ASUS ROG G15 G532LWS R1.3 G512LW-AZ003T Liquid Damage Repair

[tektechnician]

Hi,

I have a laptop an G512LW-AZ003T serial number LANRKD004568428 with what I presume is a i7-10750h and a RTX 2070 with the motherboard code G532LWS.
I bought this laptop secondhand so I don't know much about it's history but I'm certain it's had soft drink spilled into the vent area above/behind the screen.
I'm not a repair professional so I'm merely attempting this as a hobby but I'm seeking advice since i'm a little stumped at this current point in time.
The laptop was in very rough condition and I'm certain it was abused so here was what was wrong with it:

1. Black tar corrosion around UP9612R (U4602) GPU phase controller and pretty much all the filtering capacitor pairs 0805 25V 10UF above the 5030SG PQH810x/PQL810x and 5018SG cpu and GPU mosfets
2. Deadshorts through most of those filtering capacitors and also 4/6 5030SG CPU mosfets
3. broken DC jack
4. Dead jack detection circuit at the AC SYS power mosfet PPKCH2BB

I initially removed all of the shorted components and fed 19.5V power into the PRS8901 current shunt resistor and found that the charging circuit was ok, 5V, 3.3V, 10.2 (12V), 1.05V, 1.24V, 2.5V RAM, GPU 0.81V , VCCSA were all present with the GPU powering up. No fan spin, keyboard rgb.

The laptop responded to the power button but only the white FLY led turned on, long press of 30-40 seconds causes the system to reset. The motherboard also automatically turns on when the battery and AC power are plugged in.

Obviously there was no VCCCore/VCCGT so I've replaced all the filtering capacitors, 5030SG Mosfets, NCP81161 mosfet drivers and I also replaced the CPU power controller NCP81266 (PU8001).
I've also fixed the jack detection circuit so that is OK now.

I've abbreviated some details but basically after I installed those components, I still didn't have any CPU rail power until I replaced the VCCSA controller NCP81210 (PU8105) as well.
I don't have VCCGT, but VCCCore is present at 0.9V and the 12V rail is up from 10.2V. I read somewhere on the forum that VCCGT is not necessary but I don't know how true that is for this situation.

The laptop now powers up with all rgb, fan spin and the status led changes from FLY to OS however I still don't get image.
The laptop will power up for about a minute, then it will enter a cycle of power resets. If there's no ram installed then it remains static and constantly on. I've tested the 16gb of ram that came with the laptop and it functions fine in another system so I don't think it's that. I've got no evidence to suspect that the CPU is fried, but it feels like it's becoming more likely, atleast something major like pch, cpu, EC etc. I bought the laptop for $100 AUD so this repair job was only for the curiosity.

I've dumped the bios as a last resort using a RPI4b and flashrom and been trying to work through building a fresh version from an asus update or clean the DMI region of my own using the guides found on this forum,
however I don't really know where in the BIOS file the DMI region is located for CSME 14.0 and the digital signing process seems very hard for me to comprehend.
I've found that the bios dump has CSME 14.0.33.1125 Consumer H CNP/CMP-H A which is only extract form from the CSME repository, which version do I try to use? 14.0.32.1123_CON_H_A_PRD_RGN.bin?

Does anyone have any suggestions for what voltages or signals I should be looking to isolate the issue preventing post?
Apart from a bios flash success, I think I might call this laptop a write off if it requires CPU/GPU/PCH replacement since I don't really have the skillset/equipment to handle BGA properly.
I've attached some images and the bios dump.

Hopefully this is in the right section, if not I apologize in advance.

[john49884]

It seems almost fixed to me. Usually, if you have VCORE, your BIOS is probably fine and you are only missing GT voltage which is the integrated graphics and the very last rail to be enabled. The driver of VCCGT shares the enable signal with VCCCORE so it is worth to inspect PU8106 and all its components. It is still worth to reflash a clean BIOS for good measure; there is a dedicated BIOS request sub forum in the laptop forum where you can ask for help.

[tektechnician]

Hi thanks for the advice.I replaced PU8106 and additionally replaced PU8001 NCP81266 and spent some time desoldering and measuring all the resistors and capacitors related to VCCGT. The driver of VCCGT had 5V VCC, BST and enable but the PWM pin was reading 1.81V whereas all the other VCORE was around 0.21V. I pretty much gave up and reassembled the laptop and somewhere in the process the laptop now turns on to bios! It even let me read from USB and update it to version 314.
However I think there is still an issue with the laptop as it cannot boot into windows or linux. It just freezes before I would expect to see the loading circle or black screen after grub loader for ubuntu. I can't use the windows 11/10 installer either or Hirens PE boot tool. I can boot into memtest86 and run a full test with all 6 cores and 16gb memory and that passed, however my experience with another computer is that memtest86 will pass even with a dead core and only works for real memory issues like connection or faulty ram.

I suspect a CPU/power issue given that linux won't boot, but I will remove the UP9512R I replaced to disable the GPU just in case. How would you recommend I go about diagnosing this?
I don't own a oscilloscope and do not understand what the typical boot voltage profile is like for this CPU. Is it possible to measure each phase output with a multi meter?

[tektechnician]

Here's the photos of the laptop posting and running memtest86 successfully, but no boot with windows 10,11, installers and ubuntu. I know for desktop motherboards they can let you disable specific CPU cores in the BIOS, does laptops have the same ability if it were to be unlocked?

[john49884]

Those 76° C were immediately after the memtest? And yes, a failing phase could be causing your problem. Without an oscilloscope I would first monitor VCCCORE (in the output of the coils) to se if it is pulled low during boot. If that's the case you can suspect of one or more of the phases an then monitor during boot each coil at their input. The same with VCCGT.

Another options could be the thermal sensors which are near the spill zone or a false over current, but for now I think that's unlikely because in that case the laptop would shutdown suddenly. You also need to be sure that the charger is 100% in good condition and it is very useful to know how much current the laptop is taking if you do this in the future again.

Additionally I would recommend a simple MATS test to check the GPU's health.

I just saw the liquid near the gpu. did you check the gpu voltages? I don't have the exact boardview, but I think there was a thermistor under the stain just between the two dual mosfets below the gpu pic 3 of your original post.

[Edit] Now I have boardview. And it not a thermistor but a 1 Ohm resistor. probably cooked.

[tektechnician]

VCCCore seems ok, I replaced the 5030SG mosfets again and also tested each one individually by isolating the coils and powering up. Also tested different NCP81266 and NCP 81210 with no change in behaviour. Regarding the 1ohm resistor PR9422 it reads nominal so no issue with that. I can't seem to boot MATS or truthfully any other live disk OS apart from memtest86. I think honestly the CPU might be bad and maybe I've killed it somewhere along the way of diagnostics. I did inject 0.9v into the VCCCcore rail but that's within the intel specifications. Temperature was ok, the memtest took maybe 2 hours and I had covered the screen since it was running overnight.

I suppose the reality is, the laptop was pretty badly liquid damaged and the circuitry for the CPU was what completely shorted (4/6 mosfets damaged). It would not surprise me if the CPU got slapped with the full 19.5V to the cores for a short period of time. I've included a voltage plot of the VCCCcore during startup, initial blip is right after the power button press, the second broad stable voltage is blackscreen memory training or cold power up without battery, and then the slightly noisy lower voltage is the CPU playing the ROG logo sound and sitting frozen on the windows boot stage until I power it off.

Suppose if I were to attempt a CPU replacement, how difficult is that process regarding dealing with CSME clearing and the difficulties of big BGA with a heatgun and a 120c heated silicone mat?
I've seen videos of Macbooks being done in that manner (probably with a heater at 200c) but I've never personally tried.



GPU +NVVDD is fairly similar although it has very little activity occurring. Probably doesn't help that the CPU is likely braindead.


VCCSA seems relatively normal

[john49884]

At this point you are limited by your tools more than anything. While a damaged CPU is a possibility, replacing it should the last thing you do because you don't know for certain if your phases are noisy or not. For example, In the pic 1 of post #1 I see corrosion in the CPU's filtering caps, may or may not be that, could be noise in the pch/cpu datalines, so it is really difficult to try guess at this point, and I won't tell you to start chasing signals, because an oscilloscope is needed for that. These are my last recommendations considering your tools

Check the thermistors (Unlikely like I wrote before, but if it is worth you time, do it); Remove PU9001 and measure all components that start with PTR, must have ~100k Ohm at room temperature. Also all the shunt resistors which start with PRS.

Try to boot on battery only.

Flash a clean ME BIOS. Check your backup with UEFI tool to begin, and in the intel FIT you can disable turbo, disable cores among other things. in winraid forums read the Plutomaniac's guide; it covers absolutely everything that you need to know to compile it.

And regarding swapping the CPU, check if you have bootguard enabled first.

[tektechnician]

Hi, thanks again.

Yeah I understand what you mean, without an oscilloscope it's basically speculation at this point.
I'll see if I can gain access to one, but it might not be in the short term. I did remove and replace all of the filtering capacitors you've mentioned originally, since as I would do rework around the area, they would sequentially fail to short so I decided to replace them all.
I'll check the PTR across the board and let you know if I find any issues, but I did buzz out all resistors, capacitors and thermistors that are related to VCCGT drive & feedback and they were ok. Battery boot does not yield any difference.

Additionally I tried a clean ME BIOS and tried disabling turbo, active core count and max frequency at boot, but strangely the motherboard/CPU didn't obey the settings if that makes sense. Memtest86 reported all 6 cores & threads running during testing.
The clean ME BIOS with modified settings was definitely different to the cleaned original dump as I could see the exact byte change in HxD, additionally I am certain I've flashed the bios chip successfully as it reverted from 314 -> 311.

I'm not sure how to correctly check if bootguard is enabled, but when building the cleaned BIOS in Intel FIT, it says it is configured as bootguard profile 5 FVME which I believe is verified boot and TPM?
Also I tried configuring a windows 10 build to use single core and to dump a boot log and all it reads is BOOTLOG_NOT_LOADED \SystemRoot\System32\drivers\dxgknl.sys which points to an issue with the direct x kernel (graphics)?

Nothing else is in the boot log which makes me think it could be possible that VCCGT is still bad. eg CPU calls for the iGPU and it crashes because VCCGT rail never comes up during OS boot.
Where and when should I expect VCCGT rail to come up during a boot process? Is it something the CPU calls on demand or should it go up as soon as an OS boots for example?

I hate screenshots on small text, but I'll include a image of the components around NCP81266 that I checked highlighted in yellow. I checked all the resistor straps and they were fine. Additionally all the resistors, capacitors and thermistors for the VCCGT part of NCP81266 measured fine.

[john49884]

I was expecting a level 0 bootguard profile for a gaming laptop. You will have to revert to your original dump; you can't change cpu straps or anything if remember correctly. If any of the BIOS experts is reading this please comment about it.

My theory was that it was freezing when the cpu boosted to turbo and disabling it would hopefully allow a boot. There is a method to modify the nvram without modding the bios (used to change the imon slope on low power processors) which could allow you to disable turbo on boot, but I don't know the real scope of what can you do with that method.

At this point you rebuilt almost everything power related so I think that's not the problem. Regarding VCCCORE and VCCGT their mosfet drivers ere enabled at the same time but their pwm signals are controlled by PU8001 at the CPU's request through SVID protocol (VCCCORE first, check at the end attachment). If you get an oscilloscope in the future you can monitor the SVID line and it's clock along with the phases. With the multimeter you can check SVID Alert signal, but since you have VCCCORE I doubt it has any problems.

If you want to log stuff, log the SMBus/I2C on linux. Maybe you can get a clue, but since it is freezing on boot it won't log anything probably.

The other thing I noticed is that the GPU has the infamous Micron D9WCW, which eventually fails, so if you manage to repair it and the VRAM is giving you problems out of nowhere that's the reason.

Maybe there are corrosion products under the CPU from the liquid damage (there, your options are either reball or a reflow which I'm not a fan of). When you first removed the heatsink did the CPU have liquid metal? that stuff is problematic if it goes under the the components and for the CPU swap you only option is one from a system to be known to has a bootguard profile 0.

I'm out of ideas at this point, so I'm leaning to blame the CPU but I can't really tell for sure.