Here are the pictures. Sorry for bad quality but My phone can't focus.
https://drive.google.com/drive/folders/14UPEq … K7a?usp=sharing

PCI-e ones shouldn't be a problem since at worst it would just run with less PCIe lanes.

One thing to keep in mind that with these sort of finds its often a case of already non-working cards being thrown in a box where they lose surface-mount components on top of the original problem.

I suspect that the card was working before cause most of the stuff I got from this scrapper was working, although this one might be a exception. Will try to power on this card on an PCI-e x8 slot.

Okay, tried it in an 8x and 4x slot and nothing. On an crosshair 4 formula it makes the board go in some sort of a reset loop for a while then it boots fine (no GPU beeps). Although it's still not detected in windows. The cause might or might not be that I'm using an PCI VGA card, so maybe it disables the other one? On an gigabyte board it boots without any beeps but still no display.
I have also noticed that the caps at the back of the die get scorching hot and R847 is missing. Any ideas what will the R847 be?

Replaced R847 with an 1k from my research but no go. Still hangs the MSI board, so no idea what might be the issue. Any ideas?

kotel wrote on 2024-10-26, 19:15:

Replaced R847 with an 1k from my research but no go. Still hangs the MSI board, so no idea what might be the issue. Any ideas?

If there was one component knocked off, carefully check for others. Good chance there will be more. Also, sometimes a component might not appear damaged but could still be. I've seen this a few times on scrapper-picked boards. On one, it was a shorted MLCC (multi-layer ceramic capacitor) that appeared fine physically, but measure short-circuit. It was a coupling cap on a PCI-E lane going to the GPU, causing the GPU to not get detected. Another time, it was a low-value array resistor close to a RAM chip that appeared to be fine, but was actually cracked and not conducting an all pins. I caught that one by complete coincidence (don't even remember how anymore.) Probably 100% helps that I do this under natural day light close to a big window.

Munx wrote on 2024-10-20, 18:45:

PCI-e ones shouldn't be a problem since at worst it would just run with less PCIe lanes.

Technically speaking, that's true.
In practice however, I've found that some motherboards will not detect a video card if a single PCI-E lane signal from the PCI-E lane signal pair is jammed or open-circuited.

So for this reason, I always test PCI-E video cards on a 1x miner riser board in addition to testing them in the motherboard. I've ran into quite a few cards with failing GPUs where the PCI-E lanes were starting to loose connection between the GPU chip and the board, making the card not always or not ever detect in a motherboard. Luckily, with almost all of these, the bad lane signals were some higher order above 1x, so these cards worked OK in a riser board at 1x speed.

Thanks for the reply.
I have tested most of the caps on the backside under the GPU die and all were okay (no under 1 ohm). I also tried flashing a new BIOS but atiflash 4.something and 2.8something didn't detect the card. Guess I'm gonna test for short further...

Munx wrote on 2024-10-20, 18:45:

PCI-e ones shouldn't be a problem since at worst it would just run with less PCIe lanes.

One thing to keep in mind that with these sort of finds its often a case of already non-working cards being thrown in a box where they lose surface-mount components on top of the original problem.

In the real world that is often not the case. I've repaired lots of non-working cards that were missing just one of the dc blocking caps for a differential pair.

So it turns out one cap in 1x arena was shorted. Replaced it but was greeted with artifacts. Question is does MATS/MODS support this card?

Okay, wierd stuff happening. Once again... So after the artifacts the card stopped displaying and after some fiddling it displayed again. During the copying files sequence on an XP install it started having artifacts, once again. After that no display again. Then I resoldered the resistor I've replaced before cause it came undone. After that the card is alive again. I'll leave it overnight and if its gonna work I'll put the blame partially on the resistor. Then I'm gonna do some hl2/BiT tests cause MATS/MODS doesn't support my card.

Okay its dead again. This time it doesn't hang the boards up. Just undetected. No shorts on the caps near PCI-e slot. Any ideas?

With artifacts like that and the fact that the GPU doesn't always get detected 100% means GPU chip is bad.

Give it a reflow if you have the equipment (or just fry it on your stove top and with the help of a hot air gun.) If nothing comes out of it after a few reflow attempts at 215-230C, then it's time to chuck it to the scrap parts bin. And if the reflow does manage to get the GPU going again, it probably won't last long unless you keep it real cool (under 45-50C at all times, regardless of load)... though even that doesn't guarantee it will go bad.

Okay an reflow "fixed" it. That or me replacing the two transistors (forgot what model) on topside of the card with MMBT3904's.
But this ones very suspicious. Had it make my test bench (which otherwise worked fine) shutdown. It all starts with no more display, then HDD parks its head and lastly the whole board shutdowns with the PWR led on PWR button having an delay at turning off...
Not gonna dive deeper and just call it an partial success.