Colorful rtx 3070 NVVDD-NCP81610 issue

[frysu36]

hello everyone, I’m new here and apologize in advance if I did post in wrong section etc. currently I’m working on RTX3070 from colorful-igame advanced which has NCP81610 in cherge of NVVDD (vcore). Long story short, I connected to his SMBus with elmorlabs EVC2SE and dumped all regs. In register 0x23, status fault, I got bit 0 - UVP_fault and bit 3 - CLIM_fault. I already replaced NCP but it’s behaving the same way. When I toggle EN to GND (r22 10k on q5 pin3 side) I am able to start NVVDD and all other rails like FBVDD and PEX aslo starts no problem, GPU is pulling 1.5A from lab psu, all drMos warms up evenly and all seems to be normal when read 0x23 the errors are gone also 0x48 looks like reports correct Vout (~750mV). I checked all phases with oscilloscope and just ph8 amplitude spikes to ~15-20V the rest of phses looks pretty good. I don’t know where to go from there, I can replace drMos on PH8 but would like to avoid blindly replacing all elements in NVVDD circuit one by one. Worth to mention that I don’t have a correct schematic and bv for it, Im using some for gigabyte which contains NCP81610 for NVVDD just for reference. Im not sure maybe there’s some missing components which I can’t catch (R22 for NCP EN was ripped off) Anyway help advise will be appreciated.
thank you in Advance

what is also worth to mention I have Vref for like 9ms
and then is gone as on picture

[frysu36]

going down to the rabbit hole… I found out that I'm momentarily losing EN signal. this lead me to power monitoring IC and then as far as to 5V step-down converter MP1475 (after disconnecting PS_VMON_PGOOD. long story short, it has realy weird ripple on output for about ~200ms and then it’s just fine. this leds to 3V3_seq flickering like crazy as PGOOD are used to EN 3V3_seq. this condition is pretty short and after that all stuff starts working Im getting all voltage rails perfectly fine, windows detecting the GPU but with code 43 and mods test 2 gives me this stupid error `vbios invalid or rejected` 😞 anyway I would really like to fix this bad initial conditions for 5V first, connect back PS_VMON_PGOOD, and then focus on mods/mats results.

If anyone have any ideas what could cause this behaviour I will really appreciate. Side note, oscillographs from FNIRSI are before power monitoring IC ps_vmon_pgood was disconnected as I start to suspect that issue is related to it but no. last graph which is actually 5V output is from ANEG as for some reason FNIRSI didn’t want to cooperate below 20ms 🤷🏼♂️. From manual for MP1475 I got the info that if voltage drops to 90% PGOOD is pulled low and it’s purely below this value

[frysu36]

p1 p2 stands for PIN number of U29 AND GATE (inputs) also MP1475 was replaced already

[frysu36]

Here is actually signal from 5V PGOOD which enables 3V3_seq

[frysu36]

Card is fixed, it was my bench PSU which is MDP-P906 from Miniware. On pic are 3 scenarios: 12V with logic disconnect from us5650Q (U26 and U29), 12V when us5650Q logic is in place and 12V from my PC PSU during initial power on. I don’t know why NCP81610 acts this way and why it’s able to boot when its getting multiple EN (when us5650Q is disconnect) maybe playing around boot ramp time for NCP will help but what’s the point? I guess that I have to look for new bench PSU. Please note that probe for 12V was directly attached/connected to MDP output to rule out all potential voltages drops