Do the boards rely on crystal oscillators to set the clock speed? If so, I would put in a socket then try increasing by one speed grade at a time, to see how far it can go. I wouldn'y jump straight from 25 to 40 MHz in one jump, as instability at 33 might mean you kill the chip at 40.

Don't want to put you off trying though. I would imagine that the later the date of the CPU, the greater the chance of it working at higher speeds, as the technology improves over time, even for a chip that is otherwise identical.

Yes, I was taking it for granted I might need to do some old school "clock-chipping" as overclocking used to be called because of needing to switch the oscillators.

I guess one could say that all AMD 386 are "late" in that they got held up in court by intel for a couple of years, by the time they could crank them out in quantity, I think the process they were doing them on should have been very high yielding. But yes, even more laterer is still even more gooderer. I think the ones with the MS Windows compatibility logo had to have been 93 or later.

_____________________________________________________________

Been trying to figure what board I actually had before. I recalled it as having jetkey KBC on the back edge, AMIBIOS between 2 of the slots, making it actually a 5 and a half slotter rather than a true 6, 5x16bit 1x8bit. 386 in the front right area of the board, 387 socket... and first one I found was
https://www.ultimateretro.net/en/motherboards/2497
Almost, too pale..
https://www.ultimateretro.net/en/motherboards/6813
Ah, darker green, warmer, why come no AMIBIOS?
Re: 80386 BIOS image collection
BINGO, that's exactly it I think... I recall some stupid number in the top left that late 20th Century search engines, lycos, infoseek, dogpile, etc wouldn't give me anything for, why isn't it on UR though? ... oh it is, it miscategorised as a DX..
https://www.ultimateretro.net/en/motherboards/6803
Those all look to be pretty much the same board though. Now you'll notice that they have CPU speed settings, yet a crystal of 80Mhz for the 40. I am having a hard time remembering if they did anything, whether I tried to set it at a slower speed.

So it's gotta have had an Ali M1217, how good was that? ...
https://thandor.net/object/623
Pretty freaking horrible by the looks of that... I feel like mine did a bit better, on Norton Sysinfo for instance it edged past that "Compaq 386/33" and I think I thought it was scoring even with a 486sx at 20... based on a 486 board I had with an SX25 in around the same time, might have had fake cache I guess, yah I'm positive it was a 386SX not a DX... Doom seemed playable at the time on HUD plus one level of frame... IDK if that's 2021 "playable" but it was "not quite chewing the edge of the desk in frustration" playable. IDK, you wonder if you remember wrong or there was some other reason it's so off. Harris 286-25 and 5x86 @ 2x60 performance data seems all in line with what I got/expected in the mid 90s from hardware I had, IDK why I'd be specifically way off about the 386.

Edit: Just occurring to me now that those settings might be for ISA bus divider, since they say AT clock next to them.

Well damn, haven't had a lot of time to look yet with the hols, but assumed those boards were "near the surface" as it were as I thought I'd seen them in the last 5 years, but not so far where I'd thought they'd be. I wanna say the actual ~8x6 I've got is a gemlight and the ~9x7 is a middle end, something like a chaintech or biostar, but nothing looks right on ultimate retro yet.