javispedro1 wrote on Today, 13:14:

Well, specially on this forum, you should probably mention Address Windowing Extension, basically like DOS' EMS/XMS for Win32 :P […]
Show full quote

spiroyster wrote on 2026-09-14, 18:57:

Windows XP 32-bit limits processes to 2GB unless built with laa (Large Address Aware, which is a compiler flag), in which (as an app) you can access 3GB, no more. If you want/need more within a single process, you have to use file memory mapping, not RAM (which is kinda application version of using a swap file). Swapping, RAM mapping (address translation) etc is all handled by the OS since we have something called protected memory, and have done for a number of decades now.

Well, specially on this forum, you should probably mention Address Windowing Extension, basically like DOS' EMS/XMS for Win32 😜 . it does allow allocating and addressing more than 2^32 from a single process, and it is neither file mapping nor swapping (if anything it is the opposite, as AWE-allocated memory cannot be swapped out!).

The Solutor wrote on Yesterday, 21:11:

Those are mostly retrofit technical explanations to justify a nasty commercial practice. Wrong in a number of ways, purposely wrong given Chen is anything but an incompetent.

I do agree, capping client 32-bit Windows at <4GB is purely arbitrary. Exactly the same story as removing 16-bit support from 64-bit windows, or all these CPUs from Windows 11. It is all arbitrary, any technical reasons you can point are mostly irrelevant when such limits can be lifted trivially even by third parties.

E.g. if the argument is 3rd party driver compatibility, then watch how they did it recently with VBS, or how did they it did with PAE/ the NX bit itself in the same time period!
They offer the toggle to users. They have a denylist/allowlist of drivers that are known to be incompatible/compatible.
They could have simply done the same to allow >4GB on client 32-bit Windows.
So it is just a way to push users to the direction they want. Whether this direction is a good thing or not, your mileage may vary.
Now all irrelevant with the death of 32-bit OSes anyway...

Dont really want to get into the core of your argument per say but the removal of the 16bit subsystem from the OS was a hardware requirement from AMD and Intel not something MS decided they wanted to do.

4Gb might look like it was an arbitrary cap but when you know what the problem was 4Gb was the obvious place.

0x00000000 = 0 bytes
0xFFFFFFFF = 4,294,967,295 bytes

As long as physical memory never exceeds 0xFFFFFFFF, every single physical memory address in the computer can fit cleanly into a standard 32bit CPU register (EAX, EBX, etc.) and into a single 32bit hardware register on an expansion card.

An analogy of what happened would be.
Your dogs are taking a dump in your back yard so you clean it up.
Every dog in the neighbourhood is dumping in your back yard so you errect a security fence and change the yard to a garden complete with flower pots and a pagoda.

Now they didnt get the flower pots and pagoda but its pretty clear their design called for one... all you got was a clean yard.

Or

By clamping the maximum physical address to exactly 4Gb 2^32 in the client kernel HAL PAE page tables remained active
The CPU could still use the 64bit page table entries required by the hardware NX bit to enforce DEP against network worms. Every physical address remained below 0xFFFFFFFF
The Windows memory manager guaranteed that it would never hand a device driver a physical DMA buffer with a high bit set so lazy driver code never triggered bugs
A buggy 32bit driver that cast physical address pointers down to standard 32bit integers would never truncate data, because the upper 32bits were always guaranteed to be zero.

The idea of a toggle to users...
Do you remember the tyrade of complaints about UAC in Vista?
Its common now of course people expect to be asked to elevate a process to admin but back then it was the wild west baby!

You are talking about an era in computing I can only equate to the American wild west.
Now it you want to listen to music there is Youtube, Spotify, Amazon, Music CDs and all those old music formats.
Back then the default answer to wanting to listen to the latest album was to load up a Peer2Peer client and steal it!

Its quite odd looking at what happened over the years. We have go from people downloading for free, illegally but free, to people willingly paying for their music.
There are edge cases of people who do still steal it, but they are edge cases now, the vast majority of people use amazon, spotify etc etc.
The younger guys I work with now, they are much much younger than I am, they are all tech savvy etc but they wouldnt know where to start looking to get free music downloads that works every bit as well on their phone.
I remember using IRC to DCC music of other users. then came the Napster and later the emule, kazza P2P networks... later torrent sites like suprnova and then piratebay took that crown. But free music was there. Thats why I say early 2000s computing was the wild west.

The 4Gb limit wasnt just a good place to put the cap, it was out n out required there to stop the problem the driver devs caused by writing drivers that were fundamentally broken. And XP fixed them by not allowing them to do what they want to do with its 4Gb hard cap.

ElectroSoldier wrote on Today, 13:52:

Nobody is talking about the inbound SMB file sharing

I'm talking about that. It's product policy exactly like the artificial memory limitin XPSP2 and later, exactly as aero disabled on win7 starter, and so on...

Then if you don't understand the analogies, it's your problem not mine.

You said many chipsets that support 4Gb or more.

Yes, in 2004, the year the limit was introduced by SP2

lets start in the NT4 era...

I said that NT had already solved the problem of managing more than 3GB.

You realize that NT4 was a thing EIGHT YEARS before XP SP2, normal people ran it on 64MB machines not 8GB ones.

Is that hard to understand?

Nvidia driver 333.11 was released in May 2014
A month after Windows XP reached End of Life, and ten full years after XP SP2 released in 2004.

What has to do XP here? Im talking about W7/Win8 drivers.

Can you stop to post random facts just to mud the waters, please?

With that one example you have to land on a hyper specific, curated, enterprise derived driver branch from 2014 to stop the machine from falling over.

Another string of invented things

All driver before 333.11 Work W/O any problem no matter if consumer or enterprise, no matter if we are takling of Geforces or Quadros or NVS, desktop and mobile.

Period.

Its exactly like forcing WinXP onto a modern motherboard with an unsupported ACPI table (anything beyond ACPI 2.0). You can slipstream patches, bypass the 0x000000A5 stop error, and trick the installer into declaring the machine an "ACPI Multiprocessor PC." But the NT5.1 kernels ACPI driver still doesnt have the hooks or logic to properly parse power states, sleep vectors, or modern interrupt routing defined a decade later. Device Manager might look clean, and the system might boot, but you havent actually solved the architectural incompatibility youve just masked it with a hack. And a dirty hack at that.

Is plenty of fixed acpi.sys drivers for xp 2003, vista, that fixes any problem you may have, coders and hakes have spent dozen of hours to give to the community the opportunity of using the OS they like on something that MS don't want to be possible, just to face people like you capable only of criticizing their work, just because.

Really nice behavior from you part, you deserve to pay even for the air you breath.

Law212 wrote on Today, 14:26:

I think you have that backwards. If one company is paying 1000 euro and another is only paying 300 , people will flock to the company paying more. I thin you mean that a company wont pay worked 1000 euro when there are workers who will take the job for only 300.

Yeah, sure lets wait for a billion and half of Chinese or Indian or Africans move all together to France, Italy, Germany... just to make your point.

And even when few thousands of immigrants move in the western countries for real, like it's happening the effect is to move the salary and rights competition from an international threat to a national one, lowering the salary which in turn helps the return of populism and fascism. Something already happened in the US, Italy, Argentina and another long string of countries.

And it's about to happen in France, Germany and another long string of countries.

What can get wrong?

🤣. ok then. Back to sanity. Thanks for the laugh though.

ElectroSoldier wrote on Today, 13:52:

Nobody is talking about the inbound SMB file sharing

I'm talking about that. It's product policy exactly like the artificial memory limitin XPSP2 and later, exactly as aero disabled on win7 starter, and so on...

Then if you don't understand the analogies, it's your problem not mine.

You said many chipsets that support 4Gb or more.

Yes, in 2004, the year the limit was introduced by SP2

lets start in the NT4 era...

I said that NT had already solved the problem of managing more than 3GB, in 1996, not that it was common use on consumer machines.

You realize that NT4 was a thing EIGHT YEARS before XP SP2? Normal people ran it on 64MB machines not 8GB ones.

Is that hard to understand?

Nvidia driver 333.11 was released in May 2014
A month after Windows XP reached End of Life, and ten full years after XP SP2 released in 2004.

What has to do XP here? Im talking about W7/Win8 drivers.

Can you stop to post random facts just to mud the waters, please?

With that one example you have to land on a hyper specific, curated, enterprise derived driver branch from 2014 to stop the machine from falling over.

Another string of invented things

All driver before 333.11 Work W/O any problem no matter if consumer or enterprise, no matter if we are takling of Geforces or Quadros or NVS, desktop and mobile.

Period.

Its exactly like forcing WinXP onto a modern motherboard with an unsupported ACPI table (anything beyond ACPI 2.0). You can slipstream patches, bypass the 0x000000A5 stop error, and trick the installer into declaring the machine an "ACPI Multiprocessor PC." But the NT5.1 kernels ACPI driver still doesnt have the hooks or logic to properly parse power states, sleep vectors, or modern interrupt routing defined a decade later. Device Manager might look clean, and the system might boot, but you havent actually solved the architectural incompatibility youve just masked it with a hack. And a dirty hack at that.

Is plenty of fixed acpi.sys drivers for xp 2003, vista, that fixes any problem you may have, coders and hakes have spent dozen of hours to give to the community the opportunity of using the OS they like on something that MS don't want to be possible, just to face people like you capable only of criticizing their work, just because.

Really nice behavior from you part, you deserve to pay even for the air you breath.

Law212 wrote on Today, 15:01:

🤣. ok then. Back to sanity. Thanks for the laugh though.

I already wasted my time replying to your dumb message, with an intelligent one.

It will not happen again.

Live long and prosper in your alternate universe.

The Solutor wrote on Today, 14:04:

@electrosoldiers

Seriously learn to quote properly. This way is impossible to discuss with you.

I honestly dont know if there was much point in fixing it, as your follow up answer proves to me you value guile over competence, and for that reason alone I would terminate any conversation.
You dont know the hardware limits you are talking about, motherboards of the era simply could not address memory over 4Gb anyway.
I mean even if you found a stick of 2Gb RAM that could work in a 4slot board to get over 4Gb, and that was a big ask back then, the board would present 4Gb to the OS because the chipset cant map it.
Thats fundimental and destroys your idea right from the start. And its on that basis I dont think you use the real hardware because otherwise you would know this because its an actual physical limitation before we even get to Microsoft clamping down to 4Gb...
That alone is DEATH to your who idea and the ones that come from it.
I dont think you realise just how devastating that fact is to your whole argument.

The only time, back then, that I actually came up to the physical memory limit of a motherboard was on an Acorp 6A815EPD that had an i815 chipset which was limited to 512Mb.
Otherwise I remember owning and using an MSI Neo2 FIS2R with the Canterwood chipset. I had a P4 3.06HT SL6S5 and 2Gb RAM in it. Not realising that was close to its limit.

It makes what you say next a dumpster fire!

The Solutor wrote on Today, 14:48:

I'm talking about that. It's product policy exactly like the artificial memory limitin XPSP2 and later, exactly as aero disable […]
Show full quote

ElectroSoldier wrote on Today, 13:52:

Nobody is talking about the inbound SMB file sharing

I'm talking about that. It's product policy exactly like the artificial memory limitin XPSP2 and later, exactly as aero disabled on win7 starter, and so on...

Then if you don't understand the analogies, it's your problem not mine.

You said many chipsets that support 4Gb or more.

Yes, in 2004, the year the limit was introduced by SP2

lets start in the NT4 era...

I said that NT had already solved the problem of managing more than 3GB.

You realize that NT4 was a thing EIGHT YEARS before XP SP2, normal people ran it on 64MB machines not 8GB ones.

Is that hard to understand?

Nvidia driver 333.11 was released in May 2014
A month after Windows XP reached End of Life, and ten full years after XP SP2 released in 2004.

What has to do XP here? Im talking about W7/Win8 drivers.

Can you stop to post random facts just to mud the waters, please?

With that one example you have to land on a hyper specific, curated, enterprise derived driver branch from 2014 to stop the machine from falling over.

Another string of invented things

All driver before 333.11 Work W/O any problem no matter if consumer or enterprise, no matter if we are takling of Geforces or Quadros or NVS, desktop and mobile.

Period.

Its exactly like forcing WinXP onto a modern motherboard with an unsupported ACPI table (anything beyond ACPI 2.0). You can slipstream patches, bypass the 0x000000A5 stop error, and trick the installer into declaring the machine an "ACPI Multiprocessor PC." But the NT5.1 kernels ACPI driver still doesnt have the hooks or logic to properly parse power states, sleep vectors, or modern interrupt routing defined a decade later. Device Manager might look clean, and the system might boot, but you havent actually solved the architectural incompatibility youve just masked it with a hack. And a dirty hack at that.

Is plenty of fixed acpi.sys drivers for xp 2003, vista, that fixes any problem you may have, coders and hakes have spent dozen of hours to give to the community the opportunity of using the OS they like on something that MS don't want to be possible, just to face people like you capable only of criticizing their work, just because.

Really nice behavior from you part, you deserve to pay even for the air you breath.

The first part we have already killed. But it does prove you dont know why Microsoft limited the OS to 4Gb.
Baring in mind no motherboard could present more than 4Gb to the OS anyway.

It shows you dont know what the drivers were doing.
It shows you have never actually tried it on the hardware.

I would stay away from NT4 as part of the idea because it just doesnt fit with the 4Gb hard cap in XP. Using it makes no sense at all. None of those boards that were current with NT could use over 4Gb of RAM so it was a none issue.

Nvidia 333.11 was something you brought up TWICE! as an example of after this its ok once nvidia removed PAE from it...
I mean wow
Thats epic as an idea that before 333.11 they work. When people who use them in the real world report otherwise, when my own experience proves you wrong.
I mean you are more than welcome to show me if working if you want to. There are working patches now that enable more than 4Gb of RAM to be addressed by XP.

But then you hit the fact the motherboards from the time cant address that much RAM.
You are litterly locked into a dead end and the only solution is a car crash ending to your ideas.

What else was there to unpack.

SMB limits were part of the products billing structure.
4Gb memory limits tells me you dont know what the problem was in the first place.
Aero on Windows 7 Starter...
that was removing features that ONE were not paid for, TWO most hardware it was installed on didnt support it.

Mixing those things up into a coherant arguement is quite a juggle because theyre so very different.

Going by what you are saying I can gauge your actual level of knowledge on the subjects and i feel there is no longer any point in continuing the conversation about it.
And Ill leave it at that.

ElectroSoldier wrote on Today, 15:33:

I honestly dont know if there was much point in fixing it,

If you are to be dishonest even about this, we don't have anything more to say each other.

OBVIOUSLY there isn't anything to fix AFTER you fixed it, even this is too complicate to understand for your naive mind?

You dont know the hardware limits you are talking about, motherboards of the era simply could not address memory over 4Gb anyway.

Which era? You keep mudding waters, my Acer notebook next to me, a cheap Turion unit manages 4GB pretty well, and it was released in 2006 or so, just one year after the release of XP 64, Just two years after the XP SP2 release.

Intel Supports 4GB in CONSUMER chipsets sice 845P (Springdale) released in 2003, supports 8GB on consumer chipset since 955x released in 2005, while support 8GB in server chipsets as old as the 450GX released in 1995 (a whole year before the NT4 release).

As usual is pretty easy to debunk your claims.

Just as a charity I avoid to quote all remaining BS. You keep repeating your funny claims, you keep replying to something I never said, you keep acting like a broken record.

And I already wasted too much of my competence, to much of my intelligence, too much of my time.

Have a nice day.

You can't drop it can you.
What about the chipsets that were in use at the time XPSP2 was released and before?
You know the ones that people were using at the time...
What did they support?

Law212 wrote on Today, 14:26:

Which happens in places like canada when you have an influx of foreign workers willing to take a lot less and so companies fire everyone and hire lower wage workers.

This has been causing controversy literally for centuries. See Bill the Butcher's complaint about the Irish. Friedrich Engels made similar observations about the Irish in Britain driving down wages and worsening the condition of the native working class.

And I wouldn't dismiss Marx out of hand. His analysis of capitalism remains insightful. The problem is the intellectual framework he tries to impose on history.

Errius wrote on 51 minutes ago:

Law212 wrote on Today, 14:26:

Which happens in places like canada when you have an influx of foreign workers willing to take a lot less and so companies fire everyone and hire lower wage workers.

This has been causing controversy literally for centuries. See Bill the Butcher's complaint about the Irish. Friedrich Engels made similar observations about the Irish in Britain driving down wages and worsening the condition of the native working class.

And I wouldn't dismiss Marx out of hand. His analysis of capitalism remains insightful. The problem is the intellectual framework he tries to impose on history.

Every 80-100 years we tend to have a major mixing up of the world powers. This cycle of boom, bust and war was noted as far back as the Greeks.

The US was warned by a once and former reserve currency nation in 1946 of pretty much all the bad side effects we would see in the last 65 years or so.

Nothing new in the world as new nations take the lead we will see what shakes out.