I found that not only CF cards vary in behavior, but also the adapters too. Try all cards you got vs. all adapters you got.. maybe you'll get lucky and find a winning combo.

Otherwise, you can also try different floppy controllers.. also double-check the CHS parameters you're using and try to get those from tools like IDEDIAG.EXE.

wbahnassi wrote on 2024-07-04, 07:20:

I found that not only CF cards vary in behavior, but also the adapters too. Try all cards you got vs. all adapters you got.. maybe you'll get lucky and find a winning combo.

Otherwise, you can also try different floppy controllers.. also double-check the CHS parameters you're using and try to get those from tools like IDEDIAG.EXE.

Alright, I guess Ill have to find or (more likely) buy some other CF cards, do you have any brands you'd personally recommend?

This happens with old, buggy BIOSes, and looking briefly at the one on theretroweb for this system, it appears to be affected.
The issue is reading from 1f7 (acknowledging the IDE interrupt) after transferring the sector into memory and not before. Because your CF card is much faster than the disk that came with this machine, the next sector is already waiting for transfer and the interrupt gets lost.
It would be best to use XT-IDE BIOS on an ethernet card and you will never have to deal with this again. It's possible you could get away with EZDrive or OnTrack too, so long as they don't attempt to use Int 13h to read more than one sector at a time.

KELP wrote on 2024-07-04, 07:35:

Alright, I guess Ill have to find or (more likely) buy some other CF cards, do you have any brands you'd personally recommend?

I have more luck with smaller-sized cards. People here seem to swear by those Cisco low-capacity cards as they are industry-grade. They should be pretty abundant as they were integrated in many Cisco hardware. Especially for a 386, I'd aim for a 256MB or even less capacity.

jakethompson1 wrote on 2024-07-04, 07:45:

This happens with old, buggy BIOSes, and looking briefly at the one on theretroweb for this system, it appears to be affected.
The issue is reading from 1f7 (acknowledging the IDE interrupt) after transferring the sector into memory and not before. Because your CF card is much faster than the disk that came with this machine, the next sector is already waiting for transfer and the interrupt gets lost.
It would be best to use XT-IDE BIOS on an ethernet card and you will never have to deal with this again. It's possible you could get away with EZDrive or OnTrack too, so long as they don't attempt to use Int 13h to read more than one sector at a time.

Alright, Ill look into getting XT-IDE setup in one way or the other

wbahnassi wrote on 2024-07-04, 08:22:

KELP wrote on 2024-07-04, 07:35:

Alright, I guess Ill have to find or (more likely) buy some other CF cards, do you have any brands you'd personally recommend?

I have more luck with smaller-sized cards. People here seem to swear by those Cisco low-capacity cards as they are industry-grade. They should be pretty abundant as they were integrated in many Cisco hardware. Especially for a 386, I'd aim for a 256MB or even less capacity.

Ill check out cisco CF cards if my 256mb card doesnt work after I find some method of getting XT-IDE

Your 256MB card will most likely work... XTIDE bios is required, and a boot sector on the CF card.

To get a boot sector on the CF card: Boot from a floppy disk into DOS. Run Fdisk, create DOS partition on CF card, set active. Run "Format C: /s".

I'd recommend official standard SanDisk cards (be careful of fakes/clones). I've used them with a variety of builds ranging from 286 to Pentium systems without issue.

I had this issue with a 386 as well (Phoenix Bios btw). I tried various cards from various manufacturers, including small stuff like 256MB, and putting in appropriate head and cylinder values for the sizes... At the end of the day nothing worked until I booted using the XTide bios as the bootrom on a network card, at which point everything worked! The various cf cards I'd been formatting and trying to make bootable all suddenly worked, confirming that I'd been setting them up correctly all along.

jakethompson1 wrote on 2024-07-04, 07:45:

This happens with old, buggy BIOSes, and looking briefly at the one on theretroweb for this system, it appears to be affected.
The issue is reading from 1f7 (acknowledging the IDE interrupt) after transferring the sector into memory and not before. Because your CF card is much faster than the disk that came with this machine, the next sector is already waiting for transfer and the interrupt gets lost.
It would be best to use XT-IDE BIOS on an ethernet card and you will never have to deal with this again. It's possible you could get away with EZDrive or OnTrack too, so long as they don't attempt to use Int 13h to read more than one sector at a time.

Might explain why I’m having problems with the russian 286 motherboard that I got. I tried XTide once, but I forgot that I had the 386l buil in there. I’ll have to try it with a new ROM.

dukeofurl wrote on 2024-07-04, 16:44:

I had this issue with a 386 as well (Phoenix Bios btw). I tried various cards from various manufacturers, including small stuff like 256MB, and putting in appropriate head and cylinder values for the sizes... At the end of the day nothing worked until I booted using the XTide bios as the bootrom on a network card, at which point everything worked! The various cf cards I'd been formatting and trying to make bootable all suddenly worked, confirming that I'd been setting them up correctly all along.

Good to know, this samsung deskmaster I have is also phoenix bios