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?

 

[DrMrLordX]

A solution for reviewers I think would be to continue to use Windows 10 for Ryzen until Windows 11 has been sorted out.

Under "normal" circumstances, a user wouldn't swap CPUs on a Win11 install that often anyway. Some tinkerers would do it but that's about it. If reviewers can't be arsed to reinstall the OS every time they switch the CPU then yes, running Win10 on the AMD box would be the correct solution.

 

[igor_kavinski]

Not so sure, people went to Win 10 because it gave a performance boost.
I know that is the reason why I didn't stay on Win 7.
I don't see any incentive to upgrade to Win 11 now for me, it's even the opposide I would feel stupid if I do.

Have you tried Windows 11? It might be slower in benchmarks but the scheduler has been tweaked to make the UI more responsive so maybe it will feel "faster" than Windows 10.

 

[jpiniero]

I don't see any incentive to upgrade to Win 11 now for me, it's even the opposide I would feel stupid if I do.

You're going to be force upgraded to 11 sooner rather than later unless your PC doesn't meet the requirements.

 

[SamMaster]

Looks like there is no performance difference between W10 and fixed W11 to speak of, except random SSD IO (towards the end of the video).

 

[moinmoin]

Ah. So Intel might be counting on the possibility that some kernel developer will get annoyed enough to write their own scheduler routine for Alder Lake.

I expect the existing Linux schedulers to be good enough already for the developers to care. Phoronix may show whether that's the case. Michael wrote "One of the main open questions is around Intel Thread Director efficiency on Linux, but my benchmarks will be able to shed light on that shortly." so I hope his benchmarks will be able to clarify all the current confusion.

 

[Kedas]

Looks like there is no performance difference between W10 and fixed W11 to speak of, except random SSD IO (towards the end of the video).

It does indeed look about the same after the fix, but where is this big increase in random SSD IO coming from.
Was Win 10 so bad or is this more like a benchmark change instead of a real usage change. It certainly didn't load windows faster.
(That is probably for an other thread)

 

[Timorous]

Looks like there is no performance difference between W10 and fixed W11 to speak of, except random SSD IO (towards the end of the video).

In 4 games the average uplift was just over 3% with two games having 5% uplifts, one having 3% uplift and the last one having no difference. That kind of difference will make the ADL vs Zen 3 gaming battle a lot closer. I am sure that ADL will still come out ahead by a reasonable margin vs the 5900X / 5950X but when it comes to the 12600K vs 5600X that is looking more and more like it is going to be a dead heat and will entirely depend on memory, power settings, cooler and game suite chosen.

 

[igor_kavinski]

I expect the existing Linux schedulers to be good enough already for the developers to care. Phoronix may show whether that's the case. Michael wrote "One of the main open questions is around Intel Thread Director efficiency on Linux, but my benchmarks will be able to shed light on that shortly." so I hope his benchmarks will be able to clarify all the current confusion.

Pity no one tested Lakefield under Linux.

 

[TheELF]

but when it comes to the 12600K vs 5600X that is looking more and more like it is going to be a dead heat and will entirely depend on memory, power settings, cooler and game suite chosen.

Yes, they will very likely be very close since gaming benchmarks tend to be on a sterile fresh install and only running a canned bench.

But the 12600k will still have a decently strong gpu that can handle streaming, video transcoding, anything that uses opencl or gpgpu, or you can just use it for troubleshooting or to have a second display with whatever resolution or refresh you want to, without any interference to your main GPU.

And the 12600k will have essentially another 4 core CPU on top because the thread scheduler will keep the two types of cores as separated as possible, 4 cores that you can use for whatever you want be it streaming, transcoding, rendering or whatever, just as an buffer to keep your background stuff from reducing your FPS.

A lot of people will not care because they will never use anything else when gaming but a lot of people will care.

And it's going to be the same price and probably cheaper, can't beat that value.

 

[Timorous]

Yes, they will very likely be very close since gaming benchmarks tend to be on a sterile fresh install and only running a canned bench.

But the 12600k will still have a decently strong gpu that can handle streaming, video transcoding, anything that uses opencl or gpgpu, or you can just use it for troubleshooting or to have a second display with whatever resolution or refresh you want to, without any interference to your main GPU.

And the 12600k will have essentially another 4 core CPU on top because the thread scheduler will keep the two types of cores as separated as possible, 4 cores that you can use for whatever you want be it streaming, transcoding, rendering or whatever, just as an buffer to keep your background stuff from reducing your FPS.

A lot of people will not care because they will never use anything else when gaming but a lot of people will care.

And it's going to be the same price and probably cheaper, can't beat that value.

For cheap productivity the ADL processors will probably be punching up at the next tier of zen3 cpus in cases where you can use all the threads. 12600K matching 5800X and 12700K matching 5900X does seem feasible in many workloads.

As for same or cheaper. Processor to processor probably, platform to platform is going to be a bit different.

 

[Thibsie]

And it's going to be the same price and probably cheaper, can't beat that value.

That including new expensive mobo and expensive DDR5?
Value it is indeed. Not sure the cheaper one though.

 

[Joe NYC]

Considering the performance difference in most games for AMD will be less than two digit percentage points, most AMD users will just swallow their pride and keep on using Windows 11. Even those that are adamantly staying on Windows 10, will begrudgingly have a change of heart sooner or later.

It is not about pride or being adamant. It is about Windows 11 being buggy.

In 6 to 12 months Windows 11 may be on par with Windows 10, so that impediment will go away.

 

[Zebo]

I still use windows 7, blackviper'd like 9/10ths of services off and only using 400MB ram on start...why wont any sites run benches with that?

 

[moinmoin]

blackviper

Thanks for that mention, didn't know these guides exist.

 

[jpiniero]

I still use windows 7, blackviper'd like 9/10ths of services off and only using 400MB ram on start...why wont any sites run benches with that?

Sadly not supported. Skylake for instance was the last Intel processor that supports Windows 7.

 

[DrMrLordX]

I still use windows 7, blackviper'd like 9/10ths of services off and only using 400MB ram on start...why wont any sites run benches with that?

Very few people mod their Windows install in such a fashion.

 

[Panino Manino]

I still use windows 7, blackviper'd like 9/10ths of services off and only using 400MB ram on start...why wont any sites run benches with that?

How you do that?
I wish I knew exactly what services I can disable without causing problems.

 

[TheELF]

That including new expensive mobo and expensive DDR5?
Value it is indeed. Not sure the cheaper one though.

Well show me an AM4 mobo with ddr5 and pci5 support so we can compare prices...
(and add an GPU no matter how cheap and add another laptop quad cpu )
It's going to take another year until we can finally compare prices.

There will be cheaper mobos and ddr4 is always an option, according to benchmarks it's a mixed bag now and it might stay that way, with many games running better on ddr4 and many on ddr5 so if you want to save some money you can opt for ddr4 without loosing out.

 

[Shamrock]

Welp, Derbauer just dropped a video about beloved DDR5...TLDR? DDR5 is only better for synthetic benchmarks. DDR4 can even beat it in games. It wasnt cheap DDR5 either. Corsair Dominator DDR5 5200 C36.

 

[Dave3000]

I went back to Windows 10. In Windows 11, even after the MS AMD performance update and installing the latest AMD chipset drivers, the L3 cache write bandwidth is still much lower than what I was getting in Windows 10. In Windows 11 in AIDA64 the L3 cache write bandwidth of my Ryzen 5800x was in the 300's GB/s and even sometimes dropping to the 100's GB/s, despite the L3 cache latency being just a little higher than what I was getting in Windows 10. In Windows 10, AIDA64 I got in the 500's GB/s for L3 cache write bandwidth and consistently as well. In Windows 11, In Flight Simulator (2020) I noticed a frequenty microstuttering if I maxed out the terrain LOD (400) while flying over Los Angeles, however, in Windows 10 with the same settings, I hardly got any microstuttering with terrain LOD maxed out. Despite that microstuttering, my overall frame rate was a little better in this game (56 in Windows 10 and 58 in Windows 11) but I rather take a smoother experience than a slightly better overall frame rate. I don't know if this microstuttering in Flight Simulator (2020) is caused by the much lower L3 cache write bandwidth performance that I'm getting in Windows 11 according to AIDA64. My GPU is a GTX 1080 Ti.

 

[VirtualLarry]

Welp, Derbauer just dropped a video about beloved DDR5...TLDR? DDR5 is only better for synthetic benchmarks. DDR4 can even beat it in games. It wasnt cheap DDR5 either. Corsair Dominator DDR5 5200 C36.

But... is this true of DDR5 in general (in which case it would also likely apply to Zen4/5), or is this an performance anomoly, that only applies to Alder Lake?

 

[Thibsie]

At this point DDR5 will lose lantency tests against current DDR4 simply 'cos DDR4 has gone way beyond JEDEC speeds IMO.
DDR5 latency will improve with time, as it is always the case with new RAM standards...

 

[Joe NYC]

Windows 11 a complete flop.

"Microsoft Windows 11, now nearing its third month since release, is for all intents and purposes, a flop. Market research by Lansweeper, which surveyed over 10 million PCs across the commercial and personal market segments, reports that less than 0.21% of the users it surveyed, had upgraded from Windows 10 to the newer operating system. The upgrade is free of charge."
Windows 11 a Flop, Survey Claims Less Than 1% Upgraded, Microsoft Improves Start Menu | TechPowerUp

(Maybe the word got out that Windows 11 is garbage)

 

[OlyAR15]

Windows 11 a complete flop.

"Microsoft Windows 11, now nearing its third month since release, is for all intents and purposes, a flop. Market research by Lansweeper, which surveyed over 10 million PCs across the commercial and personal market segments, reports that less than 0.21% of the users it surveyed, had upgraded from Windows 10 to the newer operating system. The upgrade is free of charge."
Windows 11 a Flop, Survey Claims Less Than 1% Upgraded, Microsoft Improves Start Menu | TechPowerUp

(Maybe the word got out that Windows 11 is garbage)

Yet in the November Steam hardware survey, Win 11 is already over 8%, second to Win 10. That’s the problem with surveys. It depends on who you survey.

 

[LikeLinus]

Windows 11 a complete flop.

"Microsoft Windows 11, now nearing its third month since release, is for all intents and purposes, a flop. Market research by Lansweeper, which surveyed over 10 million PCs across the commercial and personal market segments, reports that less than 0.21% of the users it surveyed, had upgraded from Windows 10 to the newer operating system. The upgrade is free of charge."
Windows 11 a Flop, Survey Claims Less Than 1% Upgraded, Microsoft Improves Start Menu | TechPowerUp

(Maybe the word got out that Windows 11 is garbage)

Where is the link to LanSweeper or the research data? The article provides NOTHING to support it's claim.
I could only find one "LanSweeper" - a company who sells IT asset management.

Is Your Business Ready for Windows 11?

Windows 10's end-of-life is arriving soon and migration to Windows 11 is key to ensuring a safe IT environment, business adoption is increasing.

www.lansweeper.com

"Lansweeper combines market-leading scanning technologies to scour the network and collect detailed information about every connected device. "

"Better yet, our experts created a Windows 11 Readiness Audit Report that can be run against the asset inventory to check for machines that are eligible for the Windows 11 upgrade. The report presents all the information in a visual, digestible format, making it easy to create a migration plan."

"Lansweeper customers will benefit from rapid access to all the information they need to take action while saving time and resources as they prepare for the organization-wide migration to Windows 11."

Seems like a solid and trustworthy place to get research data from. They got their "data" by scanning a network of their users and collected information about devices connected and the OS that is running. Sounds like they may primarily work with corporate users, which are traditionally slow at deploying new OSs. That's why we need to see the actual research data from an independent firm. Or hell, how about a link to the actual research and data from Lansweeper?

Windows 11 on 0.21% of PCs, says survey of 10 million

Lansweeper stats make grim reading for Microsoft

www.theregister.com

"Lansweeper's research comes from 10 million Windows devices, presumably in the business world, of which the IT asset management platform reckoned less than 45 per cent could take the update to Windows 11, thanks to Microsoft's stringent hardware requirements."