Causes for GPU artifacts

S

StandartBlitz

n00b
Joined
Jul 29, 2021
Messages
60
I have a Gainward GTX 770 Phantom with heavy static artifacts and already tried reflowing it which didn't help. Then I inspected the solder connections of the RAM chips and found one that was particularly stretched - fixing it also did not work.
Now have a useless graphics card liying around and would like to know whether the GPU is still good and the RAM / board bad or vice versa so I don't throw away potentially good parts.
GPU - Z gives no readings for vmem and video memory stress test can't recognize any errors.
Are there any other IMG_20210802_145515.jpgdiagnostic tools or ways to narrow down the problem without replacing all the RAM - Chips?
 

Attachments

  • IMG_20210802_145714.jpg
    IMG_20210802_145714.jpg
    579.2 KB · Views: 0
  • IMG_20210802_145542.jpg
    IMG_20210802_145542.jpg
    582.3 KB · Views: 0
yup mem "or controller" is dying. sell it on ebay as "for parts" and call it a day.
 
I testet the "poking" technique but there was no chip affecting the pattern upon pressing down. I also inspected the pc beforehand and now tested most of the caps individually but couldn't find any faulty ones. The best thing would be to distinguish between the dead and working chips but i can't run most of the memory tests since i can't install drivers on a faulty card.
 
This is the same issue that many nvidia chipsets had of that era. Honestly, there isn't much you can do to fix it unless you are competent with re-soldering BGA.
 
AustinW said:
ceramic caps don't tend to go bad. But for the most part you are right. I thought you were testing capacities. I would look for missing caps and other small SMD parts. Also look for damaged traces
Click to expand...
How can you spot a damaged trace if it is not on the top layer?
 
I compared the pcb to an image I found of the GTX 770 Phantom but couldn't find any differences and it does not appear like there are defective traces 🤷‍♂️
 

Attachments

  • IMG_20210802_204706.jpg
    IMG_20210802_204706.jpg
    579.5 KB · Views: 0
  • IMG_20210802_204736.jpg
    IMG_20210802_204736.jpg
    487.3 KB · Views: 0
AustinW said:
are you running the card with that little heat sink setting on the core? could it have knocked a SMD off the core? I don't see any maged on the card from those photos. a few SMD caps could have a better solder flow likely coused by the reflowing
Click to expand...
More likely hiding the screen print on the GPU for some reason.
 
That heatsink is only to prevent the die from overheating while looking whether the artifacts are still press and not for high power applications. The card also didn't work before and also could not catch myself breaking off any components - I'm not that silly :-/
 
Those inductors and electrolytic caps look different from the ones that I've seen in pics of the bare PCB for this card. Did you replace those? Also looks like there is an array of 4 pads bottom-left of the GPU that were soldered over. Any reason why?
 
The inductors look a bit darker (maybe another brand) but the values seem to be the same. The solder pads on the left (at least if they are the ones you mentioned) are unused - even in the original - and just a bit fuzzy due to the hot-air-gun.
 
AustinW said:
TBH at this point I don't think its something simple. likely a bad vram chip, BGA or something in the core
Click to expand...
Yes, highly likely. Many from this era had solder that would crack on the BGA. This is why the heating trick may work, and a BGA resolder is required. Or just chuck it.
 
AustinW said:
Odd I didn’t know bad solder extended to cards that new. If you have a heat gun or hot air station you may be able to fix that if its the case.
I have a gas soldiering iron for this very case, hot air gun attachment comes in handy.
Click to expand...
I already reflowed it twice and fixed the suspicious chip.
 
You can try an artifact scanner, but from your image it looks like the VRAM may in fact be good.

https://www.raymond.cc/blog/having-problems-with-video-card-stress-test-its-memory/

Honestly, I am not that knowledgeable in fixing GPUs. I have seen images like yours, but I will usually just do an RMA.

Also, consider selling on eBay for "parts or repair". You won't get much for such an old GPU, but it's better than just throwing it out.
 
Does the problem appear immediately, or does the card need to warm up a few minutes before it artifacts?

If it's ok at first (fresh bootup after being off for at least an hour), then you can likely pipoint the bad component if it's ram or another chip. Once the issue starts, start freezing the ram (freeze spray, or use a can of compressed air upside down), 1 chip at a time until you find the one that affects the issue.

But all of that is more to satisfy curiosity than anything. Baking the card actually just tends to be a short term 'fix'.. the issue always comes back. baking also isn't actually reflowing the solder, not really. To get it hot enough to actually melt the solder (215C or so) in a "slow" way like a cooking oven, takes too long and just damages other components or the PCB, or they all go slip-n-sliding around the pcb.. and then you've really done it.
They use machines to reflow solder and it's a very quick process. You also really need an SMD solder station to replace SMD parts.
 
GoodBoy said:
Does the problem appear immediately, or does the card need to warm up a few minutes before it artifacts?

If it's ok at first (fresh bootup after being off for at least an hour), then you can likely pipoint the bad component if it's ram or another chip. Once the issue starts, start freezing the ram (freeze spray, or use a can of compressed air upside down), 1 chip at a time until you find the one that affects the issue.

But all of that is more to satisfy curiosity than anything. Baking the card actually just tends to be a short term 'fix'.. the issue always comes back. baking also isn't actually reflowing the solder, not really. To get it hot enough to actually melt the solder (215C or so) in a "slow" way like a cooking oven, takes too long and just damages other components or the PCB, or they all go slip-n-sliding around the pcb.. and then you've really done it.
They use machines to reflow solder and it's a very quick process. You also really need an SMD solder station to replace SMD parts.
Click to expand...
Thanks for the reply but the artifacts are appearing on boot even on room temperature and for reflowing I used of course a heat gun and not an oven since I'm not that silly 😋
I've come to the conclusion that it is most likely an die-trace thing since I've run across two other cards with artifacts. After reflowing only the "normal" ones disappeared but the white ones stayed. I know - it's strange that only the white horizontal ones are not affected but maybe it's an architecture thing (all three Kepler or Fermi).
 
StandartBlitz said:
Thanks for the reply but the artifacts are appearing on boot even on room temperature and for reflowing I used of course a heat gun and not an oven since I'm not that silly 😋
I've come to the conclusion that it is most likely an die-trace thing since I've run across two other cards with artifacts. After reflowing only the "normal" ones disappeared but the white ones stayed. I know - it's strange that only the white horizontal ones are not affected but maybe it's an architecture thing (all three Kepler or Fermi).
Click to expand...
When you say you're doing this with a heat gun, do you mean an actual heat gun, or are you talking about a real hot air rework station? When you do this, are you reballing the chips, or literally just heating them up and then letting them cool down uncontrollably? Are you using flux? Are you preheating the board first?

Edit: Graphics cards are manufactured using an oven. If you have an oven, even a domestic food-cooking one, that can heat the board rapidly enough, that technique can work, but again, technique is critical, and if you're only doing the heating step, it's not going to work reliably, if at all, no matter how good your equipment is.
 
RazorWind said:
When you say you're doing this with a heat gun, do you mean an actual heat gun, or are you talking about a real hot air rework station? When you do this, are you reballing the chips, or literally just heating them up and then letting them cool down uncontrollably? Are you using flux? Are you preheating the board first?
Click to expand...
I'm using this hot air gun and I pre-heat the board for around 2-4 minutes depending on the area. I also don't use flux since I had one card on which the the flux caused some of the solder balls to join. I don't know whether it was a weired one - time - incident but I have seen nobody so far using it for the job 🤷‍♂️
 
AustinW said:
whats wrong with using the oven?
Click to expand...

Cooks all of the components on the card, and potentially damages the plastic connectors. It also off-gases left over fluxes and residues that stick to the inside of your stove and whaft around the house.

Throwing anything electronic into your home oven and cooking it at a random temperature for a random amount of time is dumb and should never be done.

StandartBlitz said:
I'm using this hot air gun and I pre-heat the board for around 2-4 minutes depending on the area. I also don't use flux since I had one card on which the the flux caused some of the solder balls to join. I don't know whether it was a weired one - time - incident but I have seen nobody so far using it for the job 🤷‍♂️
Click to expand...

You're just making things worse. You MUST use flux when reflowing anything. If you heat up a dry joint without any flux, it just damages the metal by further oxidizing it and making it brittle, it will never melt properly.

Also a hot air gun is far too hot. Unless you have a thermal probe where you can monitor the temperature of the part, you need a hot air rework station where you can control the heat.
 
GiGaBiTe said:
Also a hot air gun is far too hot. Unless you have a thermal probe where you can monitor the temperature of the part, you need a hot air rework station where you can control the heat.
Click to expand...
For most PC parts, particularly those with large VRM areas, what I've found is that even a hot air rework station alone isn't really good enough. You can't get the part you need to work on hot enough to melt the solder without taking a very long time or using way too much heat, and damaging things the way a heat gun would.
 
RazorWind said:
For most PC parts, particularly those with large VRM areas, what I've found is that even a hot air rework station alone isn't really good enough. You can't get the part you need to work on hot enough to melt the solder without taking a very long time or using way too much heat, and damaging things the way a heat gun would.
Click to expand...
Yeah, I definitely have that issue on huge power plane areas. A heat gun can be beneficial in cases like that, but gotta watch the temps so stuff doesn't get burned.

I've seen people get heat guns, crank them on max like 1500W and hover them like an inch above a PCB and wonder why everything turns to charcoal, which is why I tell people to not use them.
 
Agreed, so oven is best?
I've done it nd no lasting damages, just felt weird for a few days. everything is normal now.
 
UltraTaco said:
Agreed, so oven is best?
I've done it nd no lasting damages, just felt weird for a few days. everything is normal now.
Click to expand...
It depends on exactly what you're repairing, but the best way is either a BGA rework machine or a PCB preheater and a hot air rework station.

I'd echo the sentiment that you should not bake computer parts in a domestic oven. If you won't refrain from doing that because it's not generally permanent solution, at least consider not doing it because it has the potential to contaminate your oven with nasty chemicals that may then make it into your food.
 
Okay, taco won't do it anymore. Besides my old video card won't take a second baking I doubt nd newer one has that rohs stuff that won't melt unless hell temps. Tha k you!
 
You must log in or register to reply here.
Back
Top