Repair r9 390 nitro gpu that seems to be fine

[more_stars]

Finally the second r9 390 board arrived and after some checks, all the vrm seems to be fine but the board does not power and there is no video out.

The ohm readings from ground to coil below

Aux vrm 42 ohm
Ram vrm 118 ohm
Core vrm 21 ohm
0.98v rail 0.2 ohm

After seeing that ihe 0.98v rail had so an low resistance, I desoldered the coil and seen that the low resistance is on the pad closer to the main gpu.

All the fuses are fine and the 12v rail and 5v rail is present.

I think that the problem with this board is the ram chips as they are covered in liquid from the thermal pads. Is thare a way to fix those or do I need to replace them?

Or is there a common problem with those boards? I do have an second board with some working parts.

Thank you in advance and Stay safe!

[Manul]

I don't know what is that liquid, but there is good chance that it is not conductive or corrosive so just wash those VRAM chips with alcohol and dry good.

The power rail does not look healthy. What does this 0.98v rail power? You need to track down the short.

[more_stars]

I wanted to track the rail but there is no available data about the pcb. The only thing I had seen was from the actually hardcore overclocking video https://youtu.be/U_J8uEcGCiM that said the 0.95v rail is for video output driver.

The think that made me curious was the fact that non of the voltages ware present at any of the coil sand the main gpu did not heat up what so ever even with the radiator removed.

[more_stars]

After further investigations I was able to see that the 3.2v and 1.6v are present and so, I can not understand what might be wrong. I don't have an osciloscop and so I think I can not test the mosfet driver.

In order to do further checks, what should I look for? After one minute with the boar plugged in, nothing seems to heat up.

[more_stars]

Is there any way to see what is wrong with my gpu?

[Manul]

Maybe something like this can help you.





When you said, that you measured 0.2 Ohms on a rail, do you mean the 0.95v rail, called "Display drive rail" in the second video?

Maybe you can take a picture of the card and draw everything where and what you measured.

[more_stars]

I will attach the photos here. Also I took out u4, u3 and u4010 but the short is still there but after some research I did find that some of the rail have low resistance so the 0.2 ohm reading would be normal but to check that, I need a working card or some data about this pcb.

[Manul]

Based on the fact, that it is a relatively low power rail, 0.2 Ohms looks way too little. I may be worng, but still... Also it seems, that this rail powers the PLL's, so it means no clocks without it.

[more_stars]

Is there a way to check if the pll is bad? Also, is there an option to run the gpu without the original pll?

[Manul]

I noticed, that you call graphics card a GPU. GPU (graphics processing unit) is the main chip on a graphics card, it is doing calculations. Like a CPU on the motherboard. It is just a part of the system. I'm saying because it makes confusion.

PLL's are part GPU chip and generates various clocks, for GPU itself, for memory, etc.

What kind of equipment you have? Lab power supply? Scope? Thermal camera?

[Manul]

There are ideas, for example use lab power supply, set voltage limit to something like 0.5 - 0.9 volts and inject current into the short. With thermal camera it may be possible to observe slight heating (this is a little dangerous for the card).

Another idea is to use a miliohm capable meter, or build it yourself.

Yet another idea is to inject high frequency AC to the short and use EMC probe to observe the loop made by a short circuit.

[more_stars]

I don't have a thermal camera but the first idea seems to be the best one. I will try the second one but what am I supposed to look for?

Also, thank you very much for your help.

[Manul]

Accurate resistance meter (miliohm resolution), can help find shorts, because the closer you will get to the short, the lower will be resistance. But it is still tricky, mainly because of variantion of contact resistance of the probes. Also, power rail of a graphics card will be a beefy copper trace or even plane, so resistance variantion will be very small.

Another idea is to inject current to the rail and measure voltage on various rail points. That is also not guaranteed to be useful, but may be a try. Idea is that the current will go to the point of short and come back through ground plane. Because rail has some liitle resistance, it will cause small voltage drops, which can be measured and analyzed to predict a likely point of short. This technique needs a little imagination, hard to explain universally for all possible cases.