimisations it’s called fine wine and are congratulated yet MS are derided for not having these optimisations ready from the get go.

Microsoft are rightly derided for it because this performance was already available on Linux. Windows is simply playing catch-up here. It's been the case for a long time that CPU-bound games run better on Linux. I remember World of Warcraft benchmarks from back in 2020 showing it running significantly better on Linux, long before Windows 11 was released. Wendell from Level1Techs also mentioned this in his recent video investigating Zen 5's performance. None of this is to say that gaming on Linux is better overall, as there are still very obvious downsides when it comes to compatibility and anti-cheat solutions, but for CPU-bound gaming it's been miles ahead for years. Even beyond the realm of gaming, CPU-reliant productivity benchmarks too are almost always in Linux's favour.

So yes, it's entirely fair to say that Microsoft very much are fixing something that was broken. This isn't extra performance being pulled out of thin air. It's performance that was always there and available on another operating system finally being made available to Windows users. Or at least partially so. I don't know how close this brings Windows to parity. There were some pretty big gaps to bridge (Shadow of the Tomb Raider ran 40% faster on Linux in Wendell's testing for example).

 

[ltron]

just did the update came up in windows update, interested to see what kind of increase on 5800x3d

Zero improvement for my 3900X, but that was to be expected.

Some people are saying they are getting worse performance on 23H2 with the patch, so perhaps there's something else about 24H2 that's contributing to the performance increase.

 

[alec]

Ryzen speed boosts get backported to Windows 11 23H2 with optional update

And it turns out that old Ryzen CPUs benefit almost as much as newer ones. See full article...

arstechnica.com

tempt:
Rough short summary from my understanding so far:

• Windows was purging some of the caches/buffers on the CPU, this was meant to mitigate/solve the Spectre/Meltdown problems that affected Intel CPUs. This has a negative impact on performance specially on Zen5 that has a larger branch prediction buffer and instruction window.
• Running in super/hidden Admin mode disables/bypasses some of these security features hence giving more performance.
• It appears that for Zen 3 to 5 those mitigations aren't necessary, or at least not in the way Windows was doing them, the new update addresses that.

Sums up my suspicions
Meltdown/Spectre mitigation done differently or discarded

AMD Ohshit moment after admin mode difference discovered, and quick patch thrown together by MS.
All "branch prediction" story boils down to "do things less stupid"

 

[Tee Hee Johnson]

I done a quick wee benchmark before and after installing KB5041587 .

Only Cyberpunk as I couln't be bothered running more or cinebench, etc.

Seen a small uplift, but nothing to shout about. More performance is always good tho!

Oh Windows 11 Education/Pro was fully up to date, Nvidia and AMD Chipset drivers updated last week. 5800X3D, 32GB @3600Mhz, ASUS 3080Ti TUF OC of some variety.

Ranging between .46% better to 3.67% better depending on highs, lows and averages.