Author Topic: MSI GTX970 4G OC - no vram voltage  (Read 514 times)

0 Members and 1 Guest are viewing this topic.

Offline slimax

  • Newbie
  • Posts: 3
  • Country: pl
MSI GTX970 4G OC - no vram voltage
« on: December 13, 2020, 06:34:46 pm »
Hi,
I'm trying to fix my graphics card.
GPU resistance seems to be normal, GPU voltage is 0,9V.
Ram resistance is 97ohms but 0V.
Mosfets seems to be fine, no shorts etc.

I cannot identify what kind of two-phase controller is used for RAM.
It's in WQFN-24L package, maybe someone know what is that?
Labels:
08 2E
V1R
On pin 18 there is VCC.
Lgate: 16,19
Ugate: 14, 21
It seems to be RT8809A/B or something similar, can someone confirm that?

If that's the correct chip, then it's EN/MSEL is pulled down to GND while PC is powered on.
This EN pin is described like below:
Chip Enable and Mode Selection. This pin is a tri-state input. Pull up this pin to
exceed than 4.2V, controller operation into DEM mode. Pull up this pin to between
1.2V to 3V, controller operation into ASM mode. Pull down this pin to GND,
controller will shutdown.

This pin it going to 6th pin of sot23 6 pin chip near controller, labeled:
MA
.6
pins 2,3 are shorted and goes thru 10k resistor to 12V.
pins 1,4 GND.
pin nr 5 is going to something like a divider but without resistor to ground soldered.
What is that?
« Last Edit: December 13, 2020, 08:16:39 pm by slimax »
 

Offline fzabkar

  • Frequent Contributor
  • **
  • Posts: 898
  • Country: au
« Last Edit: December 14, 2020, 08:13:36 pm by fzabkar »
 
The following users thanked this post: slimax

Offline slimax

  • Newbie
  • Posts: 3
  • Country: pl
Re: MSI GTX970 4G OC - no vram voltage
« Reply #2 on: December 15, 2020, 01:04:17 pm »
Thank you Fzabkar, thats exactly what it is.
I have reverse engendered a bit and:

There is missing enable / mode select signal used also in other components:

- GS7117ST-105-R
- EN/MSEL RT8809BZQW (indirectly using MBT3904DW1T1)
- Main power controller NCP81174 is connected to this signal to pin nr 2, which is:
VID - Voltage ID from processor.

All resistors / capacitors looks ok in this EN/MSEL signal area, no shorts to ground.

Am I right thinking that GPU is controlling this signal to enable RAM and adjust power / voltage for GPU itself?

It seems to be dead GPU (4,4ohm) ?
What should I check next?
Is there still hope for BIOS failure ?
« Last Edit: December 15, 2020, 01:09:26 pm by slimax »
 

Offline fzabkar

  • Frequent Contributor
  • **
  • Posts: 898
  • Country: au
Re: MSI GTX970 4G OC - no vram voltage
« Reply #3 on: December 15, 2020, 08:23:04 pm »
AFAIK, low resistances are normal for GPU and CPU cores. I haven't done any work on GPU chips, but AIUI the VID defaults to a safe configuration at power-on. I don't know if the GPU itself or the external logic can then adjust the settings.

On examining the datasheet, it would appear that the NCP81174 PWM controller begins regulating with a target voltage defined by the REFIN input. This is tapped off VREF (2.0V) via a potential divider. Subsequently the controller gets a VID signal from the GPU and then updates the core voltage.

https://www.onsemi.com/pub/Collateral/NCP81174-D.PDF

The Richtek chip seems to allow for two preset values. When it powers up, the output voltage (Vset) is set via a potential divider with Vref (2.0V) as the reference. In the application circuit the potential divider consists of R18 and R19. After the circuit powers up a new output voltage can be selected via the VID input. This merely shunts R4 to ground, thereby reducing the output voltage. Thus there are only two voltage settings -- high when starting, and then a lower voltage when running.
« Last Edit: December 15, 2020, 08:56:17 pm by fzabkar »
 

Offline slimax

  • Newbie
  • Posts: 3
  • Country: pl
Re: MSI GTX970 4G OC - no vram voltage
« Reply #4 on: December 18, 2020, 07:38:23 am »
Finally card repaired, it was a little tricky.
GPU voltage sometimes was present and sometimes not.
If GPU voltage was present, memory voltage wasn't   |O
After some more measuring it turned out that one little resistor was broken so NCP81174 12v monitoring pin had 0V and there should be 1,2V. Luckly when it was working I manage to measure that. After resistor replacement surprisingly card started immediately, no idea why it wasn't before when GPU had voltage...
Thank you very much for help, hopefully next time fix will go more straight :)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf