T. Ment wrote:
Now that I've fixed its broken solder joint, and I know the card works
in linux, I can verify the DOS flasher DOES NOT work. It gets far enough
to trash the chip's BIOS and then fails, replacing the card with a brick
that hangs the computer at boot.

I doubt it works with the Windows flasher either. Other sil3114 cards
may flash, but not this one. It has a r5403 BIOS soldered on. Take it or
leave it.

I found the Lattice archive for legacy SIL cards:

https://www.latticesemi.com/Support/ASSPSoftwareArchive

There you can find a 3512 BIOS, version 4.3.70. It seems unique, they
call it an "IDE" BIOS. I flashed it onto my 3114 card, and it boots.

It also works in DOS. I tried two later versions of the 3512 BIOS. Both
boot, but like the 3114 BIOS, fail on a file copy. So there's something
special about the 4.3.70 BIOS that lets it work in DOS.

They call it an "IDE" BIOS. Maybe that means IDE vs AHCI. Don't know.

Though it worked with a DOS file copy, I could not get it to boot from
the SATA drive, in the computer I had in mind for it. Maybe the BIOS is
too old and the hard drive is too big. Or maybe it's incompatible with
the motherboard BIOS. Don't know.

Looks like they gave up supporting DOS after that version. The archive
has no such BIOS for a 3114 card.

OK, and was there a "max disk size for MSDOS" ?

It can't go over 28 bit LBA or 137GB.
A 120GB SATA would work, a 160GB might be too much
(and especially if a partition "spanned" the boundary,
that's supposed to be bad).

On the old IDE drives, you could use the CLIP jumper
to clip to 33.7GB or 2GB or so. The geometry was the
same in each case, and it was the interpretation of
the OS at the time, that decided what that was. The CLIP
causes the drive to declare a magical CHS value.

I don't think SATA drives have the clip jumper, because
of course, the SATA era is 48 bit LBA (double pumped registers).
It's doubtful anybody gave a rats ass about DOS.

You can also clip a drive down with a Host Protected Area (HPA).
My other machine, the BIOS simply doesn't allow those operations
(they're locked out). On this machine, only the Jmicron port
allows setting an HPA.

*******

This is the proposal to move from 28 bit to 48 bit.
While some parts of the ecosystem might handle this,
I don't know if the whole thing (DOS part) does.
DOS was likely too late to the party for this.
DOS could be generating 28 bit addresses only
(good for the beginning of the disk... maybe).

https://web.archive.org/web/20041024...l/e00101r6.pdf

There used to be web pages that would explain every limit,
all the way up to 2.2TB. But I doubt I could find those
pages today.

Paul