PNY gtx 970 - no vram voltage

[KiwiJonas]

Hi,
I'm pretty new to repairing graphic cards and I recently took my old 970 in hope I could fix it.

I did a few measurements and it turned out that the 2 vram inductors show 0 volts.
I'm guessing theres a problem with the IC which is controlling the voltage but I cannot figure out what kind of IC that is. Its labeled:
2Y=2L
H1D
(See picture)
Maybe someone here will know what it is.

I have also noticed that the NCP5369 mosfets are only giving out 212 mV. (I did remove 2 of the 4 transistors)

[fzabkar]

RT8811CGQW, Richtek, Dual-Phase COT Buck PWM Controller with Dynamic Voltage Control, marking 2Y=xx, WQFN-4x4-24L:
https://www.richtek.com/assets/product_file/RT8811C/DS8811C-00.pdf

http://www.smdmark.com/en-US/search/code?id=2y%3D

[KiwiJonas]

This is what I needed thanks a lot!
How did u find that out?
I will continue taking measurings by tomorrow.

[fzabkar]

How did u find that out?

See my last URL.

[KiwiJonas]

So I did some measurings and I stumbled over a short between ground and both vram inductors.

The Onsemi NTMFS4C10N  MOSFET's third and second pin (both source) are in direct connection with both sides of the vram inductors and so to ground.
Might it be that one or both mosfets have failed? I'll have a picture added to exactly show the short path.
If I get this completely wrong, does anyone have an idea what this problem could be?
The transistor next to the 2 mosfets is labeled as 4CO5N but I haven't Identified it yet.

thanks in advance

[fzabkar]

http://www.smdmark.com/en-US/search/code?id=4C05N

Are you sure you are measuring a short? Usually the load resistances are very low. Check the enable pins of the corresponding PWM controllers to see whether they are being turned on.

[KiwiJonas]

If I measure between vram inductor and ground it shows a flat zero ohms but if I measure from the transistor NTMFS4C10N to the inductor it shows about a hundred ohms because there is a capacitor between NTMFS4C10N and 4C05N >-> which then goes to the inductor
I will check the enable pin of the PWM later and give feedback.

[KiwiJonas]

I measured the enable pin of the PWM and it only shows 30 mV. also PVCC gets 0 volts and it seems to be somehow connected with a fuse that goes to the 6 pin 12 volt PCIe connector. But I cannot tell whats exactly wrong. the connections seemed to be there. U14 also gets the 12 volt but that MOSFET shouldn't have anything to do with the PWM of Q16 and Q15.

[KiwiJonas]

Im guessing I will resolder the MOSFET's again and see if there is still no voltage on the vram inductors. But If so, I still have no image or detection of the gpu at all

[fzabkar]

Those fuses ("Q") appear to be polyfuses. These are resettable.

[KiwiJonas]

Oh yea indeed. I measuring 2.5 ohms on the 3rd (buttom) fuse when the system is turned on and 0 ohms when its off. The other fuses stay at 0 ohms.

[KiwiJonas]

This fuse goes from the 12V pin on the PCIe slot of the motherboard to a 5 milli ohm resistor to the Drain of NTMFS4C10N MOSFET. On this MOSFET, when I measure from Source to ground it shows a 100 ohms but my multimeter beeps in continuity mode.

[KiwiJonas]

Can you tell me what that is?

this connects to the enable pin, which only puts out 30 mV.

[KiwiJonas]

it says MA6

[KiwiJonas]

So I have been spending days figuring out the missing enable Signal and I found out that the NCP81174 IC on the backside of the card for Vcore is lacking output voltage for the Driver/MOSFET's.
Also there is no VIN voltage at the RT9059 Dropout regulator. The VIN pin goes to Vmem Inductors, so it is using the voltage of the Vmem rail to to run which is missing due to the missing enable signal at RT8811C PWM Controller.

Voltage on the Vcore is present but extremely low (~200 mV)
There is no short on Vmem, I got that wrong.

Is the RT8811C depending on the Vcore rail (NCP81174)?
If yes, does the NCP81174 rather sound to be completely dead and needs to be replaced is it still possible for a resistor to be broken? If a resistor with no value written on it is broken, I won't find out if its broken.
Does anyone have an Idea where the enable signals comes from? It's currently pulled to ground through a dual transistor (SOT 23 - EN Goes into pin 6 and is getting shorted through pin 1 when system powers on). It'll need a signal on PIN 5 to be pulled high.


Maybe its just a dead GPU?

[testpoint1]

please make sure the NVVDD is 0.9V (about), then can check the FBVDDQ (about 1.35V).

[KiwiJonas]

NVVDD is about 0.2 V.. FBVDDQ is 0 Volts. As I said thats because of the Missing phases from the PWM controller because it is not enabled due to missing enable pin voltage.
Where does the PWM get it's enable signal from?

[KiwiJonas]

I got the card working again and it was actually pretty easy but it took me 2 weeks to find the fault.

For those who might need help in the future I'm going to mention how I fixed it real quick.

At first sight of the PCB I did some measurings and I quickly noticed that there is 1st: no short 2nd: no Vmem voltage and abnormal low Vcore voltage.
As I'm new to GPU repairs I didn't pay much attention to the low Vcore voltage and was more focused on the not present Vmem voltage, because I thought this might be the reason for the low Vcore voltage.
So I spent days digging and researching how Vmem MOSFET's are connected to eachother and where every single signal/input voltage comes from. In the end I was assuming that the PWM controller is just dead but after some more digging I figured out that the Enable signal comes from Vcore voltage and thus the enable signal is very low. So I started investigating around the Mutli-Phase Buck PWM Controller (NCP81174) and thought it was dead because I got no Phase output. I replaced it but nothing changed. In the end I was very frustrated and after some more circuit diagram analysis I assumed the problem to be on the 4 Phase Driver MOSFET side (4x NCP5369's), which power the GPU. I desoldered them one by one to see If one is just dead and pulls the voltage to ground or something but it didn't work out very well. I did some more measuring and I was trying to understand how they are connected with each other and to the PWM controller and I noticed the third one had a differend bootstrap voltage on the BOOT pin then the other ones. (only 4,5 Volts instead of 5) I desoldered the third one and it didn't help so I thought it might be the dead capacitor which is between BOOT and PHASE. desoldered it too and still nothing changed. I was about to give up. I just had one last Idea to test if the 4,5 volts are normal or not and switched the second NCP5369 with the third one and suddenly the GPU got Voltage again and Vmem too. The card got detected again and was functioning normally. So it was just the broken third phase MOSFET that made the whole system fail. Anyways I ordered all NCP5369 at the beginning of my research because I had already been on the thought of a broken MOSFET when I started investigating the card. I will replace all 4 phases with the new ones and everything should be fixed. I did not stress test the GPU yet but I will do that when everything is nicely replaced and screwed together again.

Nontheless this all helped me gain some experience for future GPU repairs.

I'm thanking fzabkar for the help and datasheets he gave me.

cheers.