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]

It doesn't work like that, ZEN 3 has a multiple layers of cache and a cache controller, if any of that uses AMD IP then MS can't just use it even if it is easy for them to figure it out they don't have the legal right to do so.
Look at all of this, it's all over the place and each of that could have something under copyright making it impossible for MS to do anything without AMD.

Rocket lake just has one simple cache per core, there is nothing to do wrong there.

Except that all the relevant work about that was already done under Win10, so all the specifications must have been known.

 

[LikeLinus]

Debatable. If I knew I had a fix in the pipeline (MS does; it's worked its way into the beta and release preview channels), then I might launch anyway and just let the AMD fans stew for a bit; that being said, annoying "20-30% of the market", a percentage of the market that enjoys (and on Win10, continues to enjoy) a performance lead over the competition, seems like a bad PR move. The Win11 release window has been pretty odd all around. Win8 and Win10 were not handled in this manner. Win10 certainly wasn't.

I'm not saying the 20-30% aren't unimportant, but in the grand scheme of things, when you take into account that some of those are enthusiast, it's an even smaller part of the market. Not condoning the fact that this slipped through, but the only point I was making, from the beginning, is that Microsoft isn't the only one to blame here. Plus, take in to the fact that most people haven't upgraded yet. So the percentage shrinks and by the time people do install it, it's a non-issue.

It's also being over dramatized. It's a 2-3% loss in performance for the majority of programs. It's mostly only games (eGames AMD said, then removed from their site, for some unknown reason). Generally, anything under 5% performance increase and decrease isn't noticeable. Tom's Hardware testing of games, while not exhausted, the most they saw is 7.5%. Here too, you see the people affected dwindle because the loss variation is so fairly minor on what most people use day-to-day software wise.

You could say the same thing about the UEFI issue. That same market percentage is affected. But for some reason people want to place all the blame on MS and don't realized it's shared and two issues. AMD should haven't known this though their own testing as well. Again, they are fixing this quickly and it'll be forgotten about in a month.

Bad PR move? Sure, it's not the best news, but I haven't seen any major news outlets make a big deal about this. Do I think people are going to stop using Windows due to a small issue that lasted all of 3-4 weeks and didn't cause any actual damage to their working computers? No, sorry, I don't.

This isn't affecting huge corporations and the mass majority of users.

@lobz - AMD couldn't use their own free CPU to identify the MS scheduler flaw and their own UEFI driver issue. Wouldn't it have been in their best interest to inform consumers there are issues and to hold off updating or that a fix was coming? They had access to RTM well in advance.

 

[TheELF]

aren't unimportant,

Dude, come on.
Also a lot of this 20-30% is servers/workstations that don't run windows or console APUs that again don't run windows.

 

[Joe NYC]

I ignored him a long time ago. You are wasting your breath.

You mean since you made up a story about the list of Threadripper SKUs? And you realized you were caught making things up?

This is quite a classic excuse, when caught BSing:

If anyone besides you have asked, I would have provided. However, as you like to argue and post hundreds of forum posts, I will not be engaging in this discussion with you. Sorry.

And all I asked for was a link

 

[Joe NYC]

LOL. It's a failure? It's been out for a week and half and a free upgrade.

It could have been a success if it wasn't rushed out the door prematurely, and if it followed stated Microsoft philosophy - to run on most hardware.

There was no one in the market place asking Microsoft to release a premature unfinished abomination, besides Intel.

Plus, what agreement did MS make to a schedule? Link?

The agreement between Microsoft and Intel is implied from the way Microsoft is acting with Windows 11.

Very much unlike the update process of Windows 10, which seems like a rigorous process.

If you don't like MS or Windows, use Mac or Linux.

I care, enough to post about it, because I use Windows, and I hate seeing Microsoft shooting itself and Windows in the foot.

 

[DrMrLordX]

I'm not saying the 20-30% aren't unimportant, but in the grand scheme of things, when you take into account that some of those are enthusiast, it's an even smaller part of the market.

It's more of an issue for corporate buyers. OEMs are going to be pushing Win11 pretty hard compared to home users that have the option of staying on Win10 until 2025. For DiY home users, they have the option of not upgrading. New buyers who buy prebuilts simply won't have that option. And no, I'm not talking servers/workstations but instead the thousands of cheap AiOs/desktops that get rolled out for the office folks. AMD is trying to break into that market, albeit somewhat tepidly. Win11 having performance problems on their chips isn't going to help with that, even if it is only for a few weeks after the launch.

This isn't affecting huge corporations

Only to the extent that they may not have an upgrade scheduled for their office PCs. If they do, they're going to find a lot of Win11 systems on offer.

Also a lot of this 20-30% is servers/workstations that don't run windows or console APUs that again don't run windows.

I don't think his figure references server/workstation sales, and the consoles don't even run Windows so obviously he wasn't talking about them. If he's wrong he's wrong, but at least be honest about it.

 

[lobz]

It doesn't work like that, ZEN 3 has a multiple layers of cache and a cache controller, if any of that uses AMD IP then MS can't just use it even if it is easy for them to figure it out they don't have the legal right to do so.

Since a lot has already been done in Win 10 to understand Zen's CCD and cache structure, what you're trying to tell me here implies that a) MS really didn't care - but that's completely fine, or b) they notified AMD, wanted to solve it and AMD just jumped into DMCA copyright dancing, while singing 'Can't touch this'. You have to be joking.

 

[TheELF]

Very much unlike the update process of Windows 10, which seems like a rigorous process.

Yeah windows 10 only needed two years after zen came out to support them (fully) in the scheduler...

Hardware & Tech News - OC3D.net

OC3D is where you can find the latest PC Hardware and Gaming News & Reviews. Get updates on GPUs, Motherboards, CPUs, and more.

www.overclock3d.net

I don't think his figure references server/workstation sales, and the consoles don't even run Windows so obviously he wasn't talking about them. If he's wrong he's wrong, but at least be honest about it.

If you substract server and consoles it not even close to 30% anymore and very doubtful if it's still 20%.

It's ~22% overall CPUs which includes consoles APUs ,nobody outside AMD knows how much of a percentage that is, and ~10% server which mainly don't run windows.

AMD x86 CPU Market Share Soars, Hits 14-Year High

AMD has hit its highest market share in 14 years. The company is only a few percentage points away from its all-time x86 market share record.

www.extremetech.com

they notified AMD, wanted to solve it and AMD just jumped into DMCA copyright dancing, while singing 'Can't touch this'. You have to be joking.

Not like that, but it takes resources manpower and money for AMD to work on this together with MS, and it could easily be that AMD didn't have one or more of those to spend at the optimum time and had to push it back.

 

[igor_kavinski]

b) they notified AMD, wanted to solve it and AMD just jumped into DMCA copyright dancing, while singing 'Can't touch this'.

You, Sir, are funny!

 

[lobz]

Yeah windows 10 only needed two years after zen came out to support them (fully) in the scheduler...

Hardware & Tech News - OC3D.net

OC3D is where you can find the latest PC Hardware and Gaming News & Reviews. Get updates on GPUs, Motherboards, CPUs, and more.

www.overclock3d.net

If you substract server and consoles it not even close to 30% anymore and very doubtful if it's still 20%.

It's ~22% overall CPUs which includes consoles APUs ,nobody outside AMD knows how much of a percentage that is, and ~10% server which mainly don't run windows.

AMD x86 CPU Market Share Soars, Hits 14-Year High

AMD has hit its highest market share in 14 years. The company is only a few percentage points away from its all-time x86 market share record.

www.extremetech.com

Not like that, but it takes resources manpower and money for AMD to work on this together with MS, and it could easily be that AMD didn't have one or more of those to spend at the optimum time and had to push it back.

It has already been done. Can you enlighten me?

 

[VirtualLarry]

Most of the Windows 11 is Windows 10 (and all the way back to Windows XP etc).

Pretty simple way to deal with the few features that don't work.

Hold back the new feature or hold back the release date of the of the OS.

Microsoft set itself up for failure by agreeing to an impossible schedule, and predictably Windows 11 is seen as a fail. It could have been a success if Microsoft gave it a more customary development + alpha + beta times.

It's just Win10 21H2, with a new skin and rounded corners, and a new setup routine designed to force usage of a Microsoft ID. That's mostly all that it is.

Edit: OK, and changes to the scheduler and kernel for the upcoming Intel 'Alder Lake' CPU.

 

[TheELF]

It has already been done. Can you enlighten me?

This is what I was replying to, the rigorous updates of windows 10 took 2 years for a fix while the crappy win11 updates have a fix ready a couple of weeks after release.
With all of this I'm only saying that windows 10 wasn't all that great either.
"Very much unlike the update process of Windows 10, which seems like a rigorous process. "

 

[DrMrLordX]

Yeah windows 10 only needed two years after zen came out to support them (fully) in the scheduler...

I don't remember Zen or Zen+ systems gaining significant performance in May 2019 (just before Zen2 came out). Mine certainly didn't. Granted I ran static OCed the whole time on that machine so maybe that's why it didn't matter for me.

If you substract server and consoles it not even close to 30% anymore and very doubtful if it's still 20%.

How do you figure? According to the linked article, AMD only had 9.5% server CPU market share in Q2 2021. If AMD's overall market share was ~22%, then removing server from the picture would bolster AMD's market share, not reduce it.

It's ~22% overall CPUs which includes consoles APUs

That's a little ridiculous since Intel doesn't even sell console CPUs, but they might be including those in the tally.

 

[lobz]

This is what I was replying to, the rigorous updates of windows 10 took 2 years for a fix while the crappy win11 updates have a fix ready a couple of weeks after release.
With all of this I'm only saying that windows 10 wasn't all that great either.
"Very much unlike the update process of Windows 10, which seems like a rigorous process. "

We're just not talking about the same thing I'm sure. Do you want to imply that Win 11 was built from scratch, and all the rigorous updates that were made for Win 10 are useless for Win 11, and all those things have to be rigorously implemented once again, BUT they have found the perfect balance in development resources to arbitrarily implement a big LITTLE scheduling paradigm for products that may or may not be launched in the near future and successfully continued on the desktop? Please don't come with market share again, because AMD CPUs will have orders of magnitude higher market share than Ader Lake CPUs for the foreseeable future.

Edit: please don't take the tone personally, I just wanna convey somehow, how unfeasible it sounds to me, that this is not 100% conscious negligence towards any vendor that isn't Intel. No conspiracies, no evil overlords... but also pretty much zero fracks given.

 

[TheELF]

How do you figure? According to the linked article, AMD only had 9.5% server CPU market share in Q2 2021. If AMD's overall market share was ~22%, then removing server from the picture would bolster AMD's market share, not reduce it.



That's a little ridiculous since Intel doesn't even sell console CPUs, but they might be including those in the tally.

We are not talking about market share of CPUs but about how much of those run windows and thus are a part of MS market share and being dissed by the performance fix not coming out.
You have to subtract server CPUs and all the APUs used for consoles, and that final number is the percentage of MS/windows customers that are getting dissed.

We're just not talking about the same thing I'm sure. Do you want to imply that Win 11 was built from scratch, and all the rigorous updates that were made for Win 10 are useless for Win 11, and all those things have to be rigorously implemented once again, BUT they have found the perfect balance in development resources to arbitrarily implement a big LITTLE scheduling paradigm for products that may or may not be launched in the near future and successfully continued on the desktop? Please don't come with market share again, because AMD CPUs will have orders of magnitude higher market share than Ader Lake CPUs for the foreseeable future.

Edit: please don't take the tone personally, I just wanna convey somehow, how unfeasible it sounds to me, that this is not 100% conscious negligence towards any vendor that isn't Intel. No conspiracies, no evil overlords... but also pretty much zero fracks given.

The thing is that intel made all the software for the scheduler to work fine in windows 11.
Intel finished their work back in august.
It has nothing to do with MS or any CPU market share, intel themselves made it work well before time.

How Intel Thread Director makes Alder Lake and Windows 11 a match made in heaven

Alder Lake is the new hybrid CPU architecture from Intel, and it features a special optimization for Windows 11 in the form of Thread Director.

www.digitaltrends.com

 

[lobz]

The thing is that intel made all the software for the scheduler to work fine in windows 11.
Intel finished their work back in august.
It has nothing to do with MS or any CPU market share, intel themselves made it work well before time.

How Intel Thread Director makes Alder Lake and Windows 11 a match made in heaven

Alder Lake is the new hybrid CPU architecture from Intel, and it features a special optimization for Windows 11 in the form of Thread Director.

www.digitaltrends.com

I like it when people make my point for me, thanks 🙂

 

[igor_kavinski]

The thing is that intel made all the software for the scheduler to work fine in windows 11.
Intel finished their work back in august.
It has nothing to do with MS or any CPU market share, intel themselves made it work well before time.

This raises the following possibility:

1) Microsoft did not inform AMD that the scheduler needed to be tested for performance reasons because the part of the scheduler code that deals with Intel architecture has no bearing on how scheduling is handled on AMD architecture.

2) AMD was aware that the scheduler is seeing some internal changes due to Alder Lake's P and E core arrangement but since those changes should be confined to the Intel architecture handling branch in the scheduler code, it didn't make sense for them to test it (how are they gonna test something that won't ever execute on their CPU? As in,

if(CPU==INTEL)
do scheduling_as_per_Intel_guidelines;
else if(CPU==AMD)
do scheduling_as_per_AMD_guidelines;

)

3) Unbeknownst to AMD, the part of the scheduler code that is CPU agnostic saw some change to make it work better with the Intel specific branch of the scheduler code. Microsoft programmer(s) committing the change did not foresee this change conflicting with the AMD specific branch of the scheduler code. All their internal tests ran fine and reported no problems BECAUSE the tests don't spit out performance metrics and even if they did, no one thought to compare those with the old version of the code or with those of Windows 10's scheduler performance metrics.

When the problem was discovered, it was too close to the Windows 11 launch date, so Microsoft decided to go ahead with the release anyway, thinking that people wouldn't make a big deal about a temporary problem.

That's my take on what transpired.

 

[lobz]

This raises the following possibility:

1) Microsoft did not inform AMD that the scheduler needed to be tested for performance reasons because the part of the scheduler code that deals with Intel architecture has no bearing on how scheduling is handled on AMD architecture.

2) AMD was aware that the scheduler is seeing some internal changes due to Alder Lake's P and E core arrangement but since those changes should be confined to the Intel architecture handling branch in the scheduler code, it didn't make sense for them to test it (how are they gonna test something that won't ever execute on their CPU? As in,

if(CPU==INTEL)
do scheduling_as_per_Intel_guidelines;
else if(CPU==AMD)
do scheduling_as_per_AMD_guidelines;

)

3) Unbeknownst to AMD, the part of the scheduler code that is CPU agnostic saw some change to make it work better with the Intel specific branch of the scheduler code. Microsoft programmer(s) committing the change did not foresee this change conflicting with the AMD specific branch of the scheduler code. All their internal tests ran fine and reported no problems BECAUSE the tests don't spit out performance metrics and even if they did, no one thought to compare those with the old version of the code or with those of Windows 10's scheduler performance metrics.

When the problem was discovered, it was too close to the Windows 11 launch date, so Microsoft decided to go ahead with the release anyway, thinking that people wouldn't make a big deal about a temporary problem.

That's my take on what transpired.

That's exactly my train of thought. Intel has done all the work needed, because they're swell and diligent, right? Ah give me a break... of course they did all this work, since they actually bring something completely different to market - not only was this in their best interest, but they obviously knew it beforehand.

On the other hand, Zen CPUs have been working exceptionally good for a long time now, so if AMD did nothing, I'm not entirely sure they were allowed to know that something had changed so drastically.

The funniest thing is this: the whole thing only makes MS seem blatantly incompetent, as always, and if anything, they've also made themselves seem like Intel had a firm grip over what, how and when they're releasing... A nice way to feed the idiotic conspiracy theorists.

 

[Hitman928]

It doesn't work like that, ZEN 3 has a multiple layers of cache and a cache controller, if any of that uses AMD IP then MS can't just use it even if it is easy for them to figure it out they don't have the legal right to do so.
Look at all of this, it's all over the place and each of that could have something under copyright making it impossible for MS to do anything without AMD.

Rocket lake just has one simple cache per core, there is nothing to do wrong there.

Are you suggesting that Rocket Lake has only 1 cache level and no cache controller?

 

[Zepp]

Today no one can force a "well-informed user" to do any Windows upgrade or update.

"in the past i have been using Windows 7 for six years+Windows Update Off.Windows 10, i have same practice WU Off and no problem. "

Yes! I've been using that little app on all my win10 devices for the last couple years. It's great

 

[LikeLinus]

That's exactly my train of thought. Intel has done all the work needed, because they're swell. Well of course, since they actually bring something different to market. Zen CPUs are working exceptionally good for many months now, so if AMD did nothing, I'm not entirely sure they were allowed to know that something changed so drastically. The funniest thing is this: the whole thing only makes MS seem blatantly incompetent, as always.

LOL, no, this makes AMD look blatantly incompetent as well. Manufactures had hands on the RTM for over a month. So you're telling me that AMD just sat around, with a new OS looming, and never tested their CPUs and noticed a discrepancy in performance? What about falling flat on their face with the UEFI issue? Sounds like a really idiotic move to just say "hey, it worked before, it'll work now!" without taking the time to invest benchmarking and testing. This isn't some obscure issue. AMD has no problems running games and benchmarking software to show the performance benefits of a new CPU or platform. But you're saying they don't take time to test against a new OS release?

You're saying they have to be "told" by other people that there are problems with their own product running on the latest OS? If you want to talk about being lazy and incompetent....

"I'm not entirely sure they were allowed to know that something changed so drastically." But the article The Elf posted is from August 9th! I mean, the press knew about almost two months before the launch. AMD didn't know?

It's all laughable how you're sticking to your guns and saying anything to not admit both companies screwed up. It was all on big bad MS.

 

[DrMrLordX]

You have to subtract server CPUs and all the APUs used for consoles, and that final number is the percentage of MS/windows customers that are getting dissed.

Again, AMD's server market share was only 9.5%, versus an overall marketshare of 22%. If you remove all server CPUs from that calculation, basic mathematics would show you that AMD's marketshare among the remaining markets (e.g. not server) would be higher than 22%. You can't just subtract 9.5% from 22% and get AMD's market share in everything else. That's not how it works!

I am unsure as to how (or even if) AMD's console APUs are included in x86 marketshare analysis since AMD is the only supplier of x86 console CPUs. Including those sales would only serve to muddy any report on the competitiveness of AMD vs Intel (or vice versa). AMD reports console APU sales as a part of their embedded/semi-custom in their earnings calls which really doesn't serve to clarify much. The only thing I can say with confidence is that, in Q2 2021, MS and sony combined shipped ~3.5 million units of their current-gen consoles using AMD's APUs. How AMD counts that against its x86 market share (if at all!) is, again, unclear, since AMD may have booked the earnings from those sales in a previous quarter.

Assuming AMD DID count those shipments against their x86 marketshare for Q2 2021, we can look at the total worldwide shipment of PCs (desktops, laptops, and workstations):

PC Demand Remained Strong in the Second Quarter Amid Early Signs That Market Conditions May Be Cooling, According to IDC

IDC examines consumer markets by devices, applications, networks, and services to provide complete solutions for succeeding in these expanding markets.

www.idc.com

(yes, feel free to remove some or all of the reported sales of Apple "PC"s since new sales are mostly ARM at this point)

Compared to total worldwide PC shipments, that 3.5 million units isn't really a lot of APUs. AMD having a 100% marketshare in the x86 console segment would proportionately not affect their total industry-wide x86 marketshare of 22% by very much. Removing the APU sales would push that number down a little bit, but probably not enough to move the needle downward any more than you would upward if you took out all the EPYC and Threadripper sales to boot (along with their semi-custom stuff which we don't necessarily hear about very often). In fact, the two would probably cancel each other out.

Intel, likewise, has hardware you haven't mentioned at all in the x86 realm that will never run Windows 11 Home or Pro: comm equipment. Lots of little Atoms like Tremont spend their time in comm gear. And God only knows how much of that they sell. So you pretty much have to take THAT out of Intel's corner when calculating the market share that AMD and Intel control of CPUs that could conceivably run Windows 11.

So, yeah, based on what was sold in Q2 2021, 20-30% of x86 sales in CPUs that are not: server, workstation, console, semi-custom, or embedded may well have belonged to AMD. @LikeLinus ' numbers are probably accurate, or at least accurate enough for the purposes of this conversation.

 

[Hitman928]

Again, AMD's server market share was only 9.5%, versus an overall marketshare of 22%. If you remove all server CPUs from that calculation, basic mathematics would show you that AMD's marketshare among the remaining markets (e.g. not server) would be higher than 22%. You can't just subtract 9.5% from 22% and get AMD's market share in everything else. That's not how it works!

I am unsure as to how (or even if) AMD's console APUs are included in x86 marketshare analysis since AMD is the only supplier of x86 console CPUs. Including those sales would only serve to muddy any report on the competitiveness of AMD vs Intel (or vice versa). AMD reports console APU sales as a part of their embedded/semi-custom in their earnings calls which really doesn't serve to clarify much. The only thing I can say with confidence is that, in Q2 2021, MS and sony combined shipped ~3.5 million units of their current-gen consoles using AMD's APUs. How AMD counts that against its x86 market share (if at all!) is, again, unclear, since AMD may have booked the earnings from those sales in a previous quarter.

Assuming AMD DID count those shipments against their x86 marketshare for Q2 2021, we can look at the total worldwide shipment of PCs (desktops, laptops, and workstations):

PC Demand Remained Strong in the Second Quarter Amid Early Signs That Market Conditions May Be Cooling, According to IDC

IDC examines consumer markets by devices, applications, networks, and services to provide complete solutions for succeeding in these expanding markets.

www.idc.com

(yes, feel free to remove some or all of the reported sales of Apple "PC"s since new sales are mostly ARM at this point)

Compared to total worldwide PC shipments, that 3.5 million units isn't really a lot of APUs. AMD having a 100% marketshare in the x86 console segment would proportionately not affect their total industry-wide x86 marketshare of 22% by very much. Removing the APU sales would push that number down a little bit, but probably not enough to move the needle downward any more than you would upward if you took out all the EPYC and Threadripper sales to boot (along with their semi-custom stuff which we don't necessarily hear about very often). In fact, the two would probably cancel each other out.

Intel, likewise, has hardware you haven't mentioned at all in the x86 realm that will never run Windows 11 Home or Pro: comm equipment. Lots of little Atoms like Tremont spend their time in comm gear. And God only knows how much of that they sell. So you pretty much have to take THAT out of Intel's corner when calculating the market share that AMD and Intel control of CPUs that could conceivably run Windows 11.

So, yeah, based on what was sold in Q2 2021, 20-30% of x86 sales in CPUs that are not: server, workstation, console, semi-custom, or embedded may well have belonged to AMD. @LikeLinus ' numbers are probably accurate, or at least accurate enough for the purposes of this conversation.

The x86 market share numbers may or may not include embedded and consoles depending on the source, but it's not hard to find a breakdown for exclusively desktop and laptop share. Keep in mind that in terms of volume, laptops are the clear majority of x86 sales.

AMD's Share of x86 CPU Market Hits 14-Year High

According to Mercury Research, AMD's total share of the x86 CPU market in Q2, at 22.5%, hasn't been this high since 2007. But the increase wasn't exactly driven by desktop CPUs.

www.pcmag.com

 

[igor_kavinski]

"I'm not entirely sure they were allowed to know that something changed so drastically." But the article The Elf posted is from August 9th! I mean, the press knew about almost two months before the launch. AMD didn't know?

Maybe the people who could actually do something, both on AMD and Microsoft's side, were on vacation? It's a possibility!

 

[TheELF]

Again, AMD's server market share was only 9.5%, versus an overall marketshare of 22%. If you remove all server CPUs from that calculation, basic mathematics would show you that AMD's marketshare among the remaining markets (e.g. not server) would be higher than 22%. You can't just subtract 9.5% from 22% and get AMD's market share in everything else. That's not how it works!

"You can't just subtract 9.5% from 22% and get AMD's market share in everything else."
If I did say that somewhere then I'm sorry, no, of course not, but that 22% total still includes server and APUs that are x86, it's not 9,5 of 22, it's 9,5 of whatever the percentage of server sales is compared to desktop.

PC Demand Remained Strong in the Second Quarter Amid Early Signs That Market Conditions May Be Cooling, According to IDC

IDC examines consumer markets by devices, applications, networks, and services to provide complete solutions for succeeding in these expanding markets.

www.idc.com

(yes, feel free to remove some or all of the reported sales of Apple "PC"s since new sales are mostly ARM at this point)

The question here would be if a PC centric count would even consider console CPUs, it looks like they only list PC OEM makers.