News Older systems now won't be able to update to newer versions of Windows due to reliance on an arcane CPU instruction

[mitch074]

Thar goeth my trusty Q6600! O faithful comrade of many Doom[ed] battles!

And yes, while there are still a few Penryns, Conroes and Meroms around here, few have been powered on for years.

But I don't think that operating systems should even contain floating point code let alone any onther accelerator instructions, they should stick to isolating and preserving their register files on task switches: Linus learned very early that using fancy CPU instructions like jumping to task state segements can have horrendous consequences, as Jochen Liedtke pointed out in excrucinating detail.

And as a rule: you don't change the OS compiler settings in a minor release that upates automatically--it breaks things and that's not permitted under any circumstances.

And that is even more the case when all they're trying to do is push machine learning onto people, who really do not want Microsoft's variant on their personal computers, especially when it's very likely to be even less safe then printing, something that M$ doesn't manage to secure in 42+ years of doing operating systems.

Microsoft's desperate search for consumer value in billions of ML investments does not justify what they are doing: people depend on their personal computers to work as intended by their owners not to fulfill M$ co-plot pipedreams!

If they don't know how to manage operating systems responsibly, that task needs to be taken off their hands, just like doing browsers.

You do when you scrapped 70% of your QA team and are now using your "early adopters" customer base as alpha testers (i.e. a step further from developer build), Average Joe & Jane as beta testers and your main customers are Enterprise.
Supporting a fancy instruction is a Bad Idea when it's early; when it's been in use for 10 years with at least 2 differing hardware implementation, you're rather safe from side effects as those have already been massively documented, and usually already fixed/worked around/mitigated at the compiler's level.

 

[snemarch]

Neural net stuff probably has nothing to do with this – any sort of NN/ML feature they want to bake into the OS will probably depend on VNNI instructions, or GPU code.

As others have mentioned, POPCNT requirement is just because the "generate code for" compiler flag has been set to a slightly newer architecture. And the instruction has use outside NNs and the already-mentioned cases like memory allocation, garbage collection and sparse data structures – when algorithms are implemented with (either by hand or smart compilers) SIMD instructions, it's sorta common than an instruction sets a register to a bitmask indicating "something".

E.g. if you have a bunch of data and want the count of bytes of a specific value, or larger/smaller than some value, x86 SIMD lets you do that - with AVX256 you can process the data as chunks of 32 bytes, you add the POPCNT to your counter and proceed to the next chunk. There's a lot of SIMD instructions that involve masks like this, and compilers can autovectorize (with varying degrees of success) code that isn't written with SIMD intrinsics.

 

[AndrewJacksonZA]

an i7-6700k without issue.

Thanks, man. What edition of Win11 are you running, and do you get updates?

I have an i7-6700 non-K, plus I even bought one of the last TPM 2.0 modules for my motherboard that was available in my country. I just haven't felt the need to pull the trigger for my home system to Win11 yet.

 

[TJ Hooker]

Thanks, man. What edition of Win11 are you running, and do you get updates?

I have an i7-6700 non-K, plus I even bought one of the last TPM 2.0 modules for my motherboard that was available in my country. I just haven't felt the need to pull the trigger for my home system to Win11 yet.

Unfortunately I won't be much help here. I only ran Win11 (Pro edition) for a short period, just to try it out. I think it was shortly after Win11 came out, so I'm guessing it was version 21H2. I found it didn't really offer anything over Win10 so I decided to just go back to 'officially supported' Win10 (which still had several years until end-of-life, at the time). I had installed Win11 on a separate drive/partition, so it was trivial to switch back. I've since upgraded hardware.

As an aside, you shouldn't need a discrete TPM chip. Platform Trust Technology, Intel's embedded TPM implementation which provides TPM 2.0 functionality, is supported starting with 6th gen Skylake chips/100 series chipsets. Unless your motherboard FW is dumb and doesn't let you enable it.

 

[abufrejoval]

This has always been the case, AFAIK. I was able to install Win11, shortly after it came out, with an i7-6700k without issue.

What I find so annoying is that it runs just fine on nearly anything when it's in side a VM, as long as that VM fakes a TPM device.

I've got Haswell and Broadwell Xeons and Goldmont Atoms both of which will get refused when you try to run Windows 11 natively on them, but on top of VMware (or KVM), no problem, not even the slightest complaint. And with GPU pass-through even Windows gaming works just fine e.g. in Proxmox or oVirt.

Yet on a Skylake notebook I got to accept every version of Windos 11 via some registry hacks and installer switches, only the 23H2 release finally refused to update over the existing 23H1 release, citing "incompatible hardware".

It just shows there is no hardware reason at all, it's just all about collusion with Intel & Co. to make perfectly capable hardware absolete.

 

[evdjj3j]

I have Windows 11 on an i5-4300u in a Thinkpad. It works perfectly fine. Once an instruction that is present on ix-5xxx is necessary then it wont work.

It's still not supported, even if you got it working.

 

[TerryLaze]

It just shows there is no hardware reason at all, it's just all about collusion with Intel & Co. to make perfectly capable hardware absolete.

It's called trusted platform module, it's not about the hardware being incapable, it's about having a more secure system.

 

[bit_user]

It just shows there is no hardware reason at all, it's just all about collusion with Intel & Co. to make perfectly capable hardware absolete.

I think you've got it slightly wrong. It's about collusion with the content industry, who wants to make consumer machines & devices ever more leak-proof portals through which to view their DRM-protected streaming content and games.

 

[mitch074]

I have Windows 11 on an i5-4300u in a Thinkpad. It works perfectly fine. Once an instruction that is present on ix-5xxx is necessary then it wont work.

Considering AMD based Zen to compete with Haswell (and Zen+ is pretty much identical), you should not have trouble with it. What might happen though is that Microsoft could raise the bar up to Haswell level by the end of life of Win11 (by requiring, say, AVX support), start Win12 with the same compiler settings and THEN raise it again to e.g. Broadwell level - eventhough that's doubtful, since the only instructions added to it was RDSEED and others added to Skylake afterwards were soon deprecated.

 

[bit_user]

Microsoft could raise the bar up to Haswell level by the end of life of Win11 (by requiring, say, AVX support),

Westmere (basically, first gen Core i-series) is as far as they can go without breaking support for the 2021-era Jasper Lake SoCs.

 

[mitch074]

Westmere (basically, first gen Core i-series) is as far as they can go without breaking support for the 2021-era Jasper Lake SoCs.

I had forgotten about those. OTOH, I wouldn't be surprised if Intel and Microsoft dropped those like a hot potato for next-gen Windows.

 

[abufrejoval]

I had forgotten about those. OTOH, I wouldn't be surprised if Intel and Microsoft dropped those like a hot potato for next-gen Windows.

Don't be too quick about that: Jasper Lake only launched in 2021 and they are still being sold as new in Intel NUC11 Essential Kits.

And of course they have a TPM 2.0 so there really is no reason, except some intentional obsolescence code misidentifying them as ancient.

 

[AndrewJacksonZA]

Unfortunately I won't be much help here.

As an aside, you shouldn't need a discrete TPM chip. Platform Trust Technology, Intel's embedded TPM implementation which provides TPM 2.0 functionality, is supported starting with 6th gen Skylake chips/100 series chipsets. Unless your motherboard FW is dumb and doesn't let you enable it.

Thanks anyway. And yeah, I got an early B150M board and I did need to purchase the module separately.

 

[abufrejoval]

I think you've got it slightly wrong. It's about collusion with the content industry, who wants to make consumer machines & devices ever more leak-proof portals through which to view their DRM-protected streaming content and games.

TPM can't really be the issue, near all Skylakes have TPM 2.0 yet are refused as bare metal while they work perfectly fine as VM hosts.

Same with the Haswell and Broadwell Xeons, they have sometimes even both, integrated and discrete TPM 2.0, but Windows just disqualifies them based on the CPU generation.

It's touching that you're trying to give Microsoft any credit at all, but it seems a rather arbitrary decision made along the lines of platform age, which happend to intersect Skylake and Kaby Lake, which have near zero software visible technical gaps between them.

It was similar with Windows 7, I believe, which could not be installed on a Kaby Lake system ("too new") unless you put a hypervisor underneath. In that case Microsoft absolutely didn't want anyone continue to run Windows 7 on Kaby Lakes and together with Intel tried near everything to make that impossible e.g. with Windows 7 GPU drivers that refused to install or work on a Kaby Lake system.

While there was absolutely zero detectable difference between the Sky Lake and Kaby Lake iGPUs, even less than on the CPU side.

I believe I just had to disabled the iGPU to get around that, in any case I ran Windows 7 on Kaby Lake just to prove a similar point back then...

They quite simply assume the power to make any decision on the platform they consider "theirs" when they just happen to be a software vendor who may not actively discriminate. And they need to be put into their place, because without some pressure their caesar madness will only grow to rotten Apple dimensions.

If they choose not to support ancient hardware, fine, show a warning and refuse liability if things go south.

But these active acts of sabotage go quite beyond what any government should let them get away with.

I don't know that Steam uses or requires TPM to protect games. I sure hope they'll never start because I'd consider that a breach of contract when things stop working.

I have no idea when it comes to streaming content, YouTube doesn't seem to care and all the others I have no intent of ever using.

I very much agree that they are tring to gain ever more control about the user base, to squeeze more and more constant money streams from them while they diminish the service.

Apple started from a DRM platform and branched into PC territory. But anyone smart enough could just steer clear.

Windows started on Personal Computers beholden only to their owners. Microsoft's attempts to turn that platform into an Apple clone is an abuse of power which needs to be stopped.

 

[AndrewJacksonZA]

They quite simply assume the power to make any decision on the platform they consider "theirs" when they just happen to be a software vendor who may not actively discriminate.

Opinions of the company aside, as someone who creates and maintains code, one of the best parts is deleting code that caters to "legacy support."

 

[bit_user]

It's touching that you're trying to give Microsoft any credit at all, but it seems a rather arbitrary decision made along the lines of platform age, which happend to intersect Skylake and Kaby Lake, which have near zero software visible technical gaps between them.

Well, you touch on another explanation, which is that Microsoft probably has some policy for deprecating old hardware based on age. Since Windows is no longer the cash cow it used to be, I wouldn't be surprised if there are some bean-counters simply trying to reduce QA and support costs by dropping support for hardware of a certain age, without regard to how easy or cheap supporting it might actually be.

Windows started on Personal Computers beholden only to their owners. Microsoft's attempts to turn that platform into an Apple clone is an abuse of power which needs to be stopped.

I've almost stopped caring what MS does. My disenchantment with them goes back almost 25 years. If I didn't have to run Windows on my work PC, I might never touch it again.

 

[mitch074]

Well, you touch on another explanation, which is that Microsoft probably has some policy for deprecating old hardware based on age. Since Windows is no longer the cash cow it used to be, I wouldn't be surprised if there are some bean-counters simply trying to reduce QA and support costs by dropping support for hardware of a certain age, without regard to how easy or cheap supporting it might actually be.


I've almost stopped caring what MS does. My disenchantment with them goes back almost 25 years. If I didn't have to run Windows on my work PC, I might never touch it again.

They did - When Windows 10 came out, Microsoft scrapped 2/3 of its QA team. Problem is, QA became abysmal and every "new" release of Windows 10 came with headlines-worthy regressions and problems.
So, yes, Microsoft is deprecating support for older hardware to keep QA costs down without falling victim to bad press.
Note that I personally don't care, I run Linux on most of my systems at home and even my main workstation is running Linux.

 

[bit_user]

as someone who creates and maintains code, one of the best parts is deleting code that caters to "legacy support."

Eh, I'm a little bit the opposite. Unless legacy support adds undue complexity or limitations (and I try to design code & systems where it doesn't), it makes me a little sad, every time I see legacy support get torn out.

 

[abufrejoval]

They did - When Windows 10 came out, Microsoft scrapped 2/3 of its QA team. Problem is, QA became abysmal and every "new" release of Windows 10 came with headlines-worthy regressions and problems.
So, yes, Microsoft is deprecating support for older hardware to keep QA costs down without falling victim to bad press.
Note that I personally don't care, I run Linux on most of my systems at home and even my main workstation is running Linux.

I'd say there is a big difference between not supporting older hardware and sabotaging it.

Nobody demands Microsoft to support their Q6600 running Windows 11.

Quite a few people would appreciate it if Microsoft didn't sabotage Sandy Bridge, Ivy Bridge, Haswells, Broadwells or Skaylakes.

There is a significant difference between the two.

Just saying that Microsoft won't fix any issue that doesn't appear on the platforms that are within the supported range is enough. And whether or not Windows is running in a supported mode on a supported platform is a flag that Microsoft and set, display and use for ticket routing without any significant effort on their part.

If they have budget to force Edge and Co-Plot on consumers who really would rather do without, they can absorb whatever tiny effort that might generate.

I earn my money on Linux, too.

But when I'm done working, I tend to use Windows for fun. As does the rest of the family, who believe I work for them during after-hours.

 

[bit_user]

I'd say there is a big difference between not supporting older hardware and sabotaging it.

Nobody demands Microsoft to support their Q6600 running Windows 11.

Quite a few people would appreciate it if Microsoft didn't sabotage Sandy Bridge, Ivy Bridge, Haswells, Broadwells or Skaylakes.

There is a significant difference between the two.

You're being naive. If the OS runs, people will expect it to work and complain if it doesn't. Some of these folks will not install the OS, themselves, and therefore won't see any warnings or disclaimers. For MS, it's a liability to let the OS run on anything they don't test and support. I don't blame them for locking out older hardware, even if I think they made a poor decision where to draw the line.

 

[abufrejoval]

You're being naive.

Not the typical feedback I get...

If the OS runs, people will expect it to work and complain if it doesn't. Some of these folks will not install the OS, themselves, and therefore won't see any warnings or disclaimers. For MS, it's a liability to let the OS run on anything they don't test and support. I don't blame them for locking out older hardware, even if I think they made a poor decision where to draw the line.

People will complain, that's a tautology.

As to the liability angle: it's a disingenuous argument they invented around Windows 10 and you're falling into their propaganda trap.

Microsoft is trying to change their business model and evidently care very little about the collateral damage they do.

But when you're so "relevant", there is no free practical alternative, you can't just do whatever you want, even [especially] if you succeeded with just that before [for far too long].

That's what new EU regulation is trying to make IT giants understand or conform to even if they don't want to: if you abuse the power your market position gives you against our citizens best interests or our local mores, we will hurt you, until you change.

And we'll get better at this type of regulation, so don't try to be too smart [Apple]!

Sovereignty is determined by the ground you live and trade on, not by corporate or cloud ethos.

 

[Colif]

I'm confused why this article exists, Windows 11 never supported any CPU older that Intel 8xxx or Ryzen 2xxx.

Did you read last paragraph of article?

As previously mentioned, compatibility won't be an issue for pretty much everyone running Windows 11 already. This new requirement will only affect users who are running modified builds of Windows 11 on super old machines, such as ones featuring Pentiums or Core 2 processors. Those people will not be able to run Windows 11 24H2 unless they figure out a way to bypass the POPCNT requirement as well.

So yes, most of the people on it now won't have a problem. This is for a small number.

 

[evdjj3j]

Did you read last paragraph of article?


So yes, most of the people on it now won't have a problem. This is for a small number.

You're right I missed the last paragraph because the article was pointless to read.

 

[Colif]

The last paragraph could have been higher in the article I guess but I don't have any control over articles.