GTX980 - GPU die KO?

[magneh]

Hello!

Got this card broken, no image.
Ive checked there are a few resistors and caps missing, also a small coil as far as I can tell.

The point is that ive checked also the GPU die and Ive found 3 cracks in the corner die...

This would also need new chipset right?
Or still could work?
Just to know if its worth trying to identify the few missing components and give it a try.

Would be good to hear the opinion of someone who had several of these

Thank you

[dicky96]

Re: GTX980 - GPU die KO?

What does it do when you plug it in? Does the PC recognize there is a graphics card (one bleep on booting)

Where is the small coil missing from? pics please.

[magneh]

Re: GTX980 - GPU die KO?

What does it do when you plug it in? Does the PC recognize there is a graphics card (one bleep on booting)

Where is the small coil missing from? pics please.

Hola!

For now I have not plugged it in.
Just visual inspection, and found these components missing (see pics)
For what I know, it does not give an image and fans do spin.
I can test it later on.

In these pics in red the components missing, in yellow the ground pad.
Pics 3,4 and 5, its almost for sure a cap, so not the reason for no image problem for sure.
Pic 2 its where I say it could be a coil, but now that ive measure the ground pad, most likely a cap.

I think the most important components missing are in Pic 1, where there is for sure a resistor missing.

ive also identified the zones in the PCB in another pic.
And also the best pic of the same PCB of the same card I could find to compare.

What do you think? waste of time?

Gracias!

[dicky96]

Re: GTX980 - GPU die KO?

Waste of time? Not at all!

First find all the buck regulator coils/inductors on the other side and measure from them to ground

If you don't see any shorts then you need to plug it in and see what it does - and then report back here.

Something like 0.5 ohms on the GPU supply rail (the one that will have multi phases) is not a short by the way.

This short video should give you an idea what to check and what resistance readings you are likely to find. This is a GTX280 but it's likely to read similar to your GTX960.

This turned out to be a working card by the way - so you should see similar readings on your card if there are no shorts.

https://www.youtube.com/watch?v=Pc3E4oR_o_Q&t=3s

Rich

[magneh]

Re: GTX980 - GPU die KO?

Waste of time? Not at all!

First find all the buck regulator coils/inductors on the other side and measure from them to ground

If you don't see any shorts then you need to plug it in and see what it does - and then report back here.

Something like 0.5 ohms on the GPU supply rail (the one that will have multi phases) is not a short by the way.

This short video should give you an idea what to check and what resistance readings you are likely to find. This is a GTX280 but it's likely to read similar to your GTX960.

This turned out to be a working card by the way - so you should see similar readings on your card if there are no shorts.

https://www.youtube.com/watch?v=Pc3E4oR_o_Q&t=3s

Rich

Hey!

So done a few things.
All values seem fine and safe to turn the card on.

But before this Ive checked on the PIC1 where we can see 3 components missing, one of those is indeed one resistor, and comparing with other similar Mosfets below that one, a 1K resistor.
Ive soldered it back, and fired up the card.

Fans spin, GPU chipset heats up.All tensions are present.
Screen detects the input and resolution, but no bios screen or any image.

All other components are 90% for sure caps, so would not cause the card not to boot.

I think those cracks in the GPU die will be the sentence...

[momaka]

Re: GTX980 - GPU die KO?

What do you think? waste of time?

No, but knowing how problematic the GTX980 cards are (well, all large-die, high-power GPUs are these days), I don't suggest spending too much effort on it. Good chance the GPU is probably toast too, and not just from the cracks in the core.

With that said, probably replace all of the critical SMD parts (looks like there may be a few, but not all of them are) and give it a try. If nothing, check all voltages (which it looks like you already did.) If voltages are correct and GPU still appears dead, give it a baking/reflowing as a last resort. If that doesn't bring it back, it's probably dead for good.

The point is that ive checked also the GPU die and Ive found 3 cracks in the corner die...

The cracks in the corner in picture #2 likely aren't an issue. But the one crack/chip shown in picture #3 is not in a corner (though it is at the edge), so it could be potentially fatal.

ive also identified the zones in the PCB in another pic.

Using this picture of yours as a reference...

Area 1 is probably power detection logic circuitry, so you will likely need to find out what the missing parts there are. Otherwise, card may not POST.

Area 2: this appears to be some kind of a buck regulator controller, so MOST DEFINITELY find out what the missing parts are. Not having the correct parts here could cause an under-voltage, over-voltage, or just plain bad regulation on one of the rails, so the components in this area are critical.

Area 3 and 4: I see only chipped/missing ceramic caps, which is *usually* no big deal, as most of these just filter the RAM V_dd rail. However, any missing resistors should be replaced back. Also, a few of the ceramic caps could be signal-coupling caps... so if any of those are missing or shorted, you may get artifacts. I saved a Radeon HD6850 video card from a fault like that (had a tiny shorted ceramic cap used for signal coupling and not as V_dd filter.)

Area 5 from this image: those components are for the VGA (analog) output only. It looks like you're missing a resistor/jumper/inductor and a cap on one of the lines. But all that should do is cause missing Red, Green, or Blue color from analog output and nothing else. So no need to worry about fixing anything here, at least for testing.

If you don't see any shorts then you need to plug it in and see what it does - and then report back here.

Be careful with that method. As I mentioned above, some components - particularly those around any VRM controllers - may be critical to the operation of the card. So anything missing in those areas could certainly cause complete chaos on a VRM circuit and burn more components out. I actually had an HD3870 video card sent to me like that once. It was missing a few small SMDs. Originally, I thought of it as nothing. But good thing that I checked - one of the missing SMDs was the upper feedback resistor to the GPU V_core. Without it, the GPU V_core would have shot up as high as the VRM controller could go, which from what I remember was quite a bit more than what the GPU was designed for. Luckily, I didn't power the video card before this. Instead, I found out what the value of that resistor was from another identical GPU and then scavenged one from a dead GPU. Worst part is that video card was advertised as working... which I'm sure it was, actually. My guess is the seller didn't handle it correctly and broke the SMD parts when packing it.

So the conclusion here is to always check 2nd hand video cards, regardless of what operational condition they were sold to you. I've actually had this happen more than once to me. So I always inspect my cards and fix any issues before sticking them into a system for a test. Surely, some SMDs are not critical and those I do skip replacing sometimes.