I have a question for you Dave..
If anyone would know the answer I suspect it would be you.
On the VCF forum I asked if the actual physical head width on a genuine 720k floppy drive is larger/wider than the heads in a standard 1.44mb floppy drive.
I have long suspected it is..but was assured in wasn't the case.
Not that it really matters unless the head width is different...in which case erasing ..reformatting disks for use in a different computer might be important to avoid remnants of old data remaining.
Believing this was the case I have both a genuine 720k PC drive from a zenith supersport laptop ...and a standard 1.44mb floppy in this build.
Overkill maybe..but I'd rather know for sure.
cheers

So..took about 25 minutes but just running with no activity at all from me..
The system crashed in the same way...flashing cursor and no response to key presses.
I guess it is time to remove all the cards except the video card and the quadflop to see what happens.

@DaveDDS

Hey Dave - lots of great questions!

The original Shugart specification supported four floppy drives, on one cable. Unfortunately, IBM broke four drive support with their introduction of the twist in the cable.

Many legacy floppy controllers support four drives. In fact, the original IBM XT could support this, when implemented. In most cases, you need a floppy controller with two connectors (drives 1 & 2 and 3 & 4). This can all be done with the same base address, IRQ and DMA. This is how the Quad Flop is configured.

You can also support greater than two floppies by using multiple FDCs. This is how my machine is configured. I'm using my motherboard's onboard FDC as primary and add-on card for secondary. While my add-on FDC supports non-standard IRQs and DMAs for secondary FDC, I'm using IRQ 6 and DMA 2, for compability with the MRBIOS BIOS.

Since the motherboard's BIOS natively supports four floppy drives, no drivers are necessary. But, I did have to patch Win9x to support the recognition of four floppy drives.

Both of our FDCs support BIOS extension. But, I was not having luck with its use in Win9x. I currently have my FDC's BIOS extension disabled.

-Brian

Vipersan wrote on 2025-12-09, 11:48:

... if the actual physical head width on a genuine 720k floppy drive is larger/wider than the heads in a standard 1.44mb floppy drive ...

I don't think this is "officially" the case for 3.5" drives.

It is for 5.25" - head width is related to total number of tracks. 5.25" 350k disks have 40 tracks, 1.2 have 80 tracks - to make 80 fit where there was originally 40, the head had to be thinner (and double-stepped for 40 track media).

This is the primary reason that re-writing an already formatted/written 360k disk on a 1.2m drive can lead to problems - The thinner head on the 1.2m doesn't cover all of the original track width, and "edges" of the original 360k tracks can remain which don't affect reading on 1.2m, but can cause conflicts in magnetic domains if read on a 360k.

3.5" drives : 720k or 1.44m both have 80 tracks, and therefore use the same width heads.

But... the different media uses a different magnetic coercitivity (strength) which means the heads (and driving electronics) ARE "different" and depending on how well the drive is designed, DD may not be written quite the same way as it would have been on a 720k... Not normally a problem (I've never encountered it) - just something to be aware of.

For years, my company (DDS) sold software on floppy disks - so I ran an auto-duplicator a LOT - and I always installed the right type of drive for the media I was using in that run.

Btw, the reason I started with "officially" - again, depends on drive quality and manufacturing tolerances - some heads are slightly wider, some slightly narrower. In my experience, never enough to make a detectable difference - but there are always the "one of" cases. Drive alignment can affect this as well.

bbuchholtz wrote on 2025-12-09, 12:47:

The original Shugart specification supported four floppy drives, on one cable. Unfortunately, IBM broke four drive support with their introduction of the twist in the cable.

Indeed - I've been using floppy drives since my Altair with original SA-400 drives - IIRC the first 5.25" (35 track SD - a whopping 90k per disk)
It was kinda necessary to have as many drives as you could (and the NorthStar single-density FDC only had three selects implemented)

Many legacy floppy controllers support four drives. In fact, the original IBM XT could support this, when implemented. In most cases, you need a floppy controller with two connectors ..

Ahh... It's been so long I'd forgotten that some systems had two connectors for 3&4 .. I never liked that IBM chose to mess up the selects.
I guess they thought customers are too stupid to move a drive select jumper, and didn't want their dealers to have to carry (and stock) two different
drives (A: and B:) - And being originally "floppy" only they wanted to run only the drive being accessed - which is why they corrupted the cable,
making it so ALL drives had to be jumpered as drive-2 and bringing out separate "Motor-ON" signals for each drive (the original Sughart 34pin
floppy connector spec. had only a single "Motor-ON" for all drives)

You can also support greater than two floppies by using multiple FDCs. ...

Also a good point - I have many times added an extra FDC - but always to replace the limited internal FDC with a more capable one which could better handle non-PC formats.
I don't think I ever tried to get two FDC working at once.

Since the motherboard's BIOS natively supports four floppy drives ...

I don't recall seeing a BIOS that supported 4 drives - but I did go through a lot of systems over the years, and since my main need of floppies (ImageDisk) doesn't go through BIOS to access the drive, I doubt I really looked that close - and of course many early systems using FDCs on cards (like the XT) didn't have a BIOS at all.

And lets not even talk about newer multi-thread OSs which don't use real-mode BIOS ... since the PC FDC wasn't consistent in how you could make more than two drives accessible, you're into a world of "special drivers" for something few people wanted.

Its been well over an hour now since I pulled the sound card ...the scsi card....the network card....and the cd cable leaving all four floppy drives and obviously the quad-flop card plus the banshee vga in there.
So far no crashes.
...and I got to process another 50 3.5 inch floppy disks..
it certainly seems more stable.
I'll give it another hour ...then try making another DOS 6.22 installation on a new CF card.
If that goes well...I'll reintroduce the pulled items one at a time to see if/when the problem returns.
I certainly hope there is a resolution as MRBIOS is certainly very versatile and I would like to carry on using it.

Vipersan wrote on 2025-12-09, 14:20:

Its been well over an hour now since I pulled the sound card ...the scsi card....the network card....and the cd cable leaving all four floppy drives and obviously the quad-flop card plus the banshee vga in there.
So far no crashes.

By "leaving all four floppy drives" do you mean that you've pulled the hard drive and CD drive as well as removing the cards?

Crashing problems that "go away" when the system is more lightly loaded could also be indicative of power issues. Might be worth trying a known good supply.

Dave - checkout this link:
https://www.dropbox.com/scl/fo/xom2lc3ubwgrhf … t=wksvmhqh&dl=0

It's a repository of MRBIOS BIOSes. They were a third party developer that made BIOSes for many popular motherboards. One of the interesting things they did was add four floppy drive support to many motherboards.

-Brian

What's of particular interest to me is this:

1) where does MRBIOS save the parameters for the extra floppy drives in NVRAM
2) where does MRBIOS store the scratch pad variables for the extra floppy drives in the BDA?

Since the modified win9x drivers only seem to work with MRBIOS and not Sergey's expansion ROM, this would seem to suggest they way they respectively handle these two factors is what makes the difference in compatibility. I say this because win9x's protected mode floppy driver doesn't actually use the int13h handler whatsoever, but it DOES rely on reading CMOS parameters to ascertain drive types (and possibly does use BDA variables for something).

If MRBIOS is the de facto "standard" for 4 floppy drives in an OEM BIOS, (since AFAIK they were the only ones to do it) it would be helpful to know what memory locations they chose for these things, particularly to efforts like GLaDISK

No Dave...I left the hard drive (CF card in this case) in system along with the 4 floppy drives...
But sad to say it did eventually crash in the same way...
So that's the the cards eliminated...along with the CD drive.
It does look like I'm not going to be able to use MRBIOS...or at least ..this version with this motherboard.
The only thing left to try is to replace the CF card with spinning rust.
apart from that there is little I can do.
Gutted it is I am 🙁

Have you completely ruled out any other hardware problem? With original BIOS and standard FDC - all other cards in does it still "eventually crash"?

I'd also try spinning rust in this case, get it back to as "original" a configuration as you can, and see if it still dies.

Then, add things back one by one till failures happen again.

I'd prob. try pulling the quad-flop first, see if MRBIOS works in a more normal config...

Good advice Dave..
fwiw ..Yes no crashes at all with original bios...
..but that's not so say there isn't an incompatibility with MR bios and my video card ...or indeed CF cards...
At the moment I'm eliminating ram...by trying other sticks of 128mb or less
Problem is that the 'crash' is unpredictable ...anything from 20 minutes to an hour ..
Elimination is a slow process...but I'll try everything I can before giving up and returning to the original bios.
As Brian said..it works fine with his motherboard which although different uses the same chipset.
but this version of MR bios wasn't written with my mobo in mind.
Very much an experiment.

So ...I've downloaded the manual for MrBios..
Its cool that drivers for PS2 mouse and CD drive etc are installed by default upon detection..
I was looking for a way to disable the on board floppy interface of the mobo.
There doesn't appear to be a one stop solution for this...but perhaps setting all 4x floppy drive options to "none" and enabling the quad flop bios ....thus letting it take charge will suffice.
It occurred to me that the onboard floppy controller and the quadflops controller were at odds...and fighting for supremacy..
Worth a shot ?
After all it does say on Texelecs website >>>You may want to disable this if you have a system with an on-motherboard floppy controller, and you want BIOS support. However, it would be better to get an ISA ROM controller board as no other features of the card will function.<<<<

Ok ...I removed the quadflop card ...so it can't play a part ..
obviously this meant I could only add 2 floppy drives A and B....using the on board controller and A and B now using Mr Bios.
Still it locked up after about 10 minutes this time.
The only things left now are the banshee video card ...and the Compact flash hard drive.
I will now remove all floppy drives...set A and B to none..just leaving the CF card to boot the system.
If it still locks up I'll try and find a small capacity IDE hard drive and try and install DOS on it.
I'm running out of ideas now.
It is looking like the MrBios is simply incompatible with this motherboard....but not massively so...but beyond my capabilities to fix.
doh !!

maxtherabbit wrote on 2025-12-09, 15:07:

What's of particular interest to me is this:

1) where does MRBIOS save the parameters for the extra floppy drives in NVRAM
2) where does MRBIOS store the scratch pad variables for the extra floppy drives in the BDA?

This is exactly my interest. While the MRBIOS BIOSes are great, they end at 430TX. I would love to get this working with 440BX, and beyond...

If you have any ideas of where I should try poking around, be happy to test...

-Brian

Is this the best way to dump/analyze NVRAM:
https://manpages.ubuntu.com/manpages/xenial/e … m.4freebsd.html

Not familiar with BDA. Any recommended tools or approaches?

-Brian

I think I am close to abandoning MR bios.
I have little left to try..
Having removed all the cards except the Banshee.
I have only the basic motherboard with a 64mb strip of ram....a 6gb spinning rust hard drive holding DOS 6.22 which was built on a different PC.
No Floppy drives attached at all hence the reason the OS was built on a different PC...just the hard drice...an AT keyboard...PS2 mouse and the pci video card.
I'd call that bare metal.
I am right back to the first attempt at installing DOS on a CF card...where it crashed and hung after reboot.
...an unresponsive keyboard
DOS attempts to start and fails/hangs @ HIMEM is testing extended memory
There is basically nothing left to try except maybe a diffrent VGA card ..

OUt of curiosity Brian ..what is the motherboard you are using Brian...?
Maybe I could try and find one ...

Change A20 control method?

Either in the bios, or on the himem.sys line?

This is a k6-III, from your screenies... so the defaults should totally be fine, but we ARE doing odd stuff with the disk controllers.

Just as a diagnostic step, try using the 'keyboard' A20 gate control method? (The default on pentium and newer is 'Fast', but 'historically' the A20 line was asserted by the AT keyboard controller. 'Fast' tells the CPU / chipset to do it. There are quite a few options for A20 control. The fact it's hanging on himem.sys suggests that the hang is related to entering protected mode, or is related to memory access above 1mb. Both of those could be issue with A20. It could also be that the int13 handler code (which needs to be there for floppy C and floppy D) cant tolerate entering protected mode for some reason. Eliminating 'maybe sus A20?' Is a diagnostic step. Keyboard method A20 does NOT put the cpu in protected mode, because the KBD controller is doing the line management.)

Related light reading on A20
https://en.wikipedia.org/wiki/A20_line

And msdosdrv.txt, which has the himem.sys doc you need.

https://www.betaarchive.com/wiki/index.php?ti … 46/msdosdrv.txt

Cheers Mr Wierd...I'll look into the things suggested.
All this to get MrBios in the system.
Having had a brief taste of it..I've become fanatical 🤣

bbuchholtz wrote on Yesterday, 01:52:

Is this the best way to dump/analyze NVRAM: https://manpages.ubuntu.com/manpages/xenial/e … m.4freebsd.html […]
Show full quote

Is this the best way to dump/analyze NVRAM:
https://manpages.ubuntu.com/manpages/xenial/e … m.4freebsd.html

Not familiar with BDA. Any recommended tools or approaches?

-Brian

IMO the best way to dive into it if you're really trying to figure out both aspects would be using an emulator (as ICD)

Maybe setup an 86 box instance based around a generic 286 and use the generic 286 MRBIOS. Then you could use the built in snooper-debugger to view the contents of both NVRAM and the BDA while the system is running. Observe how NVRAM changes after changing the disk drive config in SETUP and how the BDA changes while accessing a floppy disk on drive 3/4 using int13h.

The BDA is located in base memory at 40:00, right above the IVT and stack.

https://stanislavs.org/helppc/bios_data_area.html