VOGONS


First post, by superfury

User metadata
Rank l33t++
Rank
l33t++

Can Windows 98 run on a Nexgen nx586? It has Pentium support, but without the FPU afaik(that's what UniPCemu essentially emulates, although reporting as a Pentium without FPU instead)? So can UniPCemu theoretically run Windows 98(without FPU support built in the CPU)?

I managed to fire up the setup.exe with the /NM switch, causing it to no longer throw an error about an unsupported CPU without the 'required' FPU.

Does it require an actual FPU to boot, or does it (like Windows 95) use FPU emulation of the CPU(as supported on all 80286+ CPUs)?

Edit: Apparently a 80386 is fine too? Just use the /NM switch to install? According to https://forums.anandtech.com/threads/bare-min … -win98.1240055/ , the post by aswedc?

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 2 of 9, by superfury

User metadata
Rank l33t++
Rank
l33t++

What about a Pentium without an FPU(nx586?)? Will it also crash on that? My emulation has all the Pentium instructions, just no FPU instructions have been emulated(they're all NOPs with modr/m added, except adding the required #NM exception handler mechanics for them to allow software to emulate the FPU. That's the way it can allow Descent to run at all(using the FPU emulation options and external emulation software that's on the harddisk).

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 3 of 9, by superfury

User metadata
Rank l33t++
Rank
l33t++

Setup so far seems to run inside UniPCemu(Pentium without FPU on Compaq Deskpro 386, ET4000 graphics card):

1130-Windows 98 ready to reboot..jpg
Filename
1130-Windows 98 ready to reboot..jpg
File size
117.96 KiB
Views
1077 views
File comment
Windows 98 setup - ready to reboot after ~12:30 hours of installing on a i7-4790K.
File license
Fair use/fair dealing exception

Although the time excludes all steps before the file copying step. Just the file copying took approximately 12 and a half hours.

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 4 of 9, by superfury

User metadata
Rank l33t++
Rank
l33t++

Hmmm.... First boot seems to get to a 0xFFFF #UD instruction hanging the boot process? It's emulating 12MB RAM on the Compaq Deskpro 386 architecture with Pentium(without FPU)?

Perhaps not enough RAM? Now retrying with 16MB RAM installed(of which only 15360KB is recognised by the BIOS, due to the Compaq Deskpro 386 BIOS having a limit of 15MB(+1MB high memory hole)).

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 5 of 9, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie

I’ve found the only way to get 98 on a 386 was to remove the hard drive, install using something else (like a 486) then put the hard drive in the 386.

On my 386 deskpro the HD + controller wouldn’t allow windows 95 dos to SYS and be bootable.

Reply 6 of 9, by superfury

User metadata
Rank l33t++
Rank
l33t++

What about required memory? The Deskpro 386 can apparently(with the memory holes) only support up to just under 16MB of RAM(~15.3MB actually, from 0-640K and 1M-15M, thus 14MB high+640K low?)? The 15M-16M and 640K-1M memory holes need to exist, don't they?

One thing I notice booting 98 on the 14.64MB system(UniPCemu's Compaq Deskpro 386 current limits(mainly due to bios limits)) is that at some point, it's adding some number using an add instruction having it's source/destination in the pre-1MB(640K-1MB) memory gap, resolving to nothing, always reading 0xFF(due to memory not responding)? If I remember correctly, it was a VGA address(between A0000-BFFFF), but the VGA apparently isn't responding to said address?
Perhaps the VGA somehow was instructed to disable it's VRAM window? That's weird to do while windows is still booting and rendering it?

Edit: Yup. Probably the VGA misc output register bit 1 being cleared, disabling VGA MMIO, thus resolving to the memory hole(which has no memory, thus reading 0xFF and writing into a black hole essentially)?

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 8 of 9, by superfury

User metadata
Rank l33t++
Rank
l33t++

The weird thing is that it shouldn't crash at that point anyways? According to bootlog.txt, the last entry is the first mention on loading the IOS vxd driver? It's not even at the first/second init stages yet. It's literally loading IOS(before the subsystems are loaded) when it's crashing! It can't possibly run out of memory with only that much drivers loaded, can it?

The log file is around <1500 bytes in size, so it's only at the initial loading stage(the stage directly after loading HIMEM and the filesystem real mode mapper driver).
Edit: It's the stage after loading IFSHLP.SYS.

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 9 of 9, by superfury

User metadata
Rank l33t++
Rank
l33t++

Tried the Windows 98 first reboot image on Dosbos SVN Daum. It simply rebooted (Reboot requested message) when running the boot command?

Windows 95 C from UniPCemu (as the only hard drive, imgmounted on C) seemed to boot without issues(although complaining about the 1F0/3F0(primary ATA controller) disk image and FDC(perhaps the second floppy disk?), but not about the secondary ATA controller (170/370) which isn't even mounted in the system properties hardware devices tab?).

So windows 95 c setup works correctly in UniPCemu, it's just the normal boot process that crashes on the "Initializing KERNEL" from bootlog.txt?

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io