Info AMD confirms Windows 11 slow down its CPUs up to 15%

[deasd]

AMD processors hit by performance issues on Windows 11, patch on the way

AMD has said that its CPUs are taking a performance hit on Windows 11 systems and that it has teams working on the issues. Patches should be released later this month through Windows.

www.neowin.net

AMD: Windows 11 Slows Our CPUs Up To 15%, Patch Coming

High latency bug hits fast

www.tomshardware.com

AMD: Windows 11 May Cause Performance Dips Of Up To 15% On Ryzen CPUs

AMD reported issues that their Windows 11 compatible AMD Ryze CPUs were having with performance while running certain applications.

wccftech.com

The issue mentioned in the opening paragraph is caused by functional L3 cache latency which has increased by around three times on affected hardware. The programs that will suffer include those sensitive to memory subsystem access times. AMD noted another problem too, explaining that UEFI CPPC2 may not schedule threads on the processor’s fastest core preferentially.


Regarding the latter issue, applications sensitive to the performance of one or a few CPU threads will see a performance hit. The issue will be more noticeable on greater than 8-core processors that operate at over 65W. This issue should also be fixed this month.

hmmmm, whose fault is it this time? It screw up the future hardware review if true? what about Intel side?

 

[leoneazzurro]

I gave you the evidence, you are continously ignoring it because "no reputable source". I suppose you want the pope. YOU are no reputable source. And you are continuing with the nonsense. As you don't stop trolling welcome in the ignore.

 

[LikeLinus]

I gave you the evidence, you are continously ignoring it because "no reputable source". I suppose you want the pope. YOU are no reputable source. And you are continuing with the nonsense. As you don't stop trolling welcome in the ignore.

What evidence? You provided no link with real evidence. All you posted was a Reddit thread with a thousand replies, counting it as the "truth" and you keep dogging the real question. I never said I was a reputable source, That's why I gave you links with comments directly from AMD. You can't even understand that the drive issue is being done via a software update and not a windows update because it's a driver written by AMD. But you state AMD is aware of it and was aware of it before the release? Even saying they told Microsoft and MS ignored it? LOL. Delusional with no proof.

Feel free to ignore me.

 

[DrMrLordX]

Looks like the dev channel already has a fix implemented. I just switched from beta to dev of Win11 and I got all my missing L3 performance back on my 3900X, though the L3 read and write bandwidth numbers jump around like crazy. Not sure why.

 

[leoneazzurro]

Looks like the dev channel already has a fix implemented. I just switched from beta to dev of Win11 and I got all my missing L3 performance back on my 3900X, though the L3 read and write bandwidth numbers jump around like crazy. Not sure why.

But you are not a reputable source, you know, even if you can test it directly.

 

[DrMrLordX]

But you are not a reputable source, you know, even if you can test it directly.

Heh! Sure. Meanwhile:



That's a "normal" run under 22741.1000 with the latest AIDA64. Looks like the L3 performance read/write dips when there are other applications running in the background (including the browser). Anyway, I am satisfied that MS has a good lead on how to deal with the L3 cache Ryzen issue. Not sure about the "preferred core" problem but I'm not all that worried about it myself since I usually run with static clocks anyway.

 

[LikeLinus]

But you are not a reputable source, you know, even if you can test it directly.

He's talking about the Microsoft Dev Channel, hosted by Microsoft. It's different from the Insider Program in that it has current development builds available for download. So yes, that is a very reputable source.

 

[DrMrLordX]

He's talking about the Microsoft Dev Channel, hosted by Microsoft. It's different from the Insider Program in that it has current development builds available for download. So yes, that is a very reputable source.

. . . am I?

Anyone can join the Insider Program and get into the dev channel. Granted I've been an Insider since . . . uh . . . well it's been awhile. 2014 or something? Whatever, doesn't matter. Anyway, all you do is:

Join the Insider Program, install Win11 preview <whateveritisyouget>
Right click Win11 start button
Select Settings
Select Windows Update on the left panel
Hit the Windows Insider Program button
Hit the Choose your Insider Settings button
Select the Dev Channel radio button
Then you go back and search for new updates, and bam, it should hook you up with 22741.1000

Alternatively you can possibly get an .iso of 22741.1000 and install it directly, and join that way.

 

[LikeLinus]

. . . am I?

Anyone can join the Insider Program and get into the dev channel. Granted I've been an Insider since . . . uh . . . well it's been awhile. 2014 or something? Whatever, doesn't matter. Anyway, all you do is:

Oh, I know, I'm well aware. I have it setup on my Surface Pro 7 since it's basically my "around the house" laptop. I've left all of my mission critical systems on Win10 for now. I remember the days of having MSDN binders full of CDs from Win2k, Visual Studio, SQL and all that fun stuff

My comment was towards the one who is trying to be snide with his remark.

 

[Markfw]

OK, this post will be from a fact perspective, then a little logic and reasonable conclusions.

1) fact: Windows is a Microsoft product, nobody else is allowed to code it or sell it. They alone are responsible for their product.
2) Microsoft knew that Intel was coming out with a new generation of CPUs and they being the biggest, Microsoft wanted their new OS to take advantage of this new design.
3) Microsoft obviously did not take into account CPU maker number 2, who actually was the predominant CPU producer (as far as performance in many criteria) at the time.

Now for opinion.
1) I think Microsoft really wanted to release this before Intel actually released their new product, to increase their sales, and Intels, the numerically dominant entity.
2) I am sure they knew about the AMD problem, but really did not care if it was in the first release or a subsequent bug fix version. See point 2 above.

All the discussion about is it a bug, or is it AMD's problem, are all irrelevant, its a Microsoft problem. And nobody has even talked about Linux, which 90% of my machines use.

 

[DaaQ]

"Do you think that they did not tell it to MS, and why? When they went public about the issue on the launch day?"

So, why you're also saying that AMD was aware of the scheduling issue because the issue was released on the same KB article.

https://www.amd.com/en/support/kb/faq/pa-400

You can't have it both ways. They were either aware of the issues before the launch or after. If before, they are just as complicit as MS in the handling of the situation.

What they should have done, for the UEFI issue, is proper testing on a Beta OS and already finalized a new driver for launch day. Not let it ride out and then say "Oh yeah, there is a problem with our driver, we'll fix it in the next few weeks!".

MS didn't just surprise drop Win 11. AMD had time to evaluate any performance issues and such and to warn users, BEFORE HAND, or to flat out fix the driver issues.

Either way, facts and evidence stand, both companies messed up. FYI - The conspiracy theory joke wasn't about you. It was about previous ones implying they did this for the upcoming Intel CPUs, lol.

FYI - I'm not "ignoring evidence". Feel free to point me to any facts and data in that Reddit thread that is actual evidence. I'm not scrolling through a thousand replies to find a needle in a haystack from some forum member's response that you claim is evidence that this was a "FIX" (your words) in a previous Beta and worked fine and then MS removed the fix. I gave you links and plenty of real evidence from AMD and reputable sites. You grabbing a link to Reddit and calling it evidence is laughable.

Let us know how your Alder lake runs please.

 

[LikeLinus]

Let us know how your Alder lake runs please.

Don't have one. I'm running a 5950X, 5800X, 2700X, MacBook Pro, server is dual E5 V3 Xeon CPUs, and my Surface Pro 7 is an i7.

 

[LikeLinus]

OK, this post will be from a fact perspective, then a little logic and reasonable conclusions.

1) fact: Windows is a Microsoft product, nobody else is allowed to code it or sell it. They alone are responsible for their product.
2) Microsoft knew that Intel was coming out with a new generation of CPUs and they being the biggest, Microsoft wanted their new OS to take advantage of this new design.
3) Microsoft obviously did not take into account CPU maker number 2, who actually was the predominant CPU producer (as far as performance in many criteria) at the time.

Now for opinion.
1) I think Microsoft really wanted to release this before Intel actually released their new product, to increase their sales, and Intels, the numerically dominant entity.
2) I am sure they knew about the AMD problem, but really did not care if it was in the first release or a subsequent bug fix version. See point 2 above.

All the discussion about is it a bug, or is it AMD's problem, are all irrelevant, its a Microsoft problem. And nobody has even talked about Linux, which 90% of my machines use.

#3 is not a fact, it's an opinion. While it's true that MS owns the code and no one else can sell it without an agreement, vendors do work with Microsoft to create drivers.

AMD’s UEFI CPPC2 driver is not a Microsoft problem. That is why AMD says they will update it through software/driver and not a Microsoft Update. The schedule issue is due to something Microsoft changed, but It's surprising it wasn't caught until after release. Thankfully, but AMD and MS are working on it and it'll be fixed this month.

 

[BorisTheBlade82]

Do you have a source for this story?

This is taken from the book "Inside Intel" from Tim Jackson - very nice read I can tell you.

 

[Panino Manino]

#3 is not a fact, it's an opinion. While it's true that MS owns the code and no one else can sell it without an agreement, vendors do work with Microsoft to create drivers.

AMD’s UEFI CPPC2 driver is not a Microsoft problem. That is why AMD says they will update it through software/driver and not a Microsoft Update. The schedule issue is due to something Microsoft changed, but It's surprising it wasn't caught until after release. Thankfully, but AMD and MS are working on it and it'll be fixed this month.

By your own explanations you're pointing that it's NOT AMD's fault.
You yourself said more than one time that the AMD driver (that may be contributing to the problem but may also not be the only culprit) was working as intended on Microsoft's current OS, Windows 10, probably following Microsoft's guidelines. You yourself is saying that AMD's driver is working.
What changed, the driver or the OS?

Yes, the OS.
It's AMD's fault if some new feature that should be for exclusive use by a future rival's product that wasn't even launched yet is causing problems for it's CPUs? Something that AMD have no ways to control, nor say in how that new feature should work?
The most that AMD can do is pay attention to "bugs" on the development builds and ask Microsoft to fix it, and we have evidence that this was done and it was working as it should in recent builds. It's AMD's fault if Microsoft launched W11 on a broken build? That they did this giving zero warnings to the users? Why wasn't the fix available at day 1 if it was already available?
A new driver by AMD is not a fix, it's a workaround.

And isn't concerning that Microsoft rushed a new OS that may have changed rules about how drivers should be made and may cause problems for many others besides AMD?

 

[Zucker2k]

Looks like the dev channel already has a fix implemented. I just switched from beta to dev of Win11 and I got all my missing L3 performance back on my 3900X, though the L3 read and write bandwidth numbers jump around like crazy. Not sure why.

Wait for proper fix from AMD. They're aware of the problem and working on a patch.

 

[leoneazzurro]

Wait for proper fix from AMD. They're aware of the problem and working on a patch.

The AMD patch is about the "preferred core" issues, not the L3 cache issue, the fix for this last one will come from MS.

 

[Zucker2k]

1. Windows 10 has a hybrid scheduler too, y'all remember Lakefield? It runs on Windows 10, on the hybrid scheduler.
2. Intel is using Thread Director ON TOP of any changes MS may have done to the scheduler
3. Intel's other non hybrid processors are not affected by scheduling issues
4. AMD is aware of the problem and WORKING ON A FIX.

This last point indicates that AMD, is well aware of the fact that they don't have a fix for the problem. So, all throughout the period of testing, did AMD not catch this issue with their testing? Why are they now working on a fix? How about all the beta testers using AMD hardware? Did they not catch and report the issue? What did AMD do about it? How long has AMD known about this problem? It's not like this thing only surfaced on release day? If so, that'll be an indictment of AMD's own testing.

5. MS releases new OSes around the back to school period. Are y'all suggesting MS should have held back on releasing W11 because AMD cpus suffer a 4-5% hit in performance in certain cases? The onus is on AMD, as it is with Intel, to make sure to test and assess how they're platforms run on the new OS. The fact AMD has acknowledged the problem and promised a fix should've ended the conspiracy theories. It also tells me AMD is guilty of not responding on the issues on time.

 

[Zucker2k]

The AMD patch is about the "preferred core" issues, not the L3 cache issue, the fix for this last one will come from MS.

How so? According to @DrMrLordX his 3900x is still acting funny even on the latest Dev Build, which is supposed to have fixed the L3 issue; you even linked the reddit thread with the screenshot.

 

[leoneazzurro]

How so? According to @DrMrLordX his 3900x is still acting funny even on the latest Dev Build, which is supposed to have fixed the L3 issue; you even linked the reddit thread with the screenshot.

He recovered the performance loss, so the fix is working, the AIDA fuzzy numbers could be a glitch with the application combined with the fix (i.e. this has already happened in the past with AIDA not reading properly the cache bandwidth and latency on Ryzen systems). But this is about the problem with the cache latency. The one AMD is supposed to patch is the one about the preferred core issue.

Windows 11 bug could reduce Ryzen CPU performance by up to 15%, AMD says

Software fixes for the problems are coming later this month.

arstechnica.com

Here you can clearly read that the L3 issue will be fixed in a Windows update, while the preferred core with a "software update" coming quite probably from AMD.

Edit: mispelling

 

[TheELF]

OK, this post will be from a fact perspective, then a little logic and reasonable conclusions.

1) fact: Windows is a Microsoft product, nobody else is allowed to code it or sell it. They alone are responsible for their product.
2) Microsoft knew that Intel was coming out with a new generation of CPUs and they being the biggest, Microsoft wanted their new OS to take advantage of this new design.
3) Microsoft obviously did not take into account CPU maker number 2, who actually was the predominant CPU producer (as far as performance in many criteria) at the time.

1) In the same vein, only intel/AMD have any right to use intel/AMD IP
If fixing this needs any sort of CPU internal address for example then MS can't just use that even if they know it, because they have no right to the IP, and I'm sure that windows 11 is legally a different thing from windows 10 so even if they have the numbers from windows 10 they still can't use them because they only got the rights to win 10.
2) intel send coders to MS to help implement any performance tweaks possible, that's what a CPU maker is supposed to do.
3) AMD has never been the first to develop for their own hardware, they waited for others to come up with mantle for example.

Now for opinion.
1) I think Microsoft really wanted to release this before Intel actually released their new product, to increase their sales, and Intels, the numerically dominant entity.
2) I am sure they knew about the AMD problem, but really did not care if it was in the first release or a subsequent bug fix version. See point 2 above.

All the discussion about is it a bug, or is it AMD's problem, are all irrelevant, its a Microsoft problem. And nobody has even talked about Linux, which 90% of my machines use.

1) Well obviously they wanted to release well before the holidays so that OEMs have plenty of time to dump win 11 on all of their new hardware. How stupid would they have to be to NOT do that?! Intel has the same reasons to release right now, AMD doesn't have the luxury to release whenever they want because they depend on third party FABs otherwise AMD would have something ready now.

2) Yes in a world where everybody has the internets this isn't even an issue, even if you are the first person to install win11 and loose some performance for a month or so, it's not like it's a big deal, especially since cache mostly affects benchmarks and not real world apps, but then again AMD really needs those benchmark numbers to look good.

 

[Zucker2k]

Here you can clearly read that the L3 issue will be fixed in a Windows update, while the preferred core with a "software update" coming quite probably from AMD.

I don't think you understand what this means. Just because it says Windows update doesn't mean MS is writing the entire code. AMD is going to be involved in it. This should've happened in the past. Not now.

 

[leoneazzurro]

I don't think you understand what this means. Just because it says Windows update doesn't mean MS is writing the entire code. AMD is going to be involved in it. This should've happened in the past. Not now.

Lol. I understand more than you, don't worry. And as said, the bugs were known -and MS was told- before the launch date, one for sure since August.

 

[DrMrLordX]

How so? According to @DrMrLordX his 3900x is still acting funny even on the latest Dev Build, which is supposed to have fixed the L3 issue; you even linked the reddit thread with the screenshot.

See the followup post. L3 read/write bandwidth normalized when I ran Aida64 without any other applications in the background.

 

[Zucker2k]

See the followup post. L3 read/write bandwidth normalized when I ran Aida64 without any other applications in the background.

That's all well and good. This means the fix is in the channel, which again means AMD was too late with the fix to make the RTM. Thing is MS OS releases always exhibit quirks that are ironed out in the weeks and months following release. It seems to me because of ADL release, Intel has been more active to get things right out of the door. Thread Director was necessarily a collaboration between Intel and MS. Throughout the history of MS, there has always been patches released to address specific performance issues with certain vendor hardware. This won't be the first or last.

 

[DrMrLordX]

which again means AMD was too late with the fix to make the RTM.

. . . yeah, okay. You just stick with that.