Windows 98 not installing in an AMD processor

Discuss Windows 95, 98 and ME.
Post Reply
Andreas
User avatar
FTP Access
Posts: 197
Joined: Fri Dec 14, 2012 11:12 am
Location: Jakarta , Indonesia

Re: Windows 98 not installing in an AMD processor

Post by Andreas »

What is your system specs ?

vbdasc
FTP Access
Posts: 167
Joined: Fri Apr 30, 2010 6:14 pm

Re: Windows 98 not installing in an AMD processor

Post by vbdasc »

Make sure that you're using Windows 98 SE, as it solves some time measuring problems in the NDIS stack that cause a crash when the CPU is too fast. Also, if you're trying to install Windows 98 on a too new AMD CPU, like Ryzen, or maybe even anything newer than Bulldozer, you might be out of luck, due to a nasty bug in Windows 98 kernel that was never patched, AFAIK.

http://blog.stuffedcow.net/2015/08/win9 ... ation-bug/

anonymous74
User avatar
FTP Access
Posts: 173
Joined: Mon Jan 27, 2020 9:47 pm

Re: Windows 98 not installing in an AMD processor

Post by anonymous74 »

I've heard that AMD straight up deleted an obsolete instruction that 9x needed to run. Whether I'm right or not, you might want to look into PCem or 86box, modern CPUs just can't run 9x properly, no matter what you try.
My chosen virtual machine programs:
Windows 9x+, Intel macOS: VMware Workstation Pro 16
Older than 9x: PCEm
PPC Mac OS: QEMU

Current Projects: Installing every beta build of Windows 7 in VMware (viewtopic.php?t=43052)
Past Projects: Installing every beta build of Windows XP, Longhorn, and Vista in VMware (https://www.betaarchive.com/wiki/index. ... nonymous74)

DJNW
Posts: 1
Joined: Sat May 14, 2022 9:20 am

Re: Windows 98 not installing in an AMD processor

Post by DJNW »

anonymous74 wrote:
Thu May 12, 2022 4:21 pm
I've heard that AMD straight up deleted an obsolete instruction that 9x needed to run. Whether I'm right or not, you might want to look into PCem or 86box, modern CPUs just can't run 9x properly, no matter what you try.
It's not an instruction in this case(although they did break VME with the early zen chips, but that was fixed). vbdasc linked the person who discovered the cause years ago: http://blog.stuffedcow.net/2015/08/win9 ... ation-bug/

to boil it down: Windows keeps a, uh, map that shows what's actually happening with memory (because you need to keep track of what is actually stored in your physical RAM, or been idle for so long that it's been pushed into the pagefile), similar to how the controller in an SSD has a table of where the bits of files are spread out across the flash to avoid uneven wear, but presents them to the OS as whole files, not chunks. So far, so fine. Useful thing to have. This stuff also gets cached to make it faster.
Problem is, when the bits of a processor that do all the clever guessing on what will happen a few instructions in the future to try to make the chip go faster without having to run the chip run at a higher, hotter clockrate do their thing, those caches don't get updated (search for "tlb coherence", it's a whole thing), so what's in the cache appears to be correct, but is actually wrong. The OS doesn't know the cache content is wrong, so tries to use the it and then Everything Goes A Bit Crashy.
The guy's blog lays out that the way 9x is coded goes against the programming manuals issued by CPU manufacturers at the time about interacting with this stuff.

The blogger shot down the idea of being able to provide a difference-patch that could be applied in-situ to the .VXD that has the troublemaking code in it. I dropped a comment on there, but never had a reply, suggesting that in all the various leaks of Windows code, there might be the source for this particular file and it could be tweaked and recompiled, providing a drop-in replacement to solve the issue.
If the chap that used to post on VOGONS who created the 48-bit LBA patch was still alive, I'd have been seriously tempted to throw some money his way to have a go at it.
Last edited by DJNW on Sat May 14, 2022 9:53 am, edited 1 time in total.

xkai
FTP Access
Posts: 61
Joined: Mon May 26, 2014 12:20 pm

Re: Windows 98 not installing in an AMD processor

Post by xkai »

Use Windows Me instead. Windows Me works fine in new processors.

AlphaBeta
User avatar
Donator
Posts: 2329
Joined: Sun Aug 12, 2012 4:33 pm
Location: Czechia

Re: Windows 98 not installing in an AMD processor

Post by AlphaBeta »

DJNW wrote:
Sat May 14, 2022 9:51 am
... in all the various leaks of Windows code, there might be the source for this particular file and it could be tweaked and recompiled, providing a drop-in replacement to solve the issue.
I am not aware of any Windows 9x source code leak, so the chances of that happening is zero. All the leaks are of NT, which is pretty much its own thing down there.
AlphaBeta, stop brainwashing me immediately!

Image

Omores
Posts: 1
Joined: Mon May 16, 2022 7:00 pm

Re: Windows 98 not installing in an AMD processor

Post by Omores »

Windows 98SE runs very fine on my Ryzen 3900x. I have a fully functional 98SE, with a Voodoo 3 3000 PCI and Ensoniq ES1370 sound card(this one works in Windows 3.11 too), a PCI-E to USB VIA 6212, LAN, WLAN etc.

From my observations, the problems with newer configurations are not the hardware itself but the AMI BIOS implementation.

My motherboard X470GTA has 5 BIOS releases - 07/2019 (initial), 12/2019 and 07/2020 - all are 100% compatible with Windows 3.11 and 9x.

BUT... last two releases are a different story:

- 02/2021release - Windows 98SE hangs and loading bar is crazy fast - sometimes I get VCACHE error instead.
- 07/2021 release - Windows 95A, 98SE and Me all return VCACHE error and Windows 3.11 exists to DOS.

I suspect something related to disk or RAM operation. Still investigating.

Post Reply