VOGONS


SBEMU: Sound Blaster emulation on AC97

Topic actions

Reply 1980 of 1999, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

Yes, this is the source where I found it. Some drivers works to some point that is interesting.
e.g. WINDISK solves me boot VDMA issue on SATA3 port but the driver is so poor it cause freezing later on multiple disks access/query.
WINVBE is pretty fast compareed to VBEMP but cause some win32 apps crash when open.
So it brings something new and cool but it's useless for real (I wouldn't say daily) use.
If you noticed, I already created some issues on github but don't expect much happen...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 1981 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
myne wrote on 2026-10-23, 00:53:

Look into what glidulator is doing.
I realise DMA is a possible hurdle, but it looks promising to offload to another core.
No need to care about all the memory problems once you've carved out your niche.

I don't see the need for now. it's an optional feature as SBEMU needs to run on PCs around 2000 era when there was no multi core.
BTW there is no memory problems for now, its optimizations. all memory are put into himem and when "DOS" conventional memory is needed, it's really needed that way. but it can be optimized out.
and multicore management needed to implemented by a supervisor otherwise different drivers will may conflict.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 1982 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
RayeR wrote on 2026-10-23, 06:45:
Hehe, I can give you more precise hint now. I have found that the problem I reported before (that only 1st dir entry is visible […]
Show full quote
crazii wrote on 2026-10-21, 20:58:

Congrats on that and hopefully I will improve USBDDOS in the future, but I'm still busy on VDPMI so I have to put off.

Hehe, I can give you more precise hint now.
I have found that the problem I reported before (that only 1st dir entry is visible by on disk) happens only when I have loaded SHSUCDX.COM (the MSCDEX replacement) optical drives driver in my autoexec.bat that provides 2 drive letters for 2 DVDs and following drive letter is assigned by USBDDOS. When I skip loading the SHSUCDX (or load it after USBDDOS) it works fine then and I can access all files on my USB disk and no lost clusters are reported. So it seems your driver may corrupting some other's memory.
Interesting is that WINUSB behaves almost the same but only under MSDOS 7.10 while under MSDOS 6.22 it works with SHSUCDX loaded. So it gives a hint where the AI learned 😜
Please try with SHSUCDX and optical drive on real HW if you will able to reproduce the bug....

Very interesting. Hopefully I can fix that with your info.

rasz_pl wrote on 2026-10-23, 12:30:
someone just ai slopped a TON of drivers for old stuff, including USB 1/2/3 stack for DOS, Windows 1.0x / 2.x, Windows 3.0 Real […]
Show full quote
crazii wrote on 2026-10-21, 20:58:

hopefully I will improve USBDDOS in the future

someone just ai slopped a TON of drivers for old stuff, including USB 1/2/3 stack for DOS, Windows 1.0x / 2.x, Windows 3.0 Real Mode, Standard Mode and 386 Enhanced Mode (VxD)/ Windows 95 / 98 / ME https://github.com/ages2001/WINUSB

Edit: oh, and apparently https://github.com/ages2001/WINAUDIO "WINAUDIO — HD Audio / AC'97 / Intel audio DSP sound driver for DOS, Windows 3.x and Windows 95 / 98 / ME", curious how the interrupts and SB emulation was handled by the lawnmower, whose code did it rip off 😀
more at Re: List of open-source PC hardware projects

edit2: oh its very clever, it doesnt emulate anything at all 😁 and doenst support DirectSound 😀 drivers to nowhere 😀

Thanks for the info but please stop talking that way. you don't have to do that.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 1983 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
Lualb wrote on 2026-10-13, 02:49:

Hi Crazii, how are you? Sorry to ask this—I've been waiting for this for so long—but if I add GEMMIS to the memory extender on Windows ME (with the DOS real-mode patch enabled and the io8emmok.sys driver loaded so EMM386.EXE can initialize), will I be able to boot into Windows without causing any errors?

Hi there, I'm going to test with win98 and ME, can you be more specific of the problem? As I googled I found some patch being said to fix WinME with EMM386: https://github.com/gpdm/TPC-WinMe-DOSMODE/blo … /FILESOURCES.md
Doe that patch works for you? or do you want a unpatched WinME to work?

I have a laptop with WinME and I remember only the command.com was patched years ago. but never tried EMM386 on it.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 1984 of 1999, by Lualb

User metadata
Rank Newbie
Rank
Newbie
crazii wrote on 2026-10-24, 22:54:
Hi there, I'm going to test with win98 and ME, can you be more specific of the problem? As I googled I found some patch being sa […]
Show full quote
Lualb wrote on 2026-10-13, 02:49:

Hi Crazii, how are you? Sorry to ask this—I've been waiting for this for so long—but if I add GEMMIS to the memory extender on Windows ME (with the DOS real-mode patch enabled and the io8emmok.sys driver loaded so EMM386.EXE can initialize), will I be able to boot into Windows without causing any errors?

Hi there, I'm going to test with win98 and ME, can you be more specific of the problem? As I googled I found some patch being said to fix WinME with EMM386: https://github.com/gpdm/TPC-WinMe-DOSMODE/blo … /FILESOURCES.md
Doe that patch works for you? or do you want a unpatched WinME to work?

I have a laptop with WinME and I remember only the command.com was patched years ago. but never tried EMM386 on it.

Hello! The reason I asked whether your software could be used in WinME—either in real DOS mode or from within Windows—is that, as you likely know, WinME does not support any type of expanded memory manager required to enter graphical mode. Furthermore, it does not allow EMM386 to be loaded unless you add `io8emmok.sys` to your `config.sys` file. It’s just a matter of curiosity; I understand that you aim to make it compatible with the graphical environment—much like how one might load `Sbemu.exe` from `autoexec.bat` before entering the Windows desktop—by incorporating GEMMIS into your software.

Reply 1985 of 1999, by Lualb

User metadata
Rank Newbie
Rank
Newbie

After many failed attempts to use JEMMEX.EXE, I tried VDPMI.EXE instead, which allowed me to get Sbemu working with good results. By the way, the error shown in the first image comes from Ctmouse.exe; it fails to detect the USB mouse, leaving only the keyboard functional. I’m not sure what the cause might be—perhaps an incompatibility with VDPMI.EXE ???

Reply 1986 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
Lualb wrote on 2026-10-25, 02:13:

WinME does not support any type of expanded memory manager required to enter graphical mode. Furthermore, it does not allow EMM386 to be loaded unless you add `io8emmok.sys` to your `config.sys` file.

I'm aware that EMM386 cannot load without IO8EMMOK.sys, but WinME won't launch if EMM386 is installed? With or without DOS=UMB? I'll try it later to find the result.
VDPMI works without IO8EMMOK.sys as tested, as long as DOS=UMB is REMOVED. EMS seems fine with it.
Maybe I'll have to do some dirty work to try making it launch while DOS still have UMB.

EDIT: tested with EMM386+IO8EMMOK, same result, NOEMS/RAM + DOS=UMB won't launch.

Lualb wrote on 2026-10-25, 04:06:

After many failed attempts to use JEMMEX.EXE, I tried VDPMI.EXE instead, which allowed me to get Sbemu working with good results. By the way, the error shown in the first image comes from Ctmouse.exe; it fails to detect the USB mouse, leaving only the keyboard functional. I’m not sure what the cause might be—perhaps an incompatibility with VDPMI.EXE ???

JEMMEX conflicts with WinME's builtin XMS, you may try JEMM386 with the HDPMI edition of SBEMU.
Is your keyboard also connected via USB? or have you enabled USB legacy support in your BIOS?

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 1987 of 1999, by Lualb

User metadata
Rank Newbie
Rank
Newbie
crazii wrote on 2026-10-25, 10:09:
I'm aware that EMM386 cannot load without IO8EMMOK.sys, but WinME won't launch if EMM386 is installed? With or without DOS=UMB? […]
Show full quote
Lualb wrote on 2026-10-25, 02:13:

WinME does not support any type of expanded memory manager required to enter graphical mode. Furthermore, it does not allow EMM386 to be loaded unless you add `io8emmok.sys` to your `config.sys` file.

I'm aware that EMM386 cannot load without IO8EMMOK.sys, but WinME won't launch if EMM386 is installed? With or without DOS=UMB? I'll try it later to find the result.
VDPMI works without IO8EMMOK.sys as tested, as long as DOS=UMB is REMOVED. EMS seems fine with it.
Maybe I'll have to do some dirty work to try making it launch while DOS still have UMB.

EDIT: tested with EMM386+IO8EMMOK, same result, NOEMS/RAM + DOS=UMB won't launch.

Lualb wrote on 2026-10-25, 04:06:

After many failed attempts to use JEMMEX.EXE, I tried VDPMI.EXE instead, which allowed me to get Sbemu working with good results. By the way, the error shown in the first image comes from Ctmouse.exe; it fails to detect the USB mouse, leaving only the keyboard functional. I’m not sure what the cause might be—perhaps an incompatibility with VDPMI.EXE ???

JEMMEX conflicts with WinME's builtin XMS, you may try JEMM386 with the HDPMI edition of SBEMU.
Is your keyboard also connected via USB? or have you enabled USB legacy support in your BIOS?

Hi ! ... I have checked the BIOS settings: Legacy USB Support is enabled, Port 60/64 Emulation is also enabled, and XHCI/EHCI Hand-off is enabled. The BIOS detects one USB keyboard, one USB mouse, and three hubs; my computer does not have any physical PS/2 ports. Without VDPMI, CTMOUSE installs correctly on the Virtual PS/2 port (thanks to the BIOS’s Port 60/64 emulation). However, when VDPMI is loaded, CTMOUSE no longer detects the USB mouse, even though all the relevant BIOS settings are enabled—the keyboard continues to work within the system, but CTMOUSE fails in the DOS environment.

Reply 1988 of 1999, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

I tried the latest VDPMI (from 22.10.) but I have some problem with it.

1) my tools like CPUID and MTRRLFBE crashed when acessing GDT even they are whitelisted in VDPMI.CFG (I have both files directly in C:\, loaded in config.sys by DEVICE=VDPMI.EXE)

2) when trying to start Win98 - hoping that GEMMIS works and allows take over it crashed too.

At least it works with SBEMU and Doom.
BTW I found that when use WinUSB with switch /DOSDISK it solve interference with SHSUCDX and I can read whole USB stick. Japheth tried to reproduce this bug on his AHCI-only machine with AHCI CD driver and SHSUCDX but he didn't observed this issue...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 1989 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
Lualb wrote on 2026-10-25, 14:50:

Hi ! ... I have checked the BIOS settings: Legacy USB Support is enabled, Port 60/64 Emulation is also enabled, and XHCI/EHCI Hand-off is enabled. The BIOS detects one USB keyboard, one USB mouse, and three hubs; my computer does not have any physical PS/2 ports. Without VDPMI, CTMOUSE installs correctly on the Virtual PS/2 port (thanks to the BIOS’s Port 60/64 emulation). However, when VDPMI is loaded, CTMOUSE no longer detects the USB mouse, even though all the relevant BIOS settings are enabled—the keyboard continues to work within the system, but CTMOUSE fails in the DOS environment.

There could be compatibility problems of the VDPMI builtin keyboard/mouse driver. There are bad keyboard response reported in github.
you can set VIntFlag=0 in the VDPMI.CFG to disable the builtin drivers. VDPMI.CFG is available in recent release, may not in the one you're using.

RayeR wrote on 2026-10-28, 04:11:

1) my tools like CPUID and MTRRLFBE crashed when acessing GDT even they are whitelisted in VDPMI.CFG (I have both files directly in C:\, loaded in config.sys by DEVICE=VDPMI.EXE)

I haven't tested with recent changes, it turns out broken. I've fixed it in the new attachment. it's a bug in the loader that access a freed buffer and corrupts the whitelist.

RayeR wrote on 2026-10-28, 04:11:

2) when trying to start Win98 - hoping that GEMMIS works and allows take over it crashed too.

It seems the exception context is valid, i.e. not a wild execution. It is LMSW instruction, it's OK as long as the PE bit is set. It's possible that this is a untested code path on your CPU, related to FPU settings, which I haven't encountered with my CPU. It also could be a rare chance that VDPMI failed to switch to real mode and still in v86. Anyway I've add emulation for LMSW, you can test with the new attachment below.

I've finished testing win98se and winME, and added special fix for the UMB:

new option:
/UMB80: prevent DOS kernel from being loaded to UMB.
only visible and accessible in dos 8.0 aka winme only.
other LH/DEVICEHIGH still works, only 10K DOS kernel data kept in conventional memory.
This is a hack that patches two DOS8.0 memory locations, 1 in dos loader data, 1 in dos data. along with some IVT fixes if IVT hooked by UMB address (interrupt installed by driver through LH/DEVICEHIGH), and DOS device header in UMB (loaded by DEVICEHIGH).

Here is the comparison of mem.exe otuput without and with /UMB80.

The attachment WinME_ImprovedUMB.png is no longer available
The attachment WinME_OriginalUMB.png is no longer available

With /UMB80 (WinME_ImprovedUMB) the block device tables, FILES, FCBS, BUFFERS, LASTDRIVE=Z, STACKS and still in conventional memory.
this allow WinME to startup with DOS=UMB. Note that all drivers loaded with LH/DEVICEHIGH won't be accessible in WinME GUI session, they are hidden.
To make the /UMB80 hack work, "DEVICE=C:\WINDOWS\IFSHLP.SYS" must be the last line in config.sys

I will continue with more tests and bugfixes with SBEMU on real PC.

The attachment vdpmi0.33fix.zip is no longer available

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 1990 of 1999, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

Thanks for update, good work.
Whitelist is now fixed and works.
Windows 98SE can boot with VDPMI, this is a milestone, but some issues remain/appeared:

1) VDPMI don't like with R.Loew RAMDSK64.COM and HIMEMEX.SYS - HIMEMEX.SYS loads fine (before VDPMI) but then when loading RAMDSK64.COM it just reboot the machine. I have to disable it for testing. (this is RAMDISK compatible with Windows that can use memory beyond 4GB - usefull for swap and temp)

2) It crashed when loading NDIS driver for my Realtek eth. adapter in autoexec. I have to do this trick because Realtek's windows 32b driver (that is yet compatible with Win9x) don't support my revision of RTL8111E chip but DOS NDIS driver supports it and Windows can work with NDIS interface. See below...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 1991 of 1999, by Plombo

User metadata
Rank Newbie
Rank
Newbie
crazii wrote on 2026-09-15, 00:24:

BTW SBEMU also have new features thanks to andersrodrig: PC speaker, Covox and GUS emulations. there's a test build using VDPMI: https://github.com/andersrodrig/SBEMU/releases/tag/sbemu
the GUS emu is not working properly yet.
it will be merged and available on next release.

What happened to this? I can see the PC speaker emulation was merged, but I never saw any reference to GUS in the source code, and the andersrodrig/SBEMU repo seems to have been deleted. Has the GUS emulation been canceled? If it has, I might want to take my own shot at porting DOSbox's GUS emulation to SBEMU.

Some other things I'm interested in regarding SBEMU:

  • Using the SB Live/Audigy hardware synth for MIDI playback with SoundFont support. It's always bothered me that the official Live/Audigy DOS driver uses a low-quality software synth when the cards have such good hardware synth capabilities. Implementing this in SBEMU would be a significant undertaking, and I might be the only person who cares about it, so I'll probably have to port the relevant Linux kernel code myself.
  • Running Windows 3.1 with SBEMU as a Sound Blaster and MPU-401. From what I understand, this sort of thing is what VDPMI's GEMMIS support is for, so I guess I don't need to worry about that.
  • Running Ultima VII with SBEMU. That game is incompatible with EMM386, etc., because it implements its own custom EMM. I think some other Origin games of that era use the custom EMM, too. Will VDPMI allow such games to work with SBEMU? That's not a rhetorical question or anything; I don't know very much about DOS memory management or runtime environments, so I would like someone to explain whether it's possible or not 😀

Reply 1992 of 1999, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

I still use old PCI SB Audigy so better MIDI may be interesting but used timidity emulation seems to me good enough for games. It would be nice to have direct comparison.

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 1993 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
Plombo wrote on 2026-10-30, 21:12:
What happened to this? I can see the PC speaker emulation was merged, but I never saw any reference to GUS in the source code, a […]
Show full quote
crazii wrote on 2026-09-15, 00:24:

BTW SBEMU also have new features thanks to andersrodrig: PC speaker, Covox and GUS emulations. there's a test build using VDPMI: https://github.com/andersrodrig/SBEMU/releases/tag/sbemu
the GUS emu is not working properly yet.
it will be merged and available on next release.

What happened to this? I can see the PC speaker emulation was merged, but I never saw any reference to GUS in the source code, and the andersrodrig/SBEMU repo seems to have been deleted. Has the GUS emulation been canceled? If it has, I might want to take my own shot at porting DOSbox's GUS emulation to SBEMU.

Some other things I'm interested in regarding SBEMU:

  • Using the SB Live/Audigy hardware synth for MIDI playback with SoundFont support. It's always bothered me that the official Live/Audigy DOS driver uses a low-quality software synth when the cards have such good hardware synth capabilities. Implementing this in SBEMU would be a significant undertaking, and I might be the only person who cares about it, so I'll probably have to port the relevant Linux kernel code myself.
  • Running Windows 3.1 with SBEMU as a Sound Blaster and MPU-401. From what I understand, this sort of thing is what VDPMI's GEMMIS support is for, so I guess I don't need to worry about that.
  • Running Ultima VII with SBEMU. That game is incompatible with EMM386, etc., because it implements its own custom EMM. I think some other Origin games of that era use the custom EMM, too. Will VDPMI allow such games to work with SBEMU? That's not a rhetorical question or anything; I don't know very much about DOS memory management or runtime environments, so I would like someone to explain whether it's possible or not 😀

I don't know much detail about the SB live/audigy capabilities, but I glad that you're willing to do it.
Windows support (GEMMIS) for now is only for compatibility and convenience (without restarting to switch to Windows), the technical path is not undetermined, but surely it will be done.
I don't think it's much possible to run Ultima VII with SBEMU. it has its own memory manager and it is difficult to hack it.

About the GUS support, I haven't check it yet, but according to andersrodrig, GUS emulation may need trap memory access and forward Page Fault (#FP) to SBEMU, which is not implemented. there are discussions between andersrodrig and me here: https://github.com/crazii/SBEMU/issues/200
The GUS code is not integrated to the main branch, but still in a test branch (https://github.com/crazii/SBEMU/tree/vdpmi_pc … x_gus_mergetest), which I haven't merge it yet.

RayeR wrote on 2026-10-30, 05:30:
Thanks for update, good work. Whitelist is now fixed and works. Windows 98SE can boot with VDPMI, this is a milestone, but some […]
Show full quote

Thanks for update, good work.
Whitelist is now fixed and works.
Windows 98SE can boot with VDPMI, this is a milestone, but some issues remain/appeared:

1) VDPMI don't like with R.Loew RAMDSK64.COM and HIMEMEX.SYS - HIMEMEX.SYS loads fine (before VDPMI) but then when loading RAMDSK64.COM it just reboot the machine. I have to disable it for testing. (this is RAMDISK compatible with Windows that can use memory beyond 4GB - usefull for swap and temp)

2) It crashed when loading NDIS driver for my Realtek eth. adapter in autoexec. I have to do this trick because Realtek's windows 32b driver (that is yet compatible with Win9x) don't support my revision of RTL8111E chip but DOS NDIS driver supports it and Windows can work with NDIS interface. See below...

VDPMI had design flaws that not working well with external XMM, especially for Windows, as you may already notice the compatibility notes on Windows, in last VDPMI.TXT, that it won't work with Windows with HIMEM.SYS. but I managed to fix that with another method, now it should be compatible with external XMMs. I'm not sure it related to your problem, but I will continue to improve it. 😁

I've done some simple test with my TP T540p, and I found it's so broken due to recent aggressive changes. I have fixed some of them:

1.critical bug related to a tiny buffer, causing protected mode INT10 VESA/VBE fix not working properly, seen on Warcraft2, Duke3d. both crash on my PC.
2.critical bug that Lemmings 2 crashes with /EMSX
3.VDMA related tiny bugs that not always calls the VMDA virtualization callback, seen on Prince of Persia 2, intro speech broken, on the 1st line.
4.add a halt method to wait Ctrl+Alt+Del on critical errors, and it is used when /EMSX is detected when launching Windows, it prevent anyone who doesn't read the VDPMI.TXT so doesn't know /EMSX is not compatible with Windows.
5.Ctrl+Alt+Del handling with VIntFlag=0 in VDPMI.CFG or /DPMI=0 in command line option, because the two settings will disable internal drivers that handle the keystrokes.
6.fix built-in XMM that hard limit XMS to 64M for Windows, only seen around 64M memory available compared to HIMEM.SYS.

I will continue the test and fixes with my other PCs. I will focus on the full ALSA support for SBEMU when it's stable.

EDIT: the "VIntFlag=0" in VDPMI.CFG was also broken while adding win3.x std mode support, it will mute SBEMU. please wait for the next release.

The attachment vdpmi0.34.zip is no longer available

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 1994 of 1999, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie

Thanks, new version don't crash with Realtek NDIS2 driver and network subsystem loads and works fine.
(For this configuration I need to run Windows disk subsystem in legacy 16bit mode due to BIOS VDMA/VDS bug).

RAMDISK64 still reboots. Just to note, RAMDISK64.COM use special XMS extension driver HIMEMEX.SYS which extends current XMS driver (HIMEM.SYS or any other) with functions to access memory beyond 4GB. I guess it switched to 64b mode and back - this probably interferes with VDPMI, maybe some registers are not restored properly...

BTW I found that VDPMI alows to run GLIDULATOR on multiple cores contrary to other EMMs/VMMs that's nice...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 1995 of 1999, by Plombo

User metadata
Rank Newbie
Rank
Newbie

With VDPMI 0.34, Windows 3.11 starts up and seems to run fine...unless SBEMU has been loaded first. If I run SBEMU first, VDPMI blows up with an exception (see attached photo) when trying to start Windows.

config.sys:

DOS=HIGH,UMB
FILES=40
BUFFERS=40
STACKS=9,256
DEVICE=c:\dosbin\vdpmi\vdpmi.exe

Reply 1996 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
RayeR wrote on 2026-11-07, 04:35:
Thanks, new version don't crash with Realtek NDIS2 driver and network subsystem loads and works fine. (For this configuration I […]
Show full quote

Thanks, new version don't crash with Realtek NDIS2 driver and network subsystem loads and works fine.
(For this configuration I need to run Windows disk subsystem in legacy 16bit mode due to BIOS VDMA/VDS bug).

RAMDISK64 still reboots. Just to note, RAMDISK64.COM use special XMS extension driver HIMEMEX.SYS which extends current XMS driver (HIMEM.SYS or any other) with functions to access memory beyond 4GB. I guess it switched to 64b mode and back - this probably interferes with VDPMI, maybe some registers are not restored properly...

BTW I found that VDPMI alows to run GLIDULATOR on multiple cores contrary to other EMMs/VMMs that's nice...

I checked the source of HIMEMEX.SYS and RAMDSK64.COM, it needs a VxD for Windows and VDPMI works more like a Win9x env, so the VxD is needed but VDPMI doesn't support VxD.
HIMEMEX.SYS uses privilege instruction to set CR0, it's also possible to set another whitelist for it, and load it after VDPMI.EXE but I'm not sure it will work or not eventually.

I did nothing for GLIDULATOR and never tested that. it's just happens. BTW VDPMI is currently only a EMM or V86 monitor, but not a VM Manager yet, although it's possible in the future, as it's considered in the beginning, and that's why I really didn't want to add VCPI and the "VIntFlag" thing. those break the original idea.

Plombo wrote on 2026-11-11, 04:04:

With VDPMI 0.34, Windows 3.11 starts up and seems to run fine...unless SBEMU has been loaded first. If I run SBEMU first, VDPMI blows up with an exception (see attached photo) when trying to start Windows.

I hadn't test SBEMU with windows but I didn't some changes and tests, now Windows can start with SBEMU loaded. but it DOES NOT mean sbemu can always work after windows exit.
IF windows does NOT have audio driver and after exiting to dos, sbemu is likely to work,
otherwise the audio driver may breaks thing and make SBEMU not working even when fresh-installed after windows exiting.

I've test with win3.x and win95 without sound driver and after exiting SBEMU works, on win98se with sound driver, it does work after Windows exits.

Please wait for the release of the recent changes or you can find it here: https://github.com/crazii/SBEMU/issues/183#is … ment-6908640147
it's currently being tested for a critical VDS bug that may corrupt the disk data of a IDE controller with PCI-Bus Master DMA enabled. So I don't put the release here for now.

Last edited by crazii on 2026-11-16, 06:32. Edited 4 times in total.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD

Reply 1997 of 1999, by Plombo

User metadata
Rank Newbie
Rank
Newbie
Plombo wrote on 2026-11-11, 04:04:

With VDPMI 0.34, Windows 3.11 starts up and seems to run fine...unless SBEMU has been loaded first. If I run SBEMU first, VDPMI blows up with an exception (see attached photo) when trying to start Windows.

An addendum to this: a similar error happens when replacing VDPMI with QEMM+HDPMI32. So it's probably an SBEMU issue rather than a VDPMI one.

Reply 1998 of 1999, by RayeR

User metadata
Rank Oldbie
Rank
Oldbie
crazii wrote:

I checked the source of HIMEMEX.SYS and RAMDSK64.COM, it needs a VxD for Windows and VDPMI works more like a Win9x env, so the VxD is needed but VDPMI doesn't support VxD.
HIMEMEX.SYS uses privilege instruction to set CR0, it's also possible to set another whitelist for it, and load it after VDPMI.EXE but I'm not sure it will work or not eventually.

How you mean ramdisk would need VXD? The original R.loew code don't use vxd, it's just extension of xms driver and ramdisk com file. Did you mean to make it compatible with VDPMI?

Then maybe the whole ramdisk could be rewriten as VXD and loaded in early init phase in system.ini - I think it should be possible before windows will create page file ( if VXD loaded soon enough it can provide ramdisk drive for paging file)

Also maybe other way - as VDPMI already has built-in XMM you may adopt HIMEMEX extension for 64b XMS. Or have a look at Japhets's himemsx - I guess it does the same thing - provide extended XMS over 4GB by switching to 64b mode. Japhet also set is as new standard, clearly described the API. You may rather implement this. Then r.loew ramdisk would be need to be adjusted for new API. In case that HIMEMEX doesn't do anything extra compare to HIMEMSX it should work...

I don't exactly know why Glidulator works with 4 cores (do APIC programming) and not under other VMMs,as its's dynamycally loaded ovl and I even didn't write any whitelist entry for it. Maybe it has something to do with dpmi and needs 1:1 mapping as I read somewhere elese, dunno, just stating it works...

BTW if there will be a plan to offload SBEMU to second core on SMP systems it would be good to coordinate cores allocation with Glidulator project and maybe some future projects that will use multicore. I can imagine it could be defined as some DPMI extension and provided by various DPMI servers for DOS application to avoid case where every app would manage cores its own way and one app will owerwrite worker thread of some other app...

Gigabyte GA-P67-DS3-B3, Core i7-2600K @4,5GHz, 8GB DDR3, 128GB SSD, GTX970(GF7900GT), SB Audigy + YMF724F + DreamBlaster combo + LPC2ISA

Reply 1999 of 1999, by crazii

User metadata
Rank Oldbie
Rank
Oldbie
RayeR wrote on Yesterday, 17:11:

How you mean ramdisk would need VXD? The original R.loew code don't use vxd, it's just extension of xms driver and ramdisk com file. Did you mean to make it compatible with VDPMI?

Then maybe the whole ramdisk could be rewriten as VXD and loaded in early init phase in system.ini - I think it should be possible before windows will create page file ( if VXD loaded soon enough it can provide ramdisk drive for paging file)

I found the RAMDSK64 2.0 here: https://rloewelectronics.com/distribute/RAMDISK/RAMDISK2.0/ which also have 1.0/1.2 there.
there is a PROTHOOK.VXD, and if you start RAMDSK64.COM it has a message that needs driver in windows. but it does not crash during my tests, there is a message with success but a Retry/Abort/Fail prompt when accessing the ramdisk. Are you using the 1.0 or the 1.2 version? I can test them while I was only testing 2.0.

I remember read/use himemsx with its API description some time ago. But currently it is difficult to add new features for the XMM in vdpmi, because of the real mode part is written in C, sharing the same code with the 32bit core but with a different compiler (wcc vs wcc386). and its code size bloats. it's already difficult to fit into a 4K resident memory with 96 EMB table entries, that's the result after I did some code size optimizations, it originally takes ~5KB only with code and without data. It is also possible that the 64bit support only exist in v86 monitor and being fully dropped when switching to real mode - but the EMB above 4G needs to be dropped too. or maybe like DOS8.0 the XMS driver keeps resident in himem and A20 permanently on.
all in all it's possible but needs extra work before starting this new feature. anyway VPMI can work with external XMM so I think its OK to use other XMS64 drivers, and it's much more easy to make it compatible with them.

RayeR wrote on Yesterday, 17:11:

BTW if there will be a plan to offload SBEMU to second core on SMP systems it would be good to coordinate cores allocation with Glidulator project and maybe some future projects that will use multicore. I can imagine it could be defined as some DPMI extension and provided by various DPMI servers for DOS application to avoid case where every app would manage cores its own way and one app will owerwrite worker thread of some other app...

Yes there might be, but I'm not clear about any details so I have no further ideas, but I'm open to this suggestion. for now there's only one thought, it should be centralized, put in a supervisor, like an EMM or DPMI server or its loadable components not in individual exe drivers. I might be wrong but I may have more thoughts if I actually started investigating on it.

Toshiba Satellite Pro 4300 - YMF744, Savage IX
Toshiba Satellite 2805-S501 - YMF754, GeForce 2Go
IBM Thinkpad A21p - CS4624, Mobility Radeon 128
main: Intel NUC11PHKi7C Phantom Canyon: i7-1165G7 RTX2060 64G 2T760PSDD