I would look to use XTIDE BIOS extension.

You can install XTIDE onto a NIC that allows BIOS extensions or use an ISA card like this: https://texelec.com/product/isa-compactflash-adapter/

So in theory with that solution I wouldn't have to touch the 286 bios? I like that!

+1

Problem: BIOSes from the 80s do pre-date AT-Bus/IDE.
They can talk to IDE devices but think they're WD1003 compatible MFM/RLL controllers.

And with ATA-2 specs certain details began to change so much that old BIOSes begann to fail.
Since CF cards are from late 90s/2000s, they're ATA-5 to ATA-7 specced already.
https://www.os2museum.com/wp/how-to-please-wdctrl/

For testing, you can use floppy version of XT-IDE Universal BIOS (aka XUB).
XUBDisk - floppy disk XTIDE Universal BIOS booter

PS: I've used a network card with boot ROM socket.
An 2764 or 27128 EPROM burnt with a copy of XUB is all it needs.
By old XUB backups are set to auto-detect and ready to use:
Re: IDE disk-on-module + 386

AgustinCordes wrote on 2025-06-11, 22:31:

So in theory with that solution I wouldn't have to touch the 286 bios? I like that!

Yup, no need to worry about the existing 286 BIOS.

I use XTIDE BIOS extensions with all my 286 and XT/hybrid computers.

Jo22 wrote on 2025-06-11, 22:34:

For testing, you can use floppy version of XT-IDE Universal BIOS (aka XUB).
XUBDisk - floppy disk XTIDE Universal BIOS booter

Oh my gosh, that booter is exactly what I was looking for. Many thanks, I'll give it a try shortly and report back with my findings!

This all looks very promising, thanks both again.

If you find that XUB booter works for your circumstances, consider looking for an ISA NIC with a bootprom socket. These are ideal for doing dual purpose, in getting data into/out of vintage systems, AND providing a convenient place for XTIDE BIOS to live.

https://www.youtube.com/watch?v=ofZLDAOmZYU

AgustinCordes wrote on 2025-06-11, 21:58:

Weird things happen... the card IS recognized, albeit with the incorrect CHS setup. I can partition with FDISK, format, even copy files to the card, but copying stuff from the card, or executing, hangs the system. I'm positive the issue is this mismatch between what the bios can read and how the card performs.

This actually sounds like the order of operations bug in many 286 BIOSes. When making a request to read some number of sectors, a separate IRQ is generated for each sector. The BIOS is supposed to acknowledge the IRQ, then read 512 bytes, then wait for the next IRQ. Old BIOSes do it backward: they read 512 bytes, then read the status register (acknowledging the IRQ). IDE disks and CF cards are so fast, that the next sector is already available, and an IRQ generated, in between reading that 512th byte and checking the status register. So the IRQ for the 2nd sector is lost, and the system hangs. Because it only affects reads of more than one sector, and never writes, it is why you can fdisk, format, and even install DOS, yet when you first boot or try to execute anything, the system hangs. According to my AMI BIOS book, IDE disks require a BIOS date of 1990 or later, even though IDE is supposed to appear the same as the AT disk controller to software.

Also, speaking under correction here, but since you bring it up, the AT BIOS specifies the geometry of the disk into the controller on each boot. For IDE/CF disks, and with no LBA support, you can use any geometry you want, so long as you don't exceed the total size of the disk. Because the geometry is fictitious anyway, all you are doing is telling the drive firmware what values to put into the formula (HEADS*cylinder+head)*SECTORS+sector-1. The issue is that using some random one creates issues with moving the drive between machines that don't use the exact same geometry (due to C/H/S addresses hardcoded into the partition table and used by DOS to locate partitions), and because LBA support uses a special formula to convert a size in sectors into total C/H/S values. This is why a lot of old IDE disks don't have C/H/S values printed on the case, but once LBA became common, they did.

Another option is drive overlay software
EZ-Drive Dynamic Drive Overlay

I use it on my 486 but assume it works on a 286
but anyway I just mention it as an alternative, XTIDE is still the better option

No, most modern DDOs have a 486 min requirement.

there are older versions of the DDO software that can work on a 286, but they lack really high capacity support.

XT-IDE, either in a NIC, or on a dedicated controller (like a Promise IDE, if you want 16bit card, rather than accepting 8bit, from a lotec card) is the way to go.

wierd_w wrote on 2025-06-12, 03:16:

No, most modern DDOs have a 486 min requirement.

Not sure what a “modern DDO” would be.

I’ve installed EZ drive 9.09 on a 386sx with 8mb ram.

douglar wrote on 2025-06-12, 03:43:

wierd_w wrote on 2025-06-12, 03:16:

No, most modern DDOs have a 486 min requirement.

Not sure what a “modern DDO” would be.

I’ve installed EZ drive 9.09 on a 386sx with 8mb ram.

I am unsure where I read it from, honestly. I just remember reading it, that DDOs newer than around version 5 of DiskManager, need a 386 with at least 4mb of RAM. ("Modern", in as much as these were the lastest versions of this kind of software to hit the open world.)

The 4mb of ram requirement is for the installer, presumably.

I would be leery of trying it on a 286, because 286 cannot exit protected mode. It can only transition to unreal mode, (ab)using LOADALL. Having this happen immediately on boot may have any number of unwanted consequences.

It's worthy of an empirical experiment, but I still feel that properly compiled code executed in actual real mode, so the system stays in actual real mode until changed by the user running user software, is the best approach for such a system.

wierd_w wrote on 2025-06-12, 15:38:

I am unsure where I read it from, honestly. I just remember reading it, that DDOs newer than around version 5 of DiskManager, need a 386 with at least 4mb of RAM. ("Modern", in as much as these were the lastest versions of this kind of software to hit the open world.)

Some version of Ontrack Disk Manager have some surprisingly high requirements, yes. The last version that I've seen, 10.46, the one that does LBA48, requires two floppies and 16MB ram to install. OnTrack was fond of proprietary partition types too, so I've tended to avoid it.

I'm pretty sure I've used EZ Drive 9.0x on a 286 before, but I might not have installed it on a 286. I might have set up the CF on a 486.

XTide is probably the best way to go if you can put a ROM in your system, but as with drive overlay software, you still have to make sure you are not using a 386 build.

Particularly if you end up ROMing it on a network card (or even for the XTIDE floppy) you can build an XTIDE BIOS optimized for 286