The motherboard has three jumpers with labels that are not legible--

Any idea what these mean?

I think they read:
JP2: 20M/16M (20Mhz vs 16Mhz 386sx CPU? )
JP3: NOM/-6 (memory speed? )
JP4: FGA20/EXEGA20 ( A20 Gate ???? )

I tried changing the jumpers, but they don't seem to affect anything.

@Deksor suggested I try some IO benchmarks to see if JP2 was the ISA bus divider. I see Landmark video benchmark precisely changes between 6425 chr/ms & 5093 chr/ms (26%) when I change JP2.
That suggests that the board gives a 2.5x isa clock divisor (20MHz-->8Mhz) with pins 2-3 shorted and a 2x isa clock divisor (20MHz-->10Mhz) when pins 1-2 shorted. (Open gives the same results as 1-2 shorted)

Any ideas to look at for JP3?
Are there any benchmarks that can test A20 gate performance or compatibility changes for JP4?

For JP3, have you tried to benchmark the RAM with both settings ?
If you set JP4 to "FGA", what happens if you pull out the keyboard controller ?

Also for JP2 the supposed ISA divider change may not be the only thing it does, maybe it tweaks the RAM a little and some other obscure thing in the chipset itself ?

douglar wrote on 2021-01-18, 15:58:

Are there any benchmarks that can test A20 gate performance or compatibility changes for JP4?

Try loading HIMEM with /MACHINE=2 (selects IBM PS/2 type A20 handling). The common "Fast A20" implementation is a clone of the PS/2 mechanism, whereas "Slow A20" is the AT compatible mechanism. Nearly all AT (compatible) machines support the keyboard controller implementation (which you can force using /MACHINE=1). Even if a machine supports "Fast A20", "slow A20" usually works, too.

I checked the memory timings using CacheChk.exe and didn't see any changes when I changed JP3 or JP4.

I'll have to try out the HIMEM /MACHINE=2 tomorrow. Putting in a CR2023 batter holder w/ diode tonight.