virtualbox.org

I'm not sure what this bug report is down to, either my CPU or a new release of VirtualBox (6.1.2) but I wonder if there are any people here who can try installing Windows 98 SE on a CPU other than Ryzen 3000 and see if it works without errors 'cause in my case it's 100% unusable - either the guest OS crashes or VirtualBox itself crashes.

Edit: Windows 95/98/98SE/ME are all affected - see the explanation below.

My Win98SE VM is not newly installed, but it is currently running just fine in 6.1.2 on my i7-6700 home desktop.

For that particular installation of Win98SE I followed the tutorial to get an ACPI-enabled Win98SE.

mpack wrote:My Win98SE VM is not newly installed, but it is currently running just fine in 6.1.2 on my i7-6700 home desktop.

For that particular installation of Win98SE I followed the tutorial to get an ACPI-enabled Win98SE.

VirtualBox 6.x features some changes pertaining exactly to Ryzen 3000/Zen 2 CPUs support which have already caused serious issues.

Looks like Intel users are doing fine. Hopefully VBox developers can look into these issues even further because some of them are not yet resolved.

Setting --cpu-profile to "Intel Core Duo T2600 2.16GHz" makes things worse.

Windows ME doesn't work either

Can't be installed no matter which options I try.

Nested pages/Paravirtualization interface/IO APIC/nothing helps.

Don't I remember something in a tutorial on Virtualbox, that a successful install of W9x requires VT-x to be turned off during the install? If so, Virtualbox 6.1.x has removed the ability to turn VT-x off for a guest, so W9x can't be installed. Maybe an oversight by the developers? Maybe W9x is on the way out?

Try installing in 6.0.x then upgrading to 6.1.x. To the best of my remembrance, VT-x can be on after the install is complete.

scottgus1 wrote:Don't I remember something in a tutorial on Virtualbox, that a successful install of W9x requires VT-x to be turned off during the install?

I think that's one of those things that was true for about 10 seconds, a few years ago, and has been copied around forever since. Oh, and ISTR the assertion was that VT-x had to be off for Win98 to run, not just that you couldn't install it. The Win98SE VM I have now was first installed to test that old theory, and I could not reproduce the problem.

scottgus1 wrote:Don't I remember something in a tutorial on Virtualbox, that a successful install of W9x requires VT-x to be turned off during the install? If so, Virtualbox 6.1.x has removed the ability to turn VT-x off for a guest, so W9x can't be installed. Maybe an oversight by the developers? Maybe W9x is on the way out?

Try installing in 6.0.x then upgrading to 6.1.x. To the best of my remembrance, VT-x can be on after the install is complete.

No issues with VMWare Workstation which has used HW assisted virtualization for many years now (I don't even know how to switch back to software virtualization).

Also, VirtualBox consistently crashes while running both W98 and WME guests on a Ryzen 3000 system.

Looks like some major changes introduced in 6.0/6.1 have some bugs which only manifest themselves under old Windows versions.

mpack wrote:I think that's one of those things that was true for about 10 seconds, a few years ago

I have a 98 guest, I can't remember whether I turned it off myself. I have to try it again...

I have to try this possible workaround:

michaln wrote:There is a possible workaround: VBoxManage modifyvm <vm> --cpuidset 1 00800f11 00000800 56d8220b 078bfbfd

This masks off the VME bit in CPUID. It is currently unclear why that should make a difference, but it reportedly does.

More info on the topic:

http://blog.stuffedcow.net/2015/08/win9 ... ation-bug/
http://www.os2museum.com/wp/vme-broken-on-amd-ryzen/
https://www.vogons.org/viewtopic.php?t=68205
https://communities.vmware.com/thread/579537
https://community.amd.com/thread/244263

Looks like your manual here about installing, sorry not being able to install Windows 95/98/98SE/ME guests under Ryzen CPUs should mention these facts. I don't know whether Windows 3.1(1) is affected as it's a 16 bit OS without multitasking.

Please stop using external sites as authorities, especially not 5 years old blogs. None of those count as "facts".

The only relevant documents are the user manual and the BugTracker. We also like to believe that information given on this site, in the howto section, is accurate.

mpack wrote:Please stop using external sites as authorities, especially not 5 years old blogs. None of those count as "facts".

The only relevant documents are the user manual and the BugTracker. We also like to believe that information given on this site, in the howto section, is accurate.

In the bug report I created two Oracle employees have said just the same but you apparently haven't bothered to look.

And the links I posted are from this bug report as well.

What an amazing attitude we have here.

You are absolutely correct that I have not bothered to look at your reports (much), because I don't use Ryzen and nor do I plan to. I'm somewhat puzzled why you think I should have done so. I already checked and commented earlier that I don't have the problem on Intel. Job done.

Was there something else you assumed I would be interested in?

As moderator I am however interested that you don't treat external blogs as authorities on this site.

mpack wrote:As moderator I am however interested that you don't treat external blogs as authorities on this site.

Again the links I posted were given/endorsed by Oracle developers. I'm not sure what is that we're arguing about here.

The user suggested fix:

"VBoxManage modifyvm <vm> --cpuidset 1 00800f11 00000800 56d8220b 078bfbfd"

Does not work on my Windows 98SE VM on my Surface Pro 7 with VirtualBox v6.1.14

Is Windows 95/98/ME compatibility broken for newer processors?

Is there any viable workaround relevant to Fall 2020 using VirtualBox v6.1?