Memory says it's dual p/n-fet, for switching the RAM power supply from the main 3.3V to a 3.3V supplied by the +5VSB for when using sleep mode. I know that's not much help just now, but if no one has a chance to have a look then it should be possible to work out a component that'll work.

Is there a pad for pin 2 missing?

[edit: the equivalent on a ka7 is this: https://www.vishay.com/docs/70717/70717.pdf
and here's a picture from the ka7:

]

[oh, and if possible could you let me know the marking on Q24? It looks like I didn't put mine back when I was fiddling trying to track down a RAM fault]

Thank you very much!
IC is ordered and here is the picture of Q24. It's labeled with 702v5. Is it an N channel MosFET?

...and yes, the pad is missing. But I think I can fix that quite well.

This is from a KT7A-RAID v1.1:

looks like it's probably the same part snufkin found on a KA7, but I don't know what all the markings mean. The top line matches at least.

I tried to take the picture through a sealed bag, so if you need a clearer picture, or something else from the board let me know. I'll be happy to open it if needed.

Thank you shamino that is quite sufficient!
I have already ordered a replacement. I'll get back as soon as the IC has been soldered!

I think the top line is the important bit:
https://www.vishay.com/docs/72762/partmarkinfoso8.pdf

So it has:
Part Number
Lot Code
Factory/Year/Week/Wafer

Thanks for checking that Abit hadn't changed that bit of their design going from the KA7 to the KT7. Thanks also to Theseim for checking on that Q24, it does look like it's an n-fet. I'll need to double check that it was originally fitted (very likely was), then refit one. Board works without it, although I guess sleep mode wouldn't. Maybe I ought to test that first.

As you said snufkin, it was actually not the FET at the Ram slots, but a torn resistor next to the clock IC! But either way everything is done and the board is now running again!

However, I'm still not completely worry-free... After some time, so about an hour of operation, mouse and keyboard say goodbye at the PS/2 port. First mouse and keyboard have dropouts, until they don't work at all anymore. The system is otherwise stable. A micro crack?

theseim wrote on 2023-10-29, 09:33:

As you said snufkin, it was actually not the FET at the Ram slots, but a torn resistor next to the clock IC! But either way everything is done and the board is now running again!

However, I'm still not completely worry-free... After some time, so about an hour of operation, mouse and keyboard say goodbye at the PS/2 port. First mouse and keyboard have dropouts, until they don't work at all anymore. The system is otherwise stable. A micro crack?

Sounds a lot like my FIC 486 board, whose keyboard controller would drop out and die. In that case I noticed I could trigger the problem by touching the board. After reflowing solder joints at the KBC and the keyboard DIN, the problem was gone.

This is a later era of board, but I think it might still have a discrete IC for the keyboard/mouse/serial/whateverElse functions.

Unfortunately not, it is all included in the south bridge. But BGA reflowing is always a thing… Will try to check the lanes and passives first.

Any chance it could be fuse related? There are a couple of what look like PTC fuses by the PS/2 ports that will trip if they warm up, then reset as they cool down. I assume they're either on the +5 or Gnd for the two ports. Might be worth a bit of testing and pointing a fan at them?

The fuses seem to be OK. I checked the lines with an oscilloscope and found out that the data line drops from a clean 5 V to a noisy 1.6 V after a few minutes. However, the clock line remains clean. I suspect a capacitor array to ground.

Clock and Data should both be pulled to +5 through a resistor. So perhaps noise on the data line is the resistor becoming disconnected, leaving it either floating or connected to Ground when it's being driven. Might be worth checking the solder joints on the couple of resistor packs near the PS/2 ports.

So, I have now tested a little more extensively. And the reason seems to be completely different. It turns out that the system runs 100% stable and there are no keyboard or mouse dropouts as long as the fsb is running at 100 MHz. If I set the board to 133 MHz fsb, I get mouse and keyboard dropouts after 30 minutes at the latest. The active Northbridge cooler works perfectly and I have also applied fresh thermal paste to it, so I would rule out a heat problem here.
However, as described above, I resoldered a missing resistor. It was R49 near the clock generator IC. Did I perhaps solder the wrong value? As it appears to be a pull-up resistor, I have installed a 10K (ID 103). Is this a different one on your boards?

I've only got the KA7, so even though the RAM voltage supply was the same, I expect the clocking will be different. I'll try and have a look mid-week.

I still suspect something is over-heating, but that's just a guess. I don't know, but are the PS/2 ports driven from the Southbridge?

Thank you and yes, theo are. I also no longer rule out the CPU . I have a TBred mobile here. I'll try it with a Thunderbird and see if anything changes…

KA7 layout is a bit different, but there's a 10k pull up/down in a nearby location.

Thank you very much! Then I assume that the 10k is correct here.

Sorry it took so long for me to respond - I got my board out today (KT7A-RAID v1.1)
I can confirm the above - the resistor you circled at R49 is marked "103" on my board, so should be 10K ohms.