X1900 XT displaying artifacts

[mitchmaster]

Hi all,

I would like your help to fix a X1900 XT video card that doesn't work properly, in fact it never did.

A little history: I bought the card in Oct 2006, it was fine in games and the various 3DMarks but it would display artifacts in ATITool (that stress test with the furry cube from 10 years ago if anybody remembers). I never RMAed it because I probably thought that they would send it back claiming that it worked fine.

The artifacts in ATITool got worse over time and the card eventually started displaying them in games as well but I could live the occasional artifact at first, then they started being too many and too big and I changed the card in Dec 2010. I have attached a couple of screenshots of the artifacts that it displayed back then if they can help.

After I replaced the card I used it in another PC for 2D use only for a few years, but now it has been sitting in a drawer for a some time and would like to try and fix it.

I have tried it earlier and it does the following:

- 2D is fine
- if I run 3DMark06 it shows a few red dots on the black screen while the test is loading
- when the test starts the card may either a) complete it fine, b) complete it with light artifacts, c) the card resets itself and I get a driver error, d) the PC freezes and I have to reset it.
- the above happens before the card has time to heat up a lot, except when it completes the test, and it does not go above 65° anyway.

I understand that this may be tricky to troubleshoot, we are looking for some component that degraded over time and I don't really know what to check (no lytics on the board, sorry, that would be too easy). Maybe a bad component in the power section, or a bad solder joint under a RAM chip or the GPU?

Here you can see high-res pics of the front and the back of the PCB:

http://www.vgamuseum.info/images/zaa...900XT_fwhq.jpg

http://www.vgamuseum.info/images/zaa...1900XT_bhq.jpg

If you have any tips that could be helpful I would be really thankful.

Thank you

[Pentium4]

Re: X1900 XT displaying artifacts

Yeah, my guess would be it's the solder on the RAM chips. This looks like a card from the early days of lead free solder

[ChaosLegionnaire]

Re: X1900 XT displaying artifacts

i suggest u post actual pictures of your card. i had the same problem with red dot artifacts as u with my msi 8800 gt oc. the problem turned out to be the qimondagate (infineongate) vram. problem was fixed by downclocking the vram 50-100mhz. check what brand of video ram is used on your card.

also, i hope your card is not a sapphire brand video card. sapphire cards are so poorly built that it is not uncommon for sapphire cards to be faulty out of the box even when new. i have seen many sapphire video cards with shoddy components on them on ebay and other sites.

[mitchmaster]

Re: X1900 XT displaying artifacts

Today I took the card apart to take the pictures asked by ChaosLegionnaire, cleaned it and put it back in and now it doesn't work at all for no apparent reason. It causes one of my PCs to power off and back on infinitely, in another one it stays powered with a red T_FAULT LED. Oh well...

I tried to fiddle with it for some time, checked for visible damage but couldn't find anything, so I thought I wouldn't waste any more time on it.

Thank you for your help anyway.

For the record, the card is a Sapphire and the RAM chips are by Samsung.

[momaka]

Re: X1900 XT displaying artifacts

Sell it on eBay or some other place. I'm sure there are people in your country who will buy it and repair it.

As with any high-power video card, artifacts is usually due to the GPU BGA going (exception being the GeForce 7 and 8 series, which also have the bumpgate defect). A proper reflow (with hot air or IR) has a high chance of bringing the card back to life. The RAM chips are rarely the culprit - even if they are Infineon (and if they were, downclocking the RAM, as ChaosLegionnaire suggested, is a good work-around). So it's worth a try if you have a heat gun / paint removal gun.

[ChaosLegionnaire]

Re: X1900 XT displaying artifacts

aye, should be bad bga on the gpu. i did some more "self-education" on the web and at first, it seems that sapphire luvs cheapening out on the gpu cooler and using single slot coolers on a lot of their video cards. i have owned a sapphire 512mb x1950 pro agp before and it uses a single slot cooler! however, after i googled for reviews and pictures of your card with the cooler on, it seems that sapphire didnt cheap out on the cooler with the x1900 xt if that is indeed the stock cooler of your card.

however, momaka mentioned before that coolers with the gpu and vram combined together can be problematic. the heat from the gpu could have made the vram hotter and thus the vram crapped out due to overheating. it could also work the other way. the heat from the vram made the gpu hotter so the rohs-compliant gpu bga solder balls crapped out faster due to high temperatures.

anyway, if removing the cooler and putting it back causes the video card to not work, its quite likely bad bga. the failing solder balls were still being partly held in working position by the tension of the cooler. removing the cooler caused the failing solder balls to completely go out of alignment. so when u put the cooler back, the failed solder balls dont go back to their previous position. so u get a completely undetectable video card.

another thing u can try is to flex the card's pcb left or right while its working. if it changes or eliminates the artifact pattern even momentarily or if it makes the card detectable again, u got a bad bga on the gpu.

lastly, if u want your high-end gpus to last a long time like me, always reapply the thermal paste with good/branded thermal paste. the stock tim from the factory is either crap or often poorly applied. also replace the cooler with a better one. keep your case well cooled and ventilated. gpu load temps should not exceed 70°C. research has shown that lead-free bga solder suffers from accelerated degradation if the temps exceed 70°C.

[mitchmaster]

Re: X1900 XT displaying artifacts

I have attached pictures of the actual card since you asked.

A proper reflow (with hot air or IR) has a high chance of bringing the card back to life.
...
So it's worth a try if you have a heat gun / paint removal gun.

I'll see if I can find one, a friend should have it.

it seems that sapphire didnt cheap out on the cooler with the x1900 xt if that is indeed the stock cooler of your card.

Yes, that's the stock heatsink.

anyway, if removing the cooler and putting it back causes the video card to not work, its quite likely bad bga. the failing solder balls were still being partly held in working position by the tension of the cooler. removing the cooler caused the failing solder balls to completely go out of alignment. so when u put the cooler back, the failed solder balls dont go back to their previous position. so u get a completely undetectable video card.

I don't know, I had already removed the heatsink many times before during the life of this card and I never had any problems.

another thing u can try is to flex the card's pcb left or right while its working. if it changes or eliminates the artifact pattern even momentarily or if it makes the card detectable again, u got a bad bga on the gpu.

I'll try and see what happens, the card is pretty sturdy with the cooler mounted though, I don't know if I can flex it.

There may be another problem though, I'll just post a full report of what I did the other day so you can judge for yourselves.

So what I did the other day is:

- I took off the stock heatsink
- for the first time ever in the life of this card I took off the VRM heatsink and the crappy stock thermal pad, and I probably shouldn't have done this. The problem is that the thermal pad had a thickness to it and the heatsink wouldn't make contact with the VRMs using normal thermal paste. This because the two ends of the heatsink where taller than the actual contact surface so I sanded them off. I immediately thought that the heatsink could short something but it looked fine at first then upon closer inspection (when I took the hs off again after powering the card and seeing it didn't work any longer) I saw that the bigger SMDs are just as high as the VRMs so there may have been a short anywhere inside the area below the heatsink.
- cleaned the card, brushed the back with a toothbrush to remove the dust that wouldn't come off with compressed air.
- cleaned the GPU and the VRMs, I noticed that the GPU has two chipped corners, minor stuff, shouldn't be much of a deal. I think they have been there for years but I'm not sure, I don't have any old pictures of the card with the hs off.
- applied new paste on the VRMs and the GPU, then mounted the VRM stock hs back on and a different hs on the GPU.

Then I tried the card and it didn't work. I know, I did too many different things and now I don't know where the problem is but I was totally sure it would work as usual. The one thing I had never done before is removing the VRM heatsink, I had already removed the stock hs many times in the past and mounted the other heatsink as well, so it was nothing new.

[mitchmaster]

Re: X1900 XT displaying artifacts

I have tried still another cooler which would allow me to flex the card but nothing happens, actually the motherboard doesn't go past "VGA detected".

[electromaster84]

Re: X1900 XT displaying artifacts

If the card is dead, and you don't have any other means to reflow it. You could try putting it in the oven at 385F (195C) for 5-10 minuets. I've had success in the past with graphics cards, and a few xbox 360 motherboards using this method. Here's a guide i found online if you would like to try it https://www.ifixit.com/Guide/Repair+...the+board/2240

[ChaosLegionnaire]

Re: X1900 XT displaying artifacts

do note that such a fix is only temporary. the video card will start artifacting again in a few weeks or months depending on use and the temperature it runs at. the best way is a proper reflow with bga soldering flux which is exactly what they do at the factory.

so to the OP, if u have only one video card to reflow, investing in tools like a hot air gun and bga soldering flux may not be economical. but if u have several video cards to fix, it may be more worth the $.

also another thing to note is that a number of members here have noted that ati/amd gpus have a higher reflow success rate than nvidia gpus. so this makes it more worth the money to pay for the tools to reflow your video card if u have an ati/amd gpu since u have a higher chance of success. so there's a reason why the pro members here call nvidia "no video" instead.

I noticed that the GPU has two chipped corners, minor stuff, shouldn't be much of a deal. I think they have been there for years but I'm not sure

dont really notice where is the chipped part is? is it at the bottom left and right corner of the die? (the die is the shiny/glossy raised black part of the chip) it looks more like either a cosmetic defect or a reflection due to the light to me. shouldnt affect the functionality of the chip. dont see any exposed copper on there.

or do u mean the square shaped silver metallic thing on the substrate (the substrate is the green part of the chip) called the shim? the shim is just to protect the die from cracking due to excess mounting force from the heatsink. since your gpu is protected by a shim, its not likely the die can be cracked or damaged from mounting/dismounting the heatsink repeatedly or by excess force.

[electromaster84]

Re: X1900 XT displaying artifacts

do note that such a fix is only temporary. the video card will start artifacting again in a few weeks or months depending on use and the temperature it runs at.

The xbox 360 motherboard I reflowed in my oven has worked perfectly for five years, and has seen some heavy use during that time. Though I did replace the x-clamp mounting system on the heatsinks with bolts and washers. I can see how results could differ on a gpu as the ram chips have no heatsinks with any real clamping force.

[mitchmaster]

Re: X1900 XT displaying artifacts

If the card is dead, and you don't have any other means to reflow it. You could try putting it in the oven at 385F (195C) for 5-10 minuets. I've had success in the past with graphics cards, and a few xbox 360 motherboards using this method. Here's a guide i found online if you would like to try it https://www.ifixit.com/Guide/Repair+...the+board/2240

I am going to try, I just don't understand if I the VGA should be facing up or down when I bake it.

Also, should I get some flux before attempting this?

dont really notice where is the chipped part is? is it at the bottom left and right corner of the die?

...

or do u mean the square shaped silver metallic thing on the substrate (the substrate is the green part of the chip) called the shim?

I was talking about the die. You can see the two bottom corners are lightly chipped, the right one more than the left one.

[electromaster84]

Re: X1900 XT displaying artifacts

I would bake it with the gpu and ram chips facing up myself. Also remember to insulate any parts of the card you don't want to reflow, and any plastic components on the card as well. Flux would certainly help, but is not required.

[momaka]

Re: X1900 XT displaying artifacts

- for the first time ever in the life of this card I took off the VRM heatsink and the crappy stock thermal pad, and I probably shouldn't have done this. The problem is that the thermal pad had a thickness to it and the heatsink wouldn't make contact with the VRMs using normal thermal paste.

Is the heatsink for the VRM also the same heatsink that goes on top of the GPU?

- cleaned the GPU and the VRMs, I noticed that the GPU has two chipped corners, minor stuff, shouldn't be much of a deal.

Yes, they look very small, so I don't think they will be a problem either. I've seen considerably bigger chips before with the hardware still working.

- applied new paste on the VRMs and the GPU, then mounted the VRM stock hs back on and a different hs on the GPU.

Then I tried the card and it didn't work.

What do you mean by that? Did the computer power up at all? No video but fans spinning? Video, but with artifacts.

The VRM heatsink may be shorting something out in the VRM without its original thermal pads. If the VRM heatsink is separate from the GPU heatsink, then just remove the VRM heatsink and test the video card without it.

If the card is dead, and you don't have any other means to reflow it. You could try putting it in the oven at 385F (195C) for 5-10 minuets. I've had success in the past with graphics cards, and a few xbox 360 motherboards using this method. Here's a guide i found online if you would like to try it https://www.ifixit.com/Guide/Repair+...the+board/2240

No offense to anyone, but I really hate those oven reflow guides! It's not that you can't use an oven for a reflow, but rather the fact that some boards have electrolytic and/or polymer capacitors on them - NEVER put those in the oven!!! You WILL ruin them. The card/board may work fine for a little bit, but no guarantees for how long. And *when* it fails next time (not IF), you won't even know if it's the BGA that failed again or the polymer/electrolytic caps you ruined.

So with that said, I have to say that putting an Xbox 360 motherboard in the oven is about the worst you can do. I used to work in a repair shop that fixed mostly Xbox 360s and I have seen the "works" of other "repair shops" when they tried to bake those 360 motherboards. The results were never pretty, and we rarely managed to get those fixed easily.

So all in all, I wouldn't not suggest that guide above. Not unless you put a warning about the capacitors getting damaged.

Also, 5-10 minutes at 195C won't do anything for lead-free solder. You might want to read that article again. It states to pre-heat the oven to 195C, and then heat the board to higher temperature. I suggest 225-235C for lead-free solder and 200C for leaded.

The xbox 360 motherboard I reflowed in my oven has worked perfectly for five years, and has seen some heavy use during that time.

That's one of the rare cases, though. My guess would be that you didn't get the caps hot enough to vent them and that the partial "reflow" you did combined with the modded X-clamps is what did the trick. Not to rain on your parade, but it's only a matter of time before it fails again, though. The GPU heatsink in the 360 is just undersized. Even a proper reflow, a reball, and/or a new GPU can't help that. Microsoft just designed the Xbox outside-in rather than the other way around, so the cooling is not good.

Though, I will say that I'd rather take a 360 for repair any day over a PS3.

however, momaka mentioned before that coolers with the gpu and vram combined together can be problematic. the heat from the gpu could have made the vram hotter and thus the vram crapped out due to overheating. it could also work the other way. the heat from the vram made the gpu hotter so the rohs-compliant gpu bga solder balls crapped out faster due to high temperatures.

RAM chips barely produce any heat. So the second scenario, where the RAM makes the GPU run hotter is not possible.

Rather, the problem with heatsinks that combine RAM and GPU cooling is that the thermal pads that are used on the RAM often tend to push up on the heatsink, so the contact force on the GPU core will not be as strong. This may produce weak thermal coupling between GPU core and heatsink, thus raising the GPU core temperatures. And with raised GPU temperatures, the PCB of the card also runs hotter. The RAM cools primarily through its Vcc and ground copper pads on the PCB. So with a hotter PCB, the RAM will also run much hotter. That's why GPU+RAM combined heatsinks are a stupid idea. So in short, never bother with the RAM thermal pads once they are removed.

gpu load temps should not exceed 70°C. research has shown that lead-free bga solder suffers from accelerated degradation if the temps exceed 70°C.

That.
And also large changes in temperature (30-40C) between idle and load operation. In fact, frequent large temperature changes between idle and load are just as bad (if not worse) over a constant high temperature.

[mitchmaster]

Re: X1900 XT displaying artifacts

Is the heatsink for the VRM also the same heatsink that goes on top of the GPU?

No, it's a different heatsink.

What do you mean by that? Did the computer power up at all? No video but fans spinning? Video, but with artifacts.

It causes one PC to power on and off in a loop, another one doesn't go past "VGA detected" during the POST.

The VRM heatsink may be shorting something out in the VRM without its original thermal pads. If the VRM heatsink is separate from the GPU heatsink, then just remove the VRM heatsink and test the video card without it.

Tried that, doesn't work.

No offense to anyone, but I really hate those oven reflow guides! It's not that you can't use an oven for a reflow, but rather the fact that some boards have electrolytic and/or polymer capacitors on them - NEVER put those in the oven!!! You WILL ruin them. The card/board may work fine for a little bit, but no guarantees for how long. And *when* it fails next time (not IF), you won't even know if it's the BGA that failed again or the polymer/electrolytic caps you ruined.

I was planning to shield the rest of the card with tinfoil and leave just the GPU and its rear exposed, that should do.

Now I have found some flux in gel form, but maybe a friend of mine can get me some liquid flux. Would you suggest me to use the gel flux or the liquid flux?

[electromaster84]

Re: X1900 XT displaying artifacts

No offense to anyone, but I really hate those oven reflow guides! It's not that you can't use an oven for a reflow, but rather the fact that some boards have electrolytic and/or polymer capacitors on them - NEVER put those in the oven!!! You WILL ruin them. The card/board may work fine for a little bit, but no guarantees for how long. And *when* it fails next time (not IF), you won't even know if it's the BGA that failed again or the polymer/electrolytic caps you ruined.

So with that said, I have to say that putting an Xbox 360 motherboard in the oven is about the worst you can do. I used to work in a repair shop that fixed mostly Xbox 360s and I have seen the "works" of other "repair shops" when they tried to bake those 360 motherboards. The results were never pretty, and we rarely managed to get those fixed easily.

So all in all, I wouldn't not suggest that guide above. Not unless you put a warning about the capacitors getting damaged.

Also, 5-10 minutes at 195C won't do anything for lead-free solder. You might want to read that article again. It states to pre-heat the oven to 195C, and then heat the board to higher temperature. I suggest 225-235C for lead-free solder and 200C for leaded.

Well maybe I should have said I didn't use a guide when I did it. I just pulled all the through hole caps off the board, insulated the whole motherboard except the gpu and cpu areas, and stuck it in the oven set at 450F for 8-10 minuets. Nothing was burned, scorched, melted or otherwise damaged. I have fixed 4 xbox 360 motherboards using this method, and I haven't heard of any of them failing yet. Maybe I'm just lucky, or I'm not a total idiot?

[momaka]

Re: X1900 XT displaying artifacts

I was planning to shield the rest of the card with tinfoil and leave just the GPU and its rear exposed, that should do.

Well, you don't have any aluminum electrolytic and polymer capacitors on your card, so you don't have to do this. The only thing you might want to cover with foil are plastics, since they can deform, melt, or become yellow.

Now I have found some flux in gel form, but maybe a friend of mine can get me some liquid flux. Would you suggest me to use the gel flux or the liquid flux?

The form of the flux is not important. Both gel and liquid will work if they are the right type. That said, the flux should be be "RMA" type (preferably) or standard RoHS rosin non-activated (i.e. mean for lead-free soldering, which means it won't evaporate as easily at higher temperature). *AVOID* RA (Rosin Activated) types and any flux meant for plumbing / copper pipes - those fluxes are aggressive and will eat traces on the board over time.

Now, if you are using gel flux, put a line of flux on each side of the GPU before putting the card in the oven. With liquid flux, again put some on each side with the board tilted so that the flux goes everywhere. DO NOT over-apply flux, because that can cause the balls under the chip to dislodge and bridge.

Well maybe I should have said I didn't use a guide when I did it. I just pulled all the through hole caps off the board, insulated the whole motherboard except the gpu and cpu areas, and stuck it in the oven set at 450F for 8-10 minuets. Nothing was burned, scorched, melted or otherwise damaged. I have fixed 4 xbox 360 motherboards using this method, and I haven't heard of any of them failing yet. Maybe I'm just lucky, or I'm not a total idiot?

Okay, now that you said it like that... NO, you definitely weren't lucky - you actually DID IT RIGHT by pulling those caps off! So I am not surprised you got those Xboxes fixed.

My only gripe with the above article is that it mentions *nothing* about removing caps or other sensitive components. I've seen people try to bake their PS3 boards with the CMOS batteries still in there. And guess what happened to those?

Generally, most electrolytic caps can withstand a high temperature over 105C/125C for a short period of time, which is why people can still have successful repairs even with the oven baking method. However, the longer the caps are exposed to high temperature, the more their life gets affected. So even a difference of just a few minutes more in the oven can make a difference. That's why I don't recommend the oven reflow method - you really have to know what you are doing by either removing the caps or baking the board quickly enough for the solder to melt, but not more than that if the caps are not removed from the board.

[electromaster84]

Re: X1900 XT displaying artifacts

Hey mitchmaster any update on this?

[mitchmaster]

Re: X1900 XT displaying artifacts

Hey mitchmaster any update on this?

Yes, sorry but I got busy with other stuff and then I just forgot to update you.

I eventually did the reflow, and it didn't work.

I shielded the board with tinfoil, applied flux on the sides of the chip (maybe too much) and attempted the reflow. I followed some guide I found here with temperatures up to about 230°C but something obviously went very wrong, as you can see in the pics.

After I let the card cool I took off the tinfoil and saw that the flux was burnt and spread all around the PCB, which had turned to a darker hue of red. The stickers on the back were burnt as well, even if they were under the tinfoil. I though that maybe the temperature was too high but I had been checking it with an oven thermometer.

I tried to clean the board with isopropyl alcohol but I couldn't remove the sticky flux. Then I still decided to try it, and obviously it didn't work. I kept it lying around a few more days, then I binned it.

I guess I'm done with reflows for a while.

[ChaosLegionnaire]

Re: X1900 XT displaying artifacts

uh oh... it looks like u tried to wrap the video card like a baked potato. the foil isnt supposed to be touching the card. the flux too isnt supposed to come into contact with the foil. most likely a chemical reaction occured between the aluminium and the flux causing the charring of the flux that u see. u are also supposed to remove all stickers too as they can combust and turn into carbon in the oven heat.

sorry about the card. i hate to see good ol' stuff go like that too. should have kept it for display as a museum piece.