Additional: I also added headers to JP10 and JP11 which are right behind the cache chips but trying diffrent configurations of jumpers there also either doesn't help or makes the board not post properly or at all.

Assuming two TAGS, and one cache chip is right, there may be a third jumper in addition to J14, J15 to set.

Some of the other boards on retroweb with the same chipset have manuals, and have 3-4 jumper settings. The first two specify/multiplex 128, 256KB. The third specifies 512KB.

I would trace J32, to see if it is involved with the cache, and/or look for another solitary jumper.

MikeSG,

Thanks for responding. Why do you say two TAGS? The board definately only has the one 28-pin tag socket plus the 4 32-pin cache sockets.

I understand what you say about having more than the two jumpers though, but in looking through docs it seems to me that that most boards with 1 bank of cache like this one only have two jumpers and the third jumper is usually on dual-bank boards so you can select 256k as 4x 64kx8 or 8x 32kx8. Unless I'm wrong there.

Sorry I got mixed up about the two TAGs.

This board has four jumpers to select one bank (4x 32x8kb). https://theretroweb.com/motherboards/s/a-tren … -atc-1425a#docs

But the other Amaquest board (with manual) only has the two jumpers. Written on the board between first & second PCI slots. https://theretroweb.com/motherboard/image/ap8 … 73403689772.jpg

So the two jumpers may be right.

J32 appears to also be a keyboard lock when shorted.

Ah yes I see what you mean with the A-Trend board, I did think that the J10,J11 jumpers might serve that purpose but I've not been able to 'guess' any jumper config that works to get any cache at this point. I am looking at that other Amaquest board too, it looks like it could be an earlier predessor to this board, but the CPU slection jumpers are quite different sadly so not enough of a clue.

I've buzzed out the important pins on the cpu socket which vary between intel and cyrix and it looks like for intel writethrough none of them make any difference here. In fact the the board works including windows and PXE booting from a network card quite happily with only one single jumper present that is the voltdet pin. So I dont think the cpu selection jumpers here are what's upsetting the l2 cache here.

You could try putting the resistor networks back (or jamming them in the sockets with the 74F373 chips to test). High speed signals often need pullups/downs.

Already tried it. They fit find in the sockets on their own or as you say jamming in as well. Signals look okay. My scope isn't quick enough to show how clean the signals are at 33 or even 25mhz fsb but there's no obvious ringing or distortion of the waveforms.