wierd_w wrote on 2026-09-18, 16:51:

They seemed to always *Say* a specific card was unsupported by a driver pack, that supported more expensive versions of that gpu core, and manually adding in the IDs almost always worked.

Not an exclusive Nvidia habit but, for sure, they were the king of the hill

XP's weird. Last time I messed with it, loading official NVIDIA AHCI driver for my motherboard's SATA controller would just BSOD 32-bit version. But x64 driver on XP x64 worked.

I could not use KeePassNatMsg plugin for KeePass on x86 XP, it threw exception about unsupported cipher. But worked on XP x64.

EAX didn't work on XP x64 with Audigy Rx (Daniel K's drivers) AND limiting RAM. At least not through DirectSound, OpenAL was fine.

I remember Explorer on XP x64 glitching once and taking most of the 4 GB of RAM.

2 words, holy mess!

Is it just me or are Creative drivers a dumpster fire that you deliberately start in your OS?

They were. Creative Labs' drivers for the AWE32 and AWE64 allowed games to dynamically load soundfonts to them, but almost always broke in spectacular ways, only worked with very specific configurations, and were more fragile than raku pottery.

Ive noticed this on forums now, XP got the blame for a lot of stuff it just didnt deserve.
UCyborg last post about XP's weird...
What he is talking about is a known driver problem to do with the nForce driver and NCQ on the SATA controller. It was fixed by using the generic microsoft driver, which im sure you know about on here, but it doesnt stop people saying its an XP thing when it really wasnt, XP was just there while it was happening.
Nvidia are some of the worst in the business for driver shenanigans. How else to you get, on the older cards, two basically identical cards to function completely differently!
The Quadro GeForce gap.
Famous on forums like this of course but all that messing about trying to control who gets access to the power of their card through drivers when they really needed to concentrate on making decent drivers that didnt break the OS they were running on to the point of instability. And then forcing Microsoft to remove the access to memory over 4Gb to stop the problem they were being blamed for.

ElectroSoldier wrote on 2026-09-20, 17:30:

Ive noticed this on forums now, XP got the blame for a lot of stuff it just didnt deserve.
UCyborg last post about XP's weird...
What he is talking about is a known driver problem to do with the nForce driver and NCQ on the SATA controller. It was fixed by using the generic microsoft driver, which im sure you know about on here, but it doesnt stop people saying its an XP thing when it really wasnt, XP was just there while it was happening.

I didn't know it was a known issue. I've come across the working Microsoft's driver. On modern Windows, there would be a delay with hybrid shutdown (aka. log off and hibernate) with Microsoft's driver.

I thought it was mostly just me ranting about things being buggy on XP, possibly stemming from trying to do "too modern" things on it on a computer with core components from late Vista era and a bit newer graphics card.

Are "too new" drivers for newer NVIDIA cards from first half 2010s generally still problematic on newer systems? At least with my limited experience, too new driver on XP seemed more noticeable in bug department, with breaking the screen resolution page in NVIDIA control panel. Or when I had the old GeForce4 MX 440, where the latest driver significantly degraded performance.

But then at some more recent point I noticed some glitches in maps in Half-Life (z-fighting), for which there are fixed maps, but don't remember if this was the problem with older drivers for my GTX 750 Ti too or does it only happen in newer ones...and it's been a good while since visiting XP. Though this could be design bug that happened to went unnoticed on old hardware. At least going by Mafia 1 and its z-fighting issue with blood and such, some developers did have problems in that department back then too. I remember there was a tweaker application with checkbox for specifically targeting that issue in Mafia, perhaps some of you remember it, it slipped out of my memory.

I mentioned some time ago in another thread here that I wasn't aware for the longest time that AMD Processor Driver existed for Windows XP too. It's bundled with Windows since Vista and is responsible for CPU frequency scaling on AMD CPUs, depending on the load. No apparent problems on Vista+, but loading appropriate driver on XP, it adds horrible lag spikes when frequency changes. At least that happened on my side.

In any case, I think it's fair to say getting things working properly on XP is on the more difficult side.

Teenager me had most fun on Windows 7, so there's that as well. 😉

UCyborg wrote on 2026-09-20, 20:57:

I didn't know it was a known issue. I've come across the working Microsoft's driver. On modern Windows, there would be a delay w […]
Show full quote

ElectroSoldier wrote on 2026-09-20, 17:30:

Ive noticed this on forums now, XP got the blame for a lot of stuff it just didnt deserve.
UCyborg last post about XP's weird...
What he is talking about is a known driver problem to do with the nForce driver and NCQ on the SATA controller. It was fixed by using the generic microsoft driver, which im sure you know about on here, but it doesnt stop people saying its an XP thing when it really wasnt, XP was just there while it was happening.

I didn't know it was a known issue. I've come across the working Microsoft's driver. On modern Windows, there would be a delay with hybrid shutdown (aka. log off and hibernate) with Microsoft's driver.

I thought it was mostly just me ranting about things being buggy on XP, possibly stemming from trying to do "too modern" things on it on a computer with core components from late Vista era and a bit newer graphics card.

Are "too new" drivers for newer NVIDIA cards from first half 2010s generally still problematic on newer systems? At least with my limited experience, too new driver on XP seemed more noticeable in bug department, with breaking the screen resolution page in NVIDIA control panel. Or when I had the old GeForce4 MX 440, where the latest driver significantly degraded performance.

But then at some more recent point I noticed some glitches in maps in Half-Life (z-fighting), for which there are fixed maps, but don't remember if this was the problem with older drivers for my GTX 750 Ti too or does it only happen in newer ones...and it's been a good while since visiting XP. Though this could be design bug that happened to went unnoticed on old hardware. At least going by Mafia 1 and its z-fighting issue with blood and such, some developers did have problems in that department back then too. I remember there was a tweaker application with checkbox for specifically targeting that issue in Mafia, perhaps some of you remember it, it slipped out of my memory.

I mentioned some time ago in another thread here that I wasn't aware for the longest time that AMD Processor Driver existed for Windows XP too. It's bundled with Windows since Vista and is responsible for CPU frequency scaling on AMD CPUs, depending on the load. No apparent problems on Vista+, but loading appropriate driver on XP, it adds horrible lag spikes when frequency changes. At least that happened on my side.

In any case, I think it's fair to say getting things working properly on XP is on the more difficult side.

Teenager me had most fun on Windows 7, so there's that as well. 😉

AMD processor driver existed due to the way AMD implemented the frequency changes, it was baked into the kernel in Vista but XP needed it to be added and it didnt work well because of that.
It was a change in technology that didnt work well because of the age of the OS at the time the change in hardware capability was released... You either lived with it or locked to CPU to not change frequency. There is of course more to it than what Ive just said but without going all technical its good enough.
All the tech changed and advanced so rapidly. It was hard for XP to keep up. Dont forget XP went final in 2001.

As to nvidia drivers its always best not to load the latest release drivers for those older cards especially when they are on the edge of the OS being dropped as they just didnt put in any work to iron on the edge case bugs. There was no money in it, nvidia always had its eye on big corp money paying their bills not gamers, so they dont write the best to the smaller share of their income. The gaming market cards are pocket change for nvidia today.
You only have to look at the stats on Steam to see most of the bought cards are in the mid range, I know looking at forums it looks like everybody is rockin the top of the line but their user stats say otherwise. So nvidia offer up little effort into their drivers and even limit them so we couldnt mod the cards to get quadro performance from geforce cards. But back then nvidia were trying to keep the Quadro and GeForce cards separate because they made more money from the GeForce cards but wanted to keep the high profits money they got from the Quadro cards too. So you get the driver shenanigans.
Jeff on his Craft Computing YT channel touched on this a few times without diving into what was really going on with the nvidia drivers.
He has mentioned a few times a card can do "X" but the drivers wont allow it.

As to the problems in the games you mention they were not WinXP problems in themselves, they manifested themselves in WinXP but they were not caused by XP itself... It was due to the way the newer cards changed, became "better" and thus we saw the problems created.

Yeah I see where you lived in Win7 and know that well you dont know all of what happened before to get to Win7...
I suppose it happens a lot on here, which is why I read what I read about WinXP. Which is a good reason to correct it now.

Why do Windows client editions on 32-bit x86 systems artificially limit RAM to 4 GB?

Typically, device drivers must be modified in a number of small ways. Although the actual code changes may be small, they can be difficult. This is because when not using PAE memory addressing, it is possible for a device driver to assume that physical addresses and 32-bit virtual address limits are identical. PAE memory makes this assumption untrue.

Many device drivers designed for these systems may not have been tested on system configurations with PAE enabled. In order to limit the impact to device driver compatibility, changes to the hardware abstraction layer (HAL) were made to Windows XP SP2 and Windows Server 2003 SP1 Standard Edition to limit physical address space to 4 GB.

This is straight from one of Microsoft engineers.

ElectroSoldier wrote on 2026-09-20, 21:48:

AMD processor driver existed due to the way AMD implemented the frequency changes, it was baked into the kernel in Vista but XP needed it to be added and it didnt work well because of that.
It was a change in technology that didnt work well because of the age of the OS at the time the change in hardware capability was released... You either lived with it or locked to CPU to not change frequency. There is of course more to it than what Ive just said but without going all technical its good enough.
All the tech changed and advanced so rapidly. It was hard for XP to keep up. Dont forget XP went final in 2001.

Indeed, 2001 was a quite long ago. And there's this thing with commercial software where bigger things end up only in newer major version. OK, not exclusively in commercial software, but may appear like a bigger deal for some people that they have to pay for newer version.

ElectroSoldier wrote on 2026-09-20, 21:48:

As to the problems in the games you mention they were not WinXP problems in themselves, they manifested themselves in WinXP but they were not caused by XP itself... It was due to the way the newer cards changed, became "better" and thus we saw the problems created.

I know, I didn't try to imply that they were. If it came out like that, wasn't intentional.

I do happen to be aware good number of problems occur because game doesn't do something most correctly, but due to environmental factors happened to work on old OS / drivers. I did some patching of old games myself and have been aware of the phenomenon of layman blaming Windows for games breaking for a long time now.

ElectroSoldier wrote on 2026-09-20, 21:48:

Yeah I see where you lived in Win7 and know that well you dont know all of what happened before to get to Win7...
I suppose it happens a lot on here, which is why I read what I read about WinXP. Which is a good reason to correct it now.

I was crazy about XP in its heyday when I was younger and dumber. Though that computer was rather limiting in the gaming department. It was still the time of quick changes and investments into those computers weren't big. Win95 computer we had at first was mostly for 2D educational games (with some NBA Live '97 and Interstate '76 in the mix), the newer XP computer was OKish for late 90s and up-to about 2005 games (low resolution, sometimes long loading times for the later ones).

Then a jump was made straight past all early first graphics cards supporting shaders - from NVIDIA GeForce4 MX 440 (D3D 7) right to ATI Radeon HD 4890 (D3D 10.1) and of course, Windows 7 was the future then.

Joseph_Joestar wrote on 2026-09-20, 22:04:

Why do Windows client editions on 32-bit x86 systems artificially limit RAM to 4 GB?

Typically, device drivers must be modified in a number of small ways. Although the actual code changes may be small, they can be difficult. This is because when not using PAE memory addressing, it is possible for a device driver to assume that physical addresses and 32-bit virtual address limits are identical. PAE memory makes this assumption untrue.

Many device drivers designed for these systems may not have been tested on system configurations with PAE enabled. In order to limit the impact to device driver compatibility, changes to the hardware abstraction layer (HAL) were made to Windows XP SP2 and Windows Server 2003 SP1 Standard Edition to limit physical address space to 4 GB.

This is straight from one of Microsoft engineers.

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.

First: you don't need to modify the drivers to make them working, the opposite happened actually since windows server supported more than 3.3 GB since NT4, let alone Win 2000, driver were well aware of the situation for almost a decade, THEN, after the release of the downgrading SPs, Nvidia and Intel started to downgrade their drivers as well.

Second: XP, 2003 RTM are basically the same OS, 2003 Standard edition is the same OS as 2003 Datacenter, the latter supports 64GB, no matter the SP level. They are ARTIFICIALLY differentiated using the kernel product policies, to make cheaper and more expensive versions, it's as usual matter of money not matter of engendering.

And back to the drivers, aside the video drivers situation that I already detailed, I haven't seen a single driver not working because the PAE in systems with 8GB or more even in the W10 days.

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 w […]
Show full quote

Joseph_Joestar wrote on 2026-09-20, 22:04:

Why do Windows client editions on 32-bit x86 systems artificially limit RAM to 4 GB?

Typically, device drivers must be modified in a number of small ways. Although the actual code changes may be small, they can be difficult. This is because when not using PAE memory addressing, it is possible for a device driver to assume that physical addresses and 32-bit virtual address limits are identical. PAE memory makes this assumption untrue.

Many device drivers designed for these systems may not have been tested on system configurations with PAE enabled. In order to limit the impact to device driver compatibility, changes to the hardware abstraction layer (HAL) were made to Windows XP SP2 and Windows Server 2003 SP1 Standard Edition to limit physical address space to 4 GB.

This is straight from one of Microsoft engineers.

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.

First: you don't need to modify the drivers to make them working, the opposite happened actually since windows server supported more than 3.3 GB since NT4, let alone Win 2000, driver were well aware of the situation for almost a decade, THEN, after the release of the downgrading SPs, Nvidia and Intel started to downgrade their drivers as well.

Second: XP, 2003 RTM are basically the same OS, 2003 Standard edition is the same OS as 2003 Datacenter, the latter supports 64GB, no matter the SP level. They are ARTIFICIALLY differentiated using the kernel product policies, to make cheaper and more expensive versions, it's as usual matter of money not matter of engendering.

And back to the drivers, aside the video drivers situation that I already detailed, I haven't seen a single driver not working because the PAE in systems with 8GB or more even in the W10 days.

Server platforms like the Compaq ProLiant and IBMs Netfinity with NT4 and Win2k server OSs ran specialised enterprise grade hardware and used certified drivers. Those sservers used enterprise grade SCSI HBAs and RAID controllers from companies like Adaptec, LSI and Mylex Intel server NICs whos drivers were spercifically written and certified to handle 36bit PAE and 64bit DMA. Consumer hardware in the era was a completely different world, the consumer PCI and AGP expansion cards, budget sound cards like the SB Live! cheap TV tuners and dial up modems used simple 32bit DMA controllers. If physically memory was mapped above 0xFFFFFFFF (4Gb) the drivers wither truncated high address bits, attempting to write to address 0x1_0005_0000 but actually wrote to 0x0005_0000 corrupting kernel or othr memory processes and or caused hardware exception blue screens.

Youre trying to conflate two entirely different worlds as being the same thing simply because they both use memory and drivers which just doesnt map together for anybody who knows anything at all about the first world. The enterprise grade hardware and supporting software drivers.

Then we come to your idea that after MS caps PAE with the release of SP2 and the subsequent deliberate dumbing down of drivers after that.
Well first of all there was another problem running at the time that was to do with viruses like Blaster and Sasser, MS saw this problem and so pushed DEP NX bit. SP2 introduced amoung other things enforced DEP to combat the epedemic ofbuffer overflow worms and that required PAE mode to be enabled.
The problem was they knew when they did that the drivers were written in such a way as to cause those buggy drivers to causes almost constant system crashes if a system had more than 4Gb of system RAM installed.
Raymond Chen of MS documented this at the time, and said they kept PAE active to keep DEP security because the viris problem but clamped the physical address celing to 4Gb in the kernel HAL. Which ensured no DMA buffer was allocated above that 4Gb boundry. Thus preventing the already broken drivers from dragging the system down with them.
This wasnt a MS problem, the drivers never worked in systems with more than 4Gb of RAM because the companies, think nvidia, creative etc (biggest names in the business) never write their drivers with the idea that a system could have more than 4Gb of RAM.
People these days seem to look back at those systems thinking 4Gb of RAM is a low bar but thats only because they dont seem to remember what systems really had. Most people had 512Mb and some high end systems had 1Gb. ESPECIALLY in the early 2000s, because while RAM was cheaper it wasnt cheap! And a lot of the motherboards couldnt address that much RAM anyway. So the drivers were crap by the standards of 4Gb+ systems which didnt matter until PAE was enabled for security reasons not some conspiracy theory crap of forcing companies to buy server versions and consumers to say in the home user space. or pay through the nose to "upgrade" to a server OS.

This isnt my opinion, you can get the drivers and see what theyre doing. They werent built properly from the outset. The problem isnt an abstract one its very spercific. MS provided the APIs but these companies (nvidia, creative, Realtek etc) didnt aportion the memory addresses properly, they thought they could use short cuts in the code and so they chopped off the high 32 bits. So when the OS handed the driver a memory buffer located at a physical address say 0x1_0500_000 which is just above the 4Gb mark the driver chopped off the 1 and handed the hardware 0x0500_0000
Which would overwrite, or try to, memory already allocated to another app or the kernel. iteslf which resulted in a lock up or BSOD.
It wasnt that they started to do anything wrong, that they turned down the quality of the coding for the drivers, in the days before addresses over 4Gb they were just fine. Once the hardware limit jumped over 4Gb they needed to do more work and they didnt.
But then knowing the people working on it at the time that doesnt surprise me at all. It was like telling an old timer at work how to do his job...
Its not going to end well, and there are good reasons for that.

The last part is rather amusing and made me smile I must admit.
WinXP and Win2k3 are basically the same thing...

Erm no thats wrong.
WinXP = NT5.1
Win2k3 = NT5.2
They share a common architectural heritage but NT5.2 introduced significant low level architectural rewrites. The is the complete overhaul of the memory manager and I/O subsystem and native 64bit I/O Reqest Packet handling double buffere bounce buffers in the HAL for devices that can not handle 64bit DMA over 4Gb. And significant improvements to the thread sceduler and the ACPI power management.
The reason Win2k3 server datacenter could handle up to 64GB or 32bit PAE wasnt just a licencing thing, the OS was engineered with strict drivers signing requirements for enterprise grade hardware that guarenteed bounce buffering wouldnt kill system integrity, which is something consumer WinXP never had.

Given the dramatic change in the OS from the one to the other I dont know why Win2k3 isnt more popular in the retro community space especially when resurecting older hardware to run old software.
A lot of the drivers will run on 2k3 but you dont get a lot of the drawbacks you will get with XP. As the drawbacks are baked into the drivers.

A while ago now I installed WinXP onto a HP ProDesk 600 G1 DM. I was surprised when it installed because it was a 4th gen i5 based system but got everything working, including the intel USB 3.0 ports.
But there was no Wifi, which I wanted, so off to ebay I went and found a USB Wifi adapter that said it was WinXP compatible. Which it was.
When you plug it in it installs a USB storage device and installs its own drivers.
Its a god aweful little thing. Its slow and its signal drops out for no reason at all. But its the size of a thumb drive and it basically works...
But its drivers are a crime scene! They are a total hack job!
And this is a driver that was written years after the problems were well known to exist.

ElectroSoldier wrote on Yesterday, 22:58:

Server platforms like the Compaq ProLiant and IBMs Netfinity with NT4 and Win2k server OSs ran specialised enterprise grade hardware and used certified drivers. Those sservers used enterprise grade SCSI HBAs and RAID controllers from companies like Adaptec, LSI and Mylex Intel server NICs whos drivers were spercifically written and certified to handle 36bit PAE and 64bit DMA. Consumer hardware in the era was a completely different world, the consumer PCI and AGP expansion cards, budget sound cards like the SB Live! cheap TV tuners and dial up modems used simple 32bit DMA controllers. If physically memory was mapped above 0xFFFFFFFF (4Gb) the drivers wither truncated high address bits, attempting to write to address 0x1_0005_0000 but actually wrote to 0x0005_0000 corrupting kernel or othr memory processes and or caused hardware exception blue screens.

No the world isn't made just by fortune 500 companies, not now, let alone 20 years ago.

Here is plenty of small companies ranging from 5 to 30/40 employes, and such companies, especially back in the days, hardly would have spent pointless amount of money to buy, the mambojambo enterprise grade HW you're taking about.

There are tons of servers built on "normal" hw, that work w/o a glitch and worked for decades

And no, most of the certified drivers aren't any different than their client counterpart. Except that they have .inf that artificially (again) blocks the client drivers to be installed on server, or viceversa.

Youre trying to conflate two entirely different worlds as being the same thing simply because they both use memory and drivers which just doesnt map together for anybody who knows anything at all about the first world. The enterprise grade hardware and supporting software drivers.

You are just parroting the FUD, spread by HW/SW sellers who make tons easy of money using that narrative.

They are doing their job after all, you aren't

Then we come to your idea that after MS caps PAE with the release of SP2 and the subsequent deliberate dumbing down of drivers after that.

That's not my narrative. That's a fact, no matter if you like it or not

Well first of all there was another problem running at the time that was to do with viruses like Blaster and Sasser, MS saw this problem and so pushed DEP NX bit. SP2 introduced amoung other things enforced DEP to combat the epedemic ofbuffer overflow worms and that required PAE mode to be enabled.

Yes that's was the official explanation at the time. Practically, the NX bit thing brought problems while fixing just a small niche of security problem, in exchange.

The main source of security troubles of the era was called Internet Explorer 4/5/6, Outlook [express] (and the related activex crap), MS would have spent better time fixing it/them, rather than using duck tape trying to seal holes elsewhere.

People smart enough to use Opera (and a bit of common sense), never touched a virus in their life.

nvidia, creative etc (biggest names in the business) never write their drivers with the idea that a system could have more than 4Gb of RAM.

Real world tells another story, but feel free to buy whatever story they told you.

Like I said, anything before 333.11 work w/o any problem no matter the amount of RAM you have installed.

ESPECIALLY in the early 2000s, because while RAM was cheaper it wasnt cheap! And a lot of the motherboards couldnt address that much RAM anyway.

That's true for low end or mobile chipsets, like the 945 or alike, but obviously if 4GB were supported by windows since the days of NT4 in 96' or so, in the early 2000s there where many chipsets that supported 4GB or more.

The last part is rather amusing and made me smile I must admit. WinXP and Win2k3 are basically the same thing... […]
Show full quote

The last part is rather amusing and made me smile I must admit.
WinXP and Win2k3 are basically the same thing...

Erm no thats wrong.
WinXP = NT5.1
Win2k3 = NT5.2

Obviusly XP64 has the same kernel as 2003 x64, just like XP 52 has the same kernel as 2003 x86

Just compare apples to apples!!!

Perhaps 5.1 v.s. 5.2 was almost a unique case in the MS history, a case repeated only decades later with Server 2022 where no official W10 counterpart exist (albeit is pretty easy to make one yourself)

That said a large part of what makes an OS is the userland, and userland was mostly the same between 5.1 and 5.2 with small bits updated here and there, mostly to cope with kernel's improvements.

I dont know why Win2k3 isnt more popular in the retro community space especially when resurecting older hardware to run old software.

Not sure, about the retro community, but practical people use them. I'm still use a virtualized 2003x86 as my Domain Controller at home, it does its job using less than 200MB, when even the leanest modern Core Server needs 600/700MB just to boot itself, w/o a single role installed.

My multi boot client installs include, XP52 & XP64 among everything else.

A while ago now I installed WinXP onto a HP ProDesk 600 G1 DM. I was surprised when it installed because it was a 4th gen i5 based system but got everything working, including the intel USB 3.0 ports.

That's pretty normal. XP is officially supported until Gen 3/Ivy Bridge, a single generation ahead requires really minimal skill to make everything work.

But there was no Wifi, which I wanted, so off to ebay I went and found a USB Wifi adapter that said it was WinXP compatible.

Just use a Realtek USB dongle, they are robust, well supported in XP, easy to get working in Linux (they work out of the box since kernel 6.1), they work very well in access point mode, just use the win 7/8 drivers in win 10 or win 11, not their native ones (MS artificially, again, asked to remove the hosted network part from Win10+ drivers) , and last but not least, they are dirty cheap. I used to pay 2.5$ for a 2.5/5Ghz AC flavour (8812BU) unit.

XP should have plenty of support for WIFI USB dongles, Win2k on the other hand is much harder.

Unknown_K wrote on Today, 00:27:

XP should have plenty of support for WIFI USB dongles, Win2k on the other hand is much harder.

Take your old unused Android phone, connect it to internet via WIFI, share the connection via USB, plug it in your PC, add the well known RNDIS inf to detect the connection.

Done. You have your WIFI dongle compatible with WIN2k (and everything else).

If your android is newer than 9 just set USB tethering as default, if it's older, use tasker to automate the actions needed.

The Solutor wrote on Today, 00:03:

No the world isn't made just by fortune 500 companies, not now, let alone 20 years ago. […]
Show full quote

ElectroSoldier wrote on Yesterday, 22:58:

Server platforms like the Compaq ProLiant and IBMs Netfinity with NT4 and Win2k server OSs ran specialised enterprise grade hardware and used certified drivers. Those sservers used enterprise grade SCSI HBAs and RAID controllers from companies like Adaptec, LSI and Mylex Intel server NICs whos drivers were spercifically written and certified to handle 36bit PAE and 64bit DMA. Consumer hardware in the era was a completely different world, the consumer PCI and AGP expansion cards, budget sound cards like the SB Live! cheap TV tuners and dial up modems used simple 32bit DMA controllers. If physically memory was mapped above 0xFFFFFFFF (4Gb) the drivers wither truncated high address bits, attempting to write to address 0x1_0005_0000 but actually wrote to 0x0005_0000 corrupting kernel or othr memory processes and or caused hardware exception blue screens.

No the world isn't made just by fortune 500 companies, not now, let alone 20 years ago.

Here is plenty of small companies ranging from 5 to 30/40 employes, and such companies, especially back in the days, hardly would have spent pointless amount of money to buy, the mambojambo enterprise grade HW you're taking about.

There are tons of servers built on "normal" hw, that work w/o a glitch and worked for decades

And no, most of the certified drivers aren't any different than their client counterpart. Except that they have .inf that artificially (again) blocks the client drivers to be installed on server, or viceversa.

Youre trying to conflate two entirely different worlds as being the same thing simply because they both use memory and drivers which just doesnt map together for anybody who knows anything at all about the first world. The enterprise grade hardware and supporting software drivers.

You are just parroting the FUD, spread by HW/SW sellers who make tons easy of money using that narrative.

They are doing their job after all, you aren't

Then we come to your idea that after MS caps PAE with the release of SP2 and the subsequent deliberate dumbing down of drivers after that.

That's not my narrative. That's a fact, no matter if you like it or not

Well first of all there was another problem running at the time that was to do with viruses like Blaster and Sasser, MS saw this problem and so pushed DEP NX bit. SP2 introduced amoung other things enforced DEP to combat the epedemic ofbuffer overflow worms and that required PAE mode to be enabled.

Yes that's was the official explanation at the time. Practically, the NX bit thing brought problems while fixing just a small niche of security problem, in exchange.

The main source of security troubles of the era was called Internet Explorer 4/5/6,, Outlook [express] (and the related activex crap), MS would have spent better time fixing it/them, rather than using duck tape trying to seal holes elsewhere.

People smart enough to use Opera (and a bit of common sense), never touched a virus in their life

nvidia, creative etc (biggest names in the business) never write their drivers with the idea that a system could have more than 4Gb of RAM.

Real world tells another story, but feel free to buy whatever story they told you.

ESPECIALLY in the early 2000s, because while RAM was cheaper it wasnt cheap! And a lot of the motherboards couldnt address that much RAM anyway.

That's true for low end or mobile chipsets, like the 945 or alike, but obviously if 4GB were supported by windows since the days of NT4 in 96' or so, in the early 2000s there where many chipsets that supported 4GB or more.

The last part is rather amusing and made me smile I must admit. WinXP and Win2k3 are basically the same thing... […]
Show full quote

The last part is rather amusing and made me smile I must admit.
WinXP and Win2k3 are basically the same thing...

Erm no thats wrong.
WinXP = NT5.1
Win2k3 = NT5.2

Obviusly XP64 has the same kernel as 2003 x64, just like XP 52 has the same kernel as 2003 x86

Just compare apples to apples.

Perhaps 5.1 v.s. 5.2 was almost a unique case in the MS history, a case repeated only decades later with Server 2022 where no official W10 counterpart exist (albeit is pretty easy to make one yourself)

I dont know why Win2k3 isnt more popular in the retro community space especially when resurecting older hardware to run old software.

Not sure, about the retro community, but practical people use them. I'm still use a virtualized 2003x86 as my Domain Controller at home, and XP52 & XP64 as client, it does its job using less than 200MB, when even the leanest modern Core Server needs 600/700MB just to boot itself, w/o a single role installed.

A while ago now I installed WinXP onto a HP ProDesk 600 G1 DM. I was surprised when it installed because it was a 4th gen i5 based system but got everything working, including the intel USB 3.0 ports.

That's pretty normal. XP is officially supported until Gen 3/Ivy Bridge, a single generation ahead requires really minimal skill to make everything work.

But there was no Wifi, which I wanted, so off to ebay I went and found a USB Wifi adapter that said it was WinXP compatible.

Just use a Realtek USB dongle, they are robust, well supported in XP, easy to get working in Linux (they work out of the box since kernel 6.1), they work very well in access point mode (just use the win 7/8 drivers in win 10 or win 11, not their native ones), and last but not least they are dirty cheap, I used to pay 2.5$ for 2.5/5Ghz AC flavour (8812BU).

A lot to unpack there.

I know the mom n pop businesses you are talking about. I serviced a lot of them. Here where I live it was common for such businesses to use WinXP as a server OS.
The users and the use cases doesnt change the fact it is consumer grade.
I seem to remember XP Home OEM cost about £100 while XP Pro was about £150 maybe £200.
Those businesses wanted something they didnt want to pay for.
It really does come down to that. Its not about rinsing as much money as possible out of the users, its about getting value for your product. The server OSs took more work, came with more in terms of features and support and so cost more and mom n pop businesses didnt want to pay for it so cheaped out and went with XP Pro because people like my boss at the time told them it would work rather than lose the business.
And then when they did have problems, which was inevitable, then they blamed Microsoft, which is what you are doing now.
You want something for nothing. And having spent a small amount of money you expect to get all you need.

Those servers built on normal hardware are exactly what they are. Repurposed home computers.
It makes you wonder why people would do that and expect results you would expect from a system that cost thousands upon thousands of pounds more than what they paid.

Doing it the cheap way does not mean its Microsofts fault.

It isnt FUD at all. You dont even understand what the problem is. Youre blinded by a conspiracy theory.

The idea that the resellers point these businesses towards the expensive hardware and software just to make money out of them is part of the conspiracy theory.
Skill set, knowledge and enacting it in the real world...
Those sellers know that the software isnt designed to do what you want to do with it.
They point you towards the correct product.
IF you later have problems for using home grade software to run a business then whos to blame for that.

The idea about drivers being dumbed down after the enforced implimentation of PAE because of NX bit...
Yeah it is your narrative because we can see the drivers before and after, and the facts dont fit your story! Not at all what so ever.
In order to make it even half way credible you have to leave out certain key facts such as the lack of hardware support for more than 4Gb of RAM, the fact that people just didnt have that much RAM, the fact that driver writers never thought over 4Gb of system RAM was going to be a thing, that fact they had never written drivers for a system like that before.
We can see the drivers NOW, and the facts dont fit your story at all.

NX Bit fixed more than "just a small niche"
That part alone makes me realise you have no idea what was going on at the time.
MS Blaster didnt affect a small niche, it was global and the user didnt even need to do anything, just run an unpatched system on the internet. The worm found you!

IE 4/5/6 was a problem, The fix was either opera if you wanted to pay for it or those in the know saw Pheonix>Firebird>Firefox.
Its only in recent years thats changed. But back then Opera was shareware that had a stupid big banner ad if you didnt pay for it.
There were work arounds for that, but then there were work arounds for paying for Win2k3 too, and if you are willing to crack the one then why not the other which makes the conversation fall flat on its face if you take that route.

The arguement you make then seems to depart from reality and goes into your fantasy land that viruses didnt have any meaningful effect on millions of computers.
If you are going to depart from the real world by that much then there isnt much of a conversation to be had here.

This part is just unreal.

nvidia, creative etc (biggest names in the business) never write their drivers with the idea that a system could have more than 4Gb of RAM.

Real world tells another story, but feel free to buy whatever story they told you.

ESPECIALLY in the early 2000s, because while RAM was cheaper it wasnt cheap! And a lot of the motherboards couldnt address that much RAM anyway.

That's true for low end or mobile chipsets, like the 945 or alike, but obviously if 4GB were supported by windows since the days of NT4 in 96' or so, in the early 2000s there where many chipsets that supported 4GB or more.

In the real world you can see the drivers... You clearly havent ever looked at them to see what the problem is.

This is short list, not exhaustive, just the ones I remember.
Intel 845, 848, 850 chipsets were limited to 2Gb some 3Gb of RAM and they couldnt remap over 4Gb
Intel 865PE/865G supported upto 4Gb but no remap logic means they have a max usable address space of 3.25Gb
Intel 875P has a limit of 4Gb
Intel 915P, 915G, 925X are capped at 4Gb as a limit and you lost .75 to 1Gb to MMIO.
Intel 945GM, 945PM were limited to 4Gb even on a 64bit OS they are limited to 4Gb as they cant map over that.
The first Intel chipset to remap over 4Gb was the 965 Express and 975X series.

VIA KT133, KT266< KT333< KT400, KT600, KT880 were limited to 2 or 3Gb
nForce2 SPP Iltra 400 were capped at 3Gb
SiS 735, 745, 748 were capped at 2 to 3Gb

Even when the move to the 64bit generation in the early days there were still hangers on and the VIA PT880 Pro, PT890, P4M890,P4M90 were limited to 2 or 4Gb due to remapping
SiS 671 and 672 had the same problem and there was a little known board that used the ATi Radeon Xpresss 200 that was capped to 2 to 4Gb with no capability to remap over that.

Considering your mom n pop businesses you mention earlier do any of THOSE chipsets sound familiar to you? Especially when matted to WinXP... And now you start to see why its important to write the drivers properly as there is a hardware limitation in the platform you are using. And the claims that they were tested properly and fully is a blatant lie at best! Because the boards cant do it!!!
4Gb of RAM wasnt a number envisaged as something that would be used outside server grade hardware at the time.
The Intel 815 is famous for having a 512Mb limit but you know what, it wasnt a problem because most systems never had over 256Mb anyway!

Youre confusing what an OS supports in terms of capability and what the hardware is physically capable of.
I said earlier in the thread it isnt a WinXP problem, its a platform problem and it was exserbated by poor driver writing.

The things youre saying tell me you simply dont understand what the problem was to begin with and then lay a load of conspiracy theory to explain it away.
The one company I would agree with you on is nvidia, because the graphics cards were basically the same over GeForce and Quadro, in the early days the physical diference was as simple as a zero ohm resister not placed, or placed and drivers written to seperate the two cards from each other so the capabilities were limited because while nvidia at the time took most of its income from home users segment of the market they didnt want to lose the high profit margin sector their quadro cards bought in. Even though they were basically the same cards.
But even then as clever as their drivers were they couldnt write the drivers properly!
The problem isnt the physical amount of RAM in a system, its the addressing.

Driver devs for budget PCI sound cards, 56k dialup modems and capture cards never audited their DMA routines for over 4Gb physical buffers because they were writing for an user base that averaged 256 to 512Mb of RAM. When the OS handed them a 64bit physical address buffer, sloppy drivers cast the pointer to a 32bit ULONG, truncated the high bits, and handed the hardware an address under 4 GB. The DMA controller then happily trashed kernel memory or another apps memory space causing lockups and BSODs.

YOur example of the use of 2k3 in answer to me just sums it up really.
Virtualisation in a modern context compared to using Win2k3 as a workstation on the physical hardware from the period...

Ive got no problem with calling these companies out for their profit margins. I remember thinking for years if MS lowered the prices I wouldnt have been doing what I was doing because I would be able to afford to buy it.
I can tell you about the nights spent.... I mean I will say PID 640 generations and reset 502, finding the PID for the retail MCE... the PIDs that trigger MCE or TPC installs... and leave it at that, because if you know you know.
So yeah lets call them out for that, the basic cost of an OS sure, but the conspiracy stuff mentioned now just doesnt match to reality.

Microsoft didnt lock XP SP2 into a 4Gb limit in the HAL to force mom n pop to buy a £3k OS, they did it because turning on PAE for DEP on millions of real world PCs would have triggered an unmitigated disaster of driver crashes on hardware that was never built for that kind of high memory addressing. That had to do the one, because the virus problem was amazing... Kazza, emule and Suprnova were kicking out viruses like there was no tomorrow "Free" movies and MP3s forever! huh, but that gateway allowed the spread of viruses that had no end because people woudlnt stop using those sites to download from.
And then there was MS Blast and for that all you had to do was be there, no link clicking, no downloading just log on and check out!
Its still out there and still effects unpatched XP Win2k PCs its just with the decline of XP Win2k infected PCs it takes longer than it used to. But I saw it, it didnt take 25 minutes... It was more like 5 back in '04.

The best part is what you said about XP-64 and Win2k3 64.
You prove my point without even realising it.
The problem with XP-64 was the lack of drivers for it because due to signing driver devs were forced to write drivers properly to get them WHQL approved, which lead to a lack of drivers. Surprise surprise!

Yeah it takes no skill to install XP onto a known working system. But then I also have it working on an i7 5775c with a Titan Xp and 16Gb RAM and it all works properly and it uses that RAM too. ACPI the lot! And I know theres few on here, at least the ones who talk on here can do that! Because otherwise they would have said in the thread when they had the chance to.

Your wall of nonsense doesn't make your points any better.

Just calling small and middle sized business as "Mom & Pop Businesses" alone tells a lot about how people like you think.

Those businesses wanted something they didnt want to pay for.

Yeah, because MS making tons money at the cost of duplicating a 30 cents CD/DVD, and later not even that, were the good guys, Right?

Doing it the cheap way does not mean its Microsofts fault.

Who said MS fault? Win Server did its dirty job, just like XP did, it's not about technical fault that we are talking about.

It's about, debatable, commercial practices, and about naive people like you that take them blindly as true.

You dont even understand what the problem is.

Yeah, I did 45 years of IT waiting for someone like you, that finally explained me how the world works. 😁

In order to make it even half way credible you have to leave out certain key facts such as the lack of hardware support for more than 4Gb of RAM, the fact that people just didnt have that much RAM, the fact that driver writers never thought over 4Gb of system RAM was going to be a thing, that fact they had never written drivers for a system like that before.
We can see the drivers NOW, and the facts dont fit your story at all.

Listen, I'm using 32bit OSes, on bare metal, RIGHT now, is that clear? And I do that since my machines started to be filled by more than 3GB of RAM, which means more than a decade, and I NEVER faced a single problem, once the correct GPU drivers were installed.

That part alone makes me realise you have no idea what was going on at the time.
MS Blaster didnt affect a small niche, it was global and the user didnt even need to do anything, just run an unpatched system on the internet. The worm found you!

Unpatched system are perfectly fine, in most scenarios, unless you connect them directly using a modem.

Once you are behind a router that NATs, most problems (if any) come from the person sitting in front of the monitor, no matter the patch level of the OS.

Use a decent browser, use a bit of common sense, stay behind a router/firewall, and no one harms you, unless the router itself has security holes.

IE 4/5/6 was a problem, The fix was either opera if you wanted to pay for it or those in the know saw Pheonix>Firebird>Firefox

FF was a joke compared to Opera, now that most Opera innovations were slowly photocopied by others, even FF is an option, still looks like a joke if compared to Vivaldi.

There were work arounds for that, but then there were work arounds for paying for Win2k3 too

Workarounds that were way more unethical towards a small, honest, European company than they were against the big, evil, American company that MS was.
Perhaps MS became a giant thanks the the so called piracy, a practice that MS tolerated, and even silently backed at the time (and right now), at least for individuals and small/medium sized companies. So stop with the BS about those evil users who made Bill Gates poor.

I just extracted some of your pearls.

Don't mind that I spend a night quoting and debunking all your nonsense line by line.

Being concise isn't exactly your best quality 😁

Want wifi adapters that work with EVERYTHING buy 802.11b Prism 2 chipset devices, Win98 = works, WinCE = works, Apple Newton = works, Amiga 1200 = works maybe the client sucks and won't let you use much security though.

Edit: BTW #1: Dlink 802.11g adapters are usually pretty broadly compatible. Though sometimes it's a problem figuring if you've got an alternate chipset version with same model number. Usually broadcom or atheros.

BTW #2: some "universal" adapters for 10base2 aka WAP devices, might claim they need a certain windows version. Same as some routers. That's only for the config software, set them up on a later box and they'll plug into any Ethernet NIC, even if you have to use a media adapter. Of course you then have to configure TCP/IP stack to IP/mask/DNS etc settings you set them up as. Wallplug routers are dirt cheap and will usually function as WAPs, just turn of WAN and put it on wifi and LAN and you should be good. You can "sorta" plug them into your computer if you use a "kettle cord" with passthrough and socket adapter and plug them in there on back of PSU with short patch cord to NIC.

Vivaldi is good in the same way as Redhat Linux 5.0 was good.

And it’s on the Google backend with all the bloat

The Solutor wrote on Today, 03:38:

Yeah, because MS making tons money at the cost of duplicating a 30 cents CD/DVD, and later not even that, were the good guys, Right?

Its not a case of a 30 cent CD, its all the work that went into making what was on it.
And yeah had they charged less they might have made more. None of that is in dispute. Its the only part of your argument that stands up.

The Solutor wrote on Today, 03:38:

Yeah, I did 45 years of IT waiting for someone like you, that finally explained me how the world works. 😁

In order to make it even half way credible you have to leave out certain key facts such as the lack of hardware support for more than 4Gb of RAM, the fact that people just didnt have that much RAM, the fact that driver writers never thought over 4Gb of system RAM was going to be a thing, that fact they had never written drivers for a system like that before.
We can see the drivers NOW, and the facts dont fit your story at all.

Listen, I'm using 32bit OSes, on bare metal, RIGHT now, is that clear? And I do that since my machines started to be filled by more than 3GB of RAM, which means more than a decade, and I NEVER faced a single problem, once the correct GPU drivers were installed.

You completely sidestepped the entire chipset list. You claimed early 2000s consumer chipsets supported more than 4Gb of RAM but the actual hardware specs of i845, i865, i915, KT400, and nForce2 prove they physically could not.
All your years in IT and you never noticed something as basic as that?
You said it was so few and they were mobile chipsets so it didnt matter.

And you just conceded the entire argument with one phrase "once the correct GPU drivers were installed."
That is the whole point. You have to actively hunt for the specific drivers that dont choke on high memory buffers.
Microsoft didnt build Windows XP SP2 exclusively for the 1% of enthusiasts who know how to curate working driver versions, they built it for 500 million general users who plugged in whatever cheap PCI expansion cards or OEM driver was on the disc.
You are cherry picking the argument for you. But there even you say yourself you had to cherry pick the drivers because they were broken!

As for Blaster in 2003, millions of homes were hooked directly to USB broadband modems like the SpeedTouch "frog" running dialer PPPoA straight off the PC with a raw public IP. NAT hardware routers were by no means universal in consumer households back then.
Blaster took down hospital networks, train booking systems, and entire ISPs in a weekend without a single click from the user.
45 years in IT should have taught you the difference between what works on an enthusiast test bench today and what happened when a mass market OS hit bargain bin consumer silicon in 2004. The chipset specs dont lie, no matter how much you want to blame Bill Gates for it.

The Solutor wrote on Today, 03:38:

Unpatched system are perfectly fine, in most scenarios, unless you connect them directly using a modem. […]
Show full quote

Unpatched system are perfectly fine, in most scenarios, unless you connect them directly using a modem.

Once you are behind a router that NATs, most problems (if any) come from the person sitting in front of the monitor, no matter the patch level of the OS.

Use a decent browser, use a bit of common sense, stay behind a router/firewall, and no one harms you, unless the router itself has security holes.

IE 4/5/6 was a problem, The fix was either opera if you wanted to pay for it or those in the know saw Pheonix>Firebird>Firefox

FF was a joke compared to Opera, now that most Opera innovations were slowly photocopied by others, even FF is an option, still looks like a joke if compared to Vivaldi.

There were work arounds for that, but then there were work arounds for paying for Win2k3 too

Workarounds that were way more unethical towards a small, honest, European company than they were against the big, evil, American company that MS was.
Perhaps MS became a giant thanks the the so called piracy, a practice that MS tolerated, and even silently backed at the time (and right now), at least for individuals and small/medium sized companies. So stop with the BS about those evil users who made Bill Gates poor.

I just extracted some of your pearls.

Don't mind that I spend a night quoting and debunking all your nonsense line by line.

Being concise isn't exactly your best quality 😁

I mean wow.
Youre ignoring the way people actually did use their computers back then.
It makes me wonder if you were there at the time.

Decent browser... Yeah this looks like yet another "I know more than you and the browser I use proves it" argument. Just pick the most obscure name you can think of and roll with it.
I stopped playing that game in the early 2000s...

Netscape >Opera > Firefox once it actually became useable for me but I guess most just used Netscape then Internet Explorer because it came with windows, which is what got MS into so much trouble around that time. I guess I should mention that professionally it was Netscape and then Internet Explorer .. IE literally changed the internet and how it was rendered and displayed on screen, so professionally you had very little choice till Opera and Firefox caught up.