virtualbox.org

Just to be clear, "modern CPUs" is your special code for AMD Ryzen, right? I query this because Win98SE runs just fine without tricks on my i9-10900K, which some would consider to be quite modern ty v much. The picture I attached a few posts above was back when I had a measly 6th gen i7, Win98SE obviously ran fine on that too.

Win95 has a well known problem (protection error) with fast CPUs, they don't need to be "modern". The tutorial includes a fix for this.

"My" "Special" Ryzen?

Is that an insult? AMD now has the fastest supercomputer in the world, 5950X remains the fastest desktop MT CPU, 5800X3D beats anything that Intel offers in absolute most games by a wide margin.

Yes, Zen 2 and Zen 3 cannot by default run Windows 95, 98 (SE) and ME because they all these OSes coded incorrectly which was confirmed by a VBox developer. We've had this conversation before only at that time I couldn't find a solution. Now I sort of have.

No insult, but your odd habit of referring to AMD CPUs as "modern processors" instead of clearly stating "AMD Ryzen" has caused confusion in the past. And yes, I'm not aware of anyone else with that habit, hence "your".

There's a 3rd party patch to make 9x work under Ryzen systems too: https://github.com/JHRobotics/patcher9x .

May be worth updating the tutorial with some pointers to this issue with Zen2/3 CPUs.
I don't have an AMD CPU myself so I can't really verify any of this is true. I don't know if out of the box 9x is supposed to work under AMD Zen 2/3.
What I have read:

• The problem seems to be due to "differing behavior" regarding paging in newer AMD CPUs: https://blog.stuffedcow.net/2015/08/win ... ation-bug/ , but apparently newer Intel CPUs have the same issue: https://www.vogons.org/viewtopic.php?p=1085697#p1085697
• In theory disabling "nested paging" option in VirtualBox should workaround the issue (as per the above page) since it will cause the hypervisor to be called during page table manipulation, effectively clearing the cache.
• However, some people claim disabling nested paging introduces slowness and other bugs ( See https://www.vogons.org/viewtopic.php?p=1076221#p1076221 )
• The issue affects Windows 98, 98 SE and ME ( from https://github.com/JHRobotics/patcher9x#requirements , or https://www.vogons.org/viewtopic.php?p=1076747#p1076747 )
• The 3rd party patch at https://github.com/JHRobotics/patcher9x fixes the affected code in 9x itself , so that you can use 9x even with nested paging enabled.

In other news, there is also https://www.os2museum.com/wp/windows-9x ... driver-hd/ -- a VirtualBox video driver for 9x!

javispedro wrote:The issue actually only affects Windows 98 , and not 95 nor ME

I don't agree to this part of your statement for the following reasons:

The source you cite states that it also affects Windows ME, albeit "very rarely", and rare crashes can be even worse IMHO.

The original source states that Pagewalk Coherence is an issue in all three OS variants, and I would trust this source the most.

Win98SE is wayyyyyyyyyyyyyy more popular than Win95 or WinME (I mean residual use now, not in the past - though that is also true), so it stands to reason that most problems will be encountered in Win98SE.

fth0 wrote:The source you cite states that it also affects Windows ME, albeit "very rarely", and rare crashes can be even worse IMHO.
The original source states that Pagewalk Coherence is an issue in all three OS variants, and I would trust this source the most.

You may be right so I lowered the agressiveness of the assertion In any case, I was thinking that in the Vogons thread (now linked) the author of the patch was saying that he actually _never_ saw Windows ME reach that part of the code, even thought the snippet is in there. It is also possible (I guess) that there are more instances of this snippet that may not have been found by the patch author; the original source mentions that but then only gives a sample from 98. Would probably need someone with an AMD to test

javispedro wrote:There's a 3rd party patch to make 9x work under Ryzen systems too: https://github.com/JHRobotics/patcher9x .

May be worth updating the tutorial with some pointers to this issue with Zen2/3 CPUs.
I don't have an AMD CPU myself so I can't really verify any of this is true. I don't know if out of the box 9x is supposed to work under AMD Zen 2/3.
What I have read:

• The problem seems to be due to "differing behavior" regarding paging in newer AMD CPUs: https://blog.stuffedcow.net/2015/08/win ... ation-bug/
• In theory disabling "nested paging" option in VirtualBox should workaround the issue (as per the above page) since it will cause the hypervisor to be called during page table manipulation, effectively clearing the cache.
• However, some people claim disabling nested paging introduces slowness and other bugs ( See https://www.vogons.org/viewtopic.php?p=1076221#p1076221 )
• The issue affects Windows 98 for sure , unclear if it affects 95 or ME ( from https://github.com/JHRobotics/patcher9x#requirements , or https://www.vogons.org/viewtopic.php?p=1076747#p1076747 )
• The 3rd party patch at https://github.com/JHRobotics/patcher9x fixes the affected code in 98 itself , so that you can use 98 even with nested paging enabled.

In other news, there is also https://www.os2museum.com/wp/windows-9x ... driver-hd/ -- a VirtualBox video driver for 9x!

The JHRobotics/patcher9x patch works beautifully, thanks! It's a little tricky to apply:

• You start installing Windows 98 SE as usual.
• Once the initial graphical COPYING files section completes and Windows reboots, you let it reboot and then immediately shut it down.
• You then apply the patch using option 3 (do 1 and 2 simultaneously)
• You continue with installation.

I don't think warez is accepted here but if anyone's interested please let me know and I'll send you a link to a VM prepared this way. I've done zero modifications to it except, obviously, patching VMM32.VXD and installing VBE9X drivers from here (version 2014.02.14). Unfortunately these drivers are not compatible with the VBoxVGA vGPU, so I had to change it to VBoxSVGA.

I've also researched the topic of the AC'97 codec but since I couldn't find "native" Microsoft Windows Update drivers for it I switched back to SoundBlaster16.

Realtek AC'97 drivers are quite heavy for such a light installation (~200MB).

birdie wrote:Unfortunately these drivers are not compatible with the VBoxVGA vGPU, so I had to change it to VBoxSVGA.

I'm afraid that makes little sense to me. The BearWindows drivers are simple VESA drivers, they have no access to any of the simulated graphics controllers. They use the VESA interface exclusively.

And in any case I have been successfully using Win98SE VMs with BearWindows and VBoxVGA for years and am doing so right now. The option to use a different controller didn't even exist before VirtualBox v6, and I think we would all have noticed if Win98 didn't run.

I'm a newbie here. I just signed up a few minutes ago. I have a quick question for all. I hope you can answer it easily. Are people still using Windows 95/98? I already upgraded my windows to 11. What about you?

This topic is for discussing the Win9x guest tutorial. If you wish to strike up a different conversation then please start your own topic.

To briefly answer: some people need to keep old systems alive to run old software. However I'd hazard a guess that most of us like to have old Windows versions - because we can, or for nostalgia reasons, or to remind ourselves of how past technologies worked. Not for daily use. VirtualBox will not run on Win9x hosts so naturally our daily PC is something else.

Just FYI I did try the "new" driver suggested by javispedro and I would advise everyone not to use it:

after that I had to boot into safe mode and revert to the old universal driver: https://bearwindows.zcm.com.au/vbe9x.htm

I have read some feedback that the paging issue also affects newer Intel CPUs too -- 12th generation at least. The same patch fixes the issue. I've updated my post.

Just FYI I did try the "new" driver suggested by javispedro and I would advise everyone not to use it:

I have been using the OS/2 museum driver for weeks now and never had that problem, what did you do to trigger it ?

javispedro wrote:I have read some feedback that the paging issue also affects newer Intel CPUs too -- 12th generation at least.

I'm not so sure. The post was only two hours before the announcement of a newer version which additionally handles the CPU speed issue. Did you clearly read that it's related to the TLB invalidation issue somewhere?