Win3x.Org

Done !

Here's a picture of the motherboard jumpers on http://www.win3x.org/uh19/motherboard/show/6283



That's the best photo I was able to get with my Iphone.

Any idea what these mean?

I think they read:
JP2: 20M/16M (processor speed? changing the jumper has no effect on processor speed)
JP3: NOM/-6 (memory speeds? changing the jumper has no effect on memory latency or throughput)
JP4: FGA20/EXEGA20 (A20 Gate ????)

Maybe JP2 for the speed is made for when you swap the crystal to parameter the board properly ?
I mean maybe 20MHz instead of 16 puts the chipset or something else (ISA ?) out of spec and you have to select the CPU speed there to adjust properly the motherboard (waitstates, dividers, etc).

For example in order to have ISA running at 8MHz, you need a 2x divider at 16MHz, but you need a 2.5x divider at 20 (or else you get 10MHz ISA).

Try to benchmark your motherboard when parametered for 16MHz operation and 20 and compare. If it does what I think it does, your board should be slightly faster with the "16Mhz" setting instead of 20.

So if it was a ISA bus divider, we'd expect to see:
[*]one setting with a 2 divider
[*]the other setting with a 2.5 divider.
[*]25% faster with one setting

Ended up being a little harder than expected to test the disk response.
[*]Speedsys locks up when trying to do the disk benchmarks under dos on this board. I'd love to find the DR Dos for it, but I'm stuck with the BIOS that I got for now.
[*]Disktest.exe has too much variation. Sometimes it showed a 15% increase, other times no improvement. Don't like that tool.
[*]Win98se won't install on a 386sx for me, even with a math coprocessor & /nm switch, but Win95b installed.
[*]ATTO won't run under Win95b. Maybe I should have used Win98 (not SE). Took att day to install win95b, so it is what it is right now.
[*]Then I lucked out. Speedsys will run disk tests from a win95 dos prompt with only a 1% variation between runs. I see a pretty clear 4-6% increase in disk IO and 6-7% decrease in seek latency.
[*]So I see an improvement, but not as clear as I had hoped.

Then I remembered "yea olde" Landmark video benchmark. Got vividly clear, repeatable results:
[*]20M/16M in position 1-2 : 6425 chr/ms
[*]20M/16M in position 2-3 : 5093 chr/ms

That a 26% performance increase with the "20M/16M" jumper in position 1-2 over the jumper in the 2-3 position.

And that's right in line with the ISA bus divider that we expected to see.

Great find ! So yeah it's not here to set the board at a specific speed but rather configuring it to run properly depending on the speed.

What's weird to me though is why did they put a jumper there ? You can't swap the CPU and you can't swap the crystal so you're never supposed to tamper with it. They could have simply soldered a piece of wire depending of the components they installed ...
Oh well that's none of their problem now ^^