Zotac RTX 3060 , another learning experience

[90Ninety]

Disclaimer Firstly I am by no means expert and this is still a learning experience , I buy broken stuff , with the intention to fix or learn something by failing.. I wanna break this into a few sections, please feel free to mention the section if replying , to assist me on getting a better understanding

Part 1 , identifying Rails .. So for the record , I understand on NVIDIA cards there are several Rails. Along with their Supposed Voltages ( Voltage depends on age/ generation )



Part 2. Blown / Missing component check .. I purchased this from the original owner , no repairs attempted .. Upon inspection , the 12V PCIE lane had a cracked /blown inductor ( Filter ? ) , which I guess acts as a fuse , or at least in this case it has, I replaced it with a similar size one from a donor ( shown below in red ) . I then read a short from the 12V to the Mosfet / Driver chip on the GPU Power rail


Part 3.
Removing Short
I have a Milliohm meter , so I removed the lowest Ohm driver chip , however I ended up removing all .. Only then the short disappeared .. I did not identify the suspected shorted Driver chip , however the last one did read a bit different ( Lower ohm - by half ) . I soldered all back , while marking the suspect Driver chip .. Mysteriously even with all the driver chips soldered back , the short disappeared 🤨 .. I took off the suspect Driver as a matter of ' Process of removal ' ( Shown below red ) .



Part 4 .. In progress .. Here I will measure the voltages on each rail but first I want to clarify which rail is , and what voltage I should be seeing

[90Ninety]

OK I done some Google-Fu

A vlogger on youtube called 'Northridge Fix' has a video fixing this exact card here .. So I followed some of the diagnosis steps . I also identified the Rails

All resistances check out fine ( to Ground)

12V 10's K Ohm rising - Jumps to Megaohm's after several seconds for some reason 🤨
3.3V Thousands K Ohm rising
1.8V Kilohm Rising
5V coil Kio Ohm Rising
PEX 60 Ohm ( double than video shown )
Memory Rail 30 Ohm
Checked pins on PCIE connector (PCIE lanes, signal Pex Return ect) , all seem good!


I powered up the card to check voltages . There is just the 12V, 1.8V , There is no VCore, No PEX , No 5V , No MEM .. There is one driver removed , though I presumed there would still be voltage

[90Ninety]

I created a video of troubleshooting so far

https://youtu.be/Rl353_YETQ4

[90Ninety]

Fixed now , by replacing the BL00 DRMOS

This is one of two I fixed with the same fault , same phase also ..

[90Ninety]

:/

Damn one of the DRMOS blew again . This time I replaced all the BL00 DRMOS phase chips. However now again there is no VCORE voltage, however all other voltages are present on the other coils ( VMEM, 1.8V PEX, 5V)

Should I look into replacing the 9512Rs controller, I checked the Enable PIN , there is just 1V

[DynaxSC]

Hi, I have same card which came also with shorted mosfet but not only. Also an electrolytic cap on the 12V line was shorted as well as some 5 pin vrm near the video ports. Even the R47 coil was blown. I guess the electrolytic caps on these boards are crap. They even look like crap.They are small dimensions with quite high capacity, so they are probably manufactured to achieve highest parameters in smallest casing but at cost of reliability w/o any security margins. I replaced them all to one step smaller capacity in same case size, and added 2 more caps, as there are places for them on the pcb. I took the caps from a premium ASUS motherboard I dedicated for parts. These are most probably japanese Nichicon caps which are best in class. The card works now, but gives artifacts on one memory chip. I now wait for a new mem chip arriving this week, and will replace it later. Hopefully its the memory, not the GPU failing.

In your case if the mosfet failed again there must be a reason for it. Maybe the vrm controller, maybe some failing cap shorting for a microsecond, or maybe something else. But I would definitely replace the controller in the first priority. I believe also replacing the electrolytic caps on the primary vrm side for some high quality ones might be a good idea.

Also in my opinion this card is very "downsized" by accountants, the board has space for 2 more phases, but the space is not populated. Five phases for such a gpu is really a total bottom line. Looks really strange and seams like a struggle between engineers and accountants. Long live all the profit driven corporations rated on the stock exchange. Make Shareholder Value grate again (read: don't care about quality and reliability, ... and "customer" - WTF who is this?").

[DynaxSC]

Hi, I have today received the memory chip and exchanged it.
After exchanging the memory chip the artifacts are gone, so this means the GPU is OK, at least as it concerns the artifatcs, but ...

I started a Furmark test and the card worked ca. 10 seconds correctly, then the power supply suddenly switched off.
Tried to power again, but no luck, short protection active at PSU, short at 12V GPU Power.

Took the card out of the board, measured 12V supply's and the short was only on the 12V PCIe slot power supply, the 12V line from the PSU had no short.
Looking on the BV I discovered that at least my version of the card has 3 phases which are powered form the PCIe slot 12V power, and separately the other 2 phases are supplied directly from the PSU and the 12V ATX plug at the card edge - strange, have'nt seen such solution before.
The short was on the first mosfet looking from the pcie slot edge.
Replaced the mosfet, card started again, run again furmark, and again power off and another short of the same mosfet (fortunately ordered 10 of them).
Checked the all the resistors and caps around the mosfet, all OK.

Then I exchanged both the mosfet (on my card NCP252160's) and the VRM Controller - on my cards it's a uP9511R, not as on your card a uP9512R.
Card worked again after starting the board, started Furmark - and voila, no short any more !!!

So I can only recommend you to exchange the VRM vontroller on your card. It seems they can go broken "partially" and doing demage to the card under load only.

Although the card worked now, the fan did not - zero RPM (but this was also before so). I could't test Furmark long, as hot spot temperature rised too 100 C in a 3 minutes or so - no air cooling).
So I exchanged the fan to one from a GTX1070 Blower Card - didn't have anything else.
Fan started to work, and shows also RPM under GPU-Z, so seems to be fitting for this card. GPU temperature now not rising to 100C any more.
Tested the fan under MSI Afterburner - 3800 RPM at 100% - so very good RPM's.

But, unfortunately I have a different problem now, i.e. the card works not with full capacity, like on the below screnshot:




GPU-Z shows 100% GPU load but framerate is only 24 frames at Full HD resolution - which is way below the capability of this card at this resolution.
GPU Clock is OK, changing from some 200MHz to 1530 MHz at Furmark test.
GPU temperatures are also low. Board Power draw is also only 61 Watt (only 36% of TDP), so much less as it should be.
CPU load is under 5%, PCIe also working at full PCIe 3.0 x16 (10-th gen CPU, so no PCIe 4.0 present), so no bottlenecks here.
Removed all drivers with DDU, reinstalled drivers, no change.
Changed windows test installation to another M.2 disk I had, same behaviour, so no issue with OS or drivers for sure.
Looked with scope on all the 5 phases - very nice square signal at each phase - so all are working.
GPU voltage is changing between ca 0,6...V and 0,95..V at max. load, tried to make higher with MSI Afterburner, but Afterburner has no effect at the GPU voltage - so seems something with the VRM control is not OK.

Somebody any ideas what could be the problem ?
Could it be that the exchange of the fan to one from a 1070 Blower card could have such effect, new vrm chip from AliExpress not OK ?

Fan already excluded - disconnected it, no change.

Below pictures of my card:



And the VRM part:

[DynaxSC]

Measured now the physical GPU VCORE voltage with DMM, and surprise, it is only 0.540V instead of ~0.85V showing at GPU-Z - seems there is something wrong with the voltage measuring network on the card.

90Ninety - could you pls kindly post a high resolution picture of the area around the vrm chip on the backside, from your card, eg. like the below area:


I have the suspition, that somebody messed around the vrm chip before, so I'd like to compare which elements are populated, which not.
I have some BV of a similar MSI card, but there are much more elements populated, so can't compare with it, and a ZOTAC BV is not available.

Below the link to the BV's.
The one similar to my cards is V397_v40.

https://www.badcaps.net/forum/troubl...=90093&page=55

[DynaxSC]

Tried to change the VRM controller to:

1. uP9511P -> VCORE present but card not recognized by the BIOS
2. uP9511S -> VCORE not present, card not recognized in BIOS
3. A second uP9511R -> same bahaviour, so VRM chip seems to be OK.

I wounder if it shoudn't be a uP9512R, will order one.
Would try also another BIOS, but no BIOS'es for this card available at Techpowerup.

Link to my BIOS request:
https://www.badcaps.net/forum/troubl...=90093&page=55

[SenaDani]

Why don't you just try Twin Edge bios? Board looks to be the same.

[DynaxSC]

Just measured a few low ohm resistors and shunts, all OK.
Then tried a few BIOS'es for similar looking cards, including the one for Twin Edge, and the effect is the card is gone to heaven for now.
VCORE is only 0,3V and fan 100% RPM.
No idea what happened, but seems the problem is become even more mysterious.
Measured VCORE resistance with mohm meter, and it's quite low, 0,09 oms (90 mohms) - this is in a range, where the core might be already faulty, at least if I compare to GTX1070, where the core has usually 150-190 mohms.
Might be that something stopped contacting (core ?), because during mesuring shunt resistors, I needed to press the probes quite strong on the card to become good contact.
Think must do a brake with repair of this card, cause I do not have any more ideas, except core reflow.
Will look in the meantime for a second same type of card, comparing two same type of cards helpes usually in repair a lot.

[90Ninety]

Measured now the physical GPU VCORE voltage with DMM, and surprise, it is only 0.540V instead of ~0.85V showing at GPU-Z - seems there is something wrong with the voltage measuring network on the card.

90Ninety - could you pls kindly post a high resolution picture of the area around the vrm chip on the backside, from your card

Hi buddy

Thanks for your input on this thread, I hope it helps with your card also. Sorry for the delayed update, here is the image.

[DynaxSC]

Thank @90Ninety.

[90Ninety]

This card is driving me insane


After changing the PWM controller, there is no difference, still no VCORE.

I am beginning to lean on that there is a faulty/ blown component on the logic circuitry. I checked the card with an IR camera, there seems to be one SN74LV1T08 (of a pair) that is heating up to around 60 degrees. As shown on the right of the IR photo


The IR image is slightly off center but I am sure it is the component below

Here is a closer up




I removed this component, the resistance is much higher on the legs, though there seems to be low resistance (24 Ohm to ground) on the Data pads on the PCBA board where A & B (see 1 & 2 in image below) legs go . I presume this is why the component is heating up? Compared to the exact component next to it, the resistance a lot lower.







I have no idea where these data tracks go to .. I would like to send this card to someone who can diagnose better. I am now out my depth

[DynaxSC]

Have you looked at the boardview's I posted, maybe they can be of help for you. Does the vrm controller gets EN signal?

[90Ninety]

Have you looked at the boardview's I posted, maybe they can be of help for you. Does the vrm controller gets EN signal?

I did have a look yes, I checked the EN pin, there is 1V , should this be higher ?

[90Ninety]

Have you looked at the boardview's I posted, maybe they can be of help for you. Does the vrm controller gets EN signal?

Sorry, I'm an idiot, there is no EN signal


I guess I have to trace out the logic going to this pin (pin 5 of 9512Rs)

[90Ninety]

So I am looking at a board view suggested above, believed to be for an MSI 3060. The board I have shares a very similar PCB but does not have all the components populated compared to the MSI

Here is the progress so far, tracing out where this enable signal comes from.

PICTURED BELOW
From pin 36 of 9512R to a 10K resistor

To a SC70_6 (Pin 3) PS_NVVDD_EN_R











































To a 0Hm resistor

To a SC70_(5 Pin 2) PS_NVVDD_AND_IN2






So next is to check voltages on the above,.. what voltages I should be looking for on the components above, and how much further I need to trace back to find the voltage drop off. Is there a typical failure point for this kind of fault?