kotel wrote on 2025-04-22, 10:57:

I have since replaced the blown MOSFETs (again) and the MOSFET drivers, and there's no more MOSFETs blowing up! Sadly Vcore is missing and the CPU is stuck in reset.....
I did also buy an ADP3166 by accident (meant to buy an 3168), but 3166 has an CROWBAR pin on pin 6 instead of VID5 like 3168 has. There's little info of what VID5 does on the 3168 besides "Voltage Identification DAC Input", so I have no idea if it can be replaced with "CROWBAR", which provides on-demand protection for the CPU.

Any ideas?

From ADP3166 datasheet:

The ADP3166 is a highly efficient, multiphase, synchronous buck switching regulator controller optimized for converting a 12 V main supply into the core supply voltage required by high performance AMD processors

So it's meant for AMD CPU, not Intel. I don't think it will work well with P4. ADP3166 was used on Asus K8V K8N models. The pin #6 wasn't connected there.

kotel wrote on 2025-04-22, 14:07:

After some troubleshooting I found the culprits, missing FB, FBRTN, PWRGD. FB and FBRTN are all connected to their respective ra […]
Show full quote

After some troubleshooting I found the culprits, missing FB, FBRTN, PWRGD. FB and FBRTN are all connected to their respective rails. Injecting FB makes PWRGD still stay low....
Injecting Vcore directly onto the coils makes the CPU get out of reset, but it's still not posting...

All I know is that the ADP3168 sees some unknown fault and halts the switching of the VRM and that's it.

Any ideas?

Sorry, no ideas. I suppose so many damages was inserted into the board during the experiments that the board is good only as a donor of ICH5.
Don't get me wrong but many-many i865 i875 boards work (with some workarounds) although they have only problem in USB part of chipset.

Huh. Well TBH, it should work with an celeron 2.6gHz SL9VV just for POST testing. Only would need to make VID4-1 be low and VID0 to be high in order to get the desired 1.525v for Vcore and lift the crowbar pin up (which results in no overvoltage protection).
I also did testing with the 3168 (before destroying pin 11) when the board was off and injecting 1.4v Vcore makes the FB, CSREF and PWRGD pull to the desired voltage. So there's still some hope.

I have a working D865GBF board (not installed in a system) which is not the same as yours but likely has a lot in common with your board. If you'd like some measurements for reference let me know and I'll see what I can do.