It's far more likely from a particular manufacturer rather than a model of card that there might be poor soldering. I can't tell you which manufaturer had a poor soldering facility unfortunately.

The other issue that could affect cards is those with poor cooling. When the card heats up, the solder balls on a BGA chip will heat up and expand. When the PC is switched off, the balls cool down and shrink. This constant shrinking and expanding of the balls will eventually cause them to form dry joints and thus card failure.

I think it was also on some GF8 and maybe 9 series cards, especially the mobile versions. Since it's a thermal problem, not sure what you can do there other than applying new thermal paste.

Pretty sure it's nvidia cards using ball grid arrays for solder. I think the 8 and 9 series got a reputation for it on the GPU chips but I have and have had 6 and 7 series cards fail. Sometimes it's just the memory chips.

Never had a problem with GF4 Ti or FX series, and a gainward 7800gs+ I had from new [with a massive hsf] never failed and nor did a vanilla 6600. But I've had 6600gt, 6800gt, 7950gx2 all fail with bad solder.

My surmise is that any of the 6 series onwards that get repeatedly heat stressed, don't have good cooling, or are physically too heavy/whatever that are more prone to it. It must be a physical issue caused by solder flexing and cracking.

Don't xboxes get a similar problem causing ROD?

You mean the 360? But you know that's all but Nvidia, right? 😜 It wasn't ATI's fault though since Microsoft used a cheap ass cooling solution on the early models.

The term that you might be looking for is micro-cracking due to thermal stress between full load and room temp. Lead-tin solder that Nvidia used during those generations wasn't of the best quality and even Apple got a rep during those years for the same issue. Geforce 6 and 7 were the worst for this but it continued well into the geforce 8/9. ATI/AMD switched over years before but at a higher cost to a silver-tin based solder for stronger bonds that allow for higher temps. The late Geforce 9 models particularly the 55nm models did use silver-tin.

Simple reflow can for a short time fix this problem but it only last a short time. So it is best to keep the temps as low as reasonably possible. Some brands like Asus or MSI may have used the higher quality solder but I can't be certain.

Got recently 2 Quadro cards based on GF4 Ti and both had typical broken image artefacts. These older cards just worked more years, before they died on this. Never ones usually die after 3 years. And doesnt care from whom you buy them. Card makers don't care and customer must buy new one.

vlask wrote:

Got recently 2 Quadro cards based on GF4 Ti and both had typical broken image artefacts. These older cards just worked more years, before they died on this. Never ones usually die after 3 years. And doesnt care from whom you buy them. Card makers don't care and customer must buy new one.

Take off the coolers and toss in the oven on cookie sheets then bake for 10min at 200c then allow for a gradual cool for 30min. It doesn't fix bad dies but broken bonds it will. For US/UK you want an oven that can handle at the very least 385f and bake for the same time.

My last job was related to this as i was repairing these problems. Most of the times the GPUs are just fine, it's the solder that gives in. Removing the GPU, cleaning the solder and applying new one with lead usually fixes the problem (see reballing on the internet). Lead free ROHS solder is real crappy. Not only GPUs fail. Northbridges, Southbridges, memory chips, etc.. Silver is another solution, but as mentioned, comes at a higher price and the melting point is higher.

The oven trick... not good. Remember that every component is stressed at +200ºC... Short term solution is a reflow over the GPU with the surrounding area isolated with aluminum tape. A proper machine is required for optimal results, but it can be done with hot air or even a good hair dryer. 😉

Yes, it is lead free ROHS solder that is the failure, This stuff just sucks. I work on other electronics , not computer related and they have been using this stuff now for years. Anything which generates a heat up/cool down cycle always fails with it. It's not limited to any one industry.

I have 2 dead cards with the problem, an 8800GT and a 9800GX2. I lack the proper equipment to re-solder them. 🙁

Lead free solder + large temperature changes = cracked solder joints.
Also, video card manufacturers don't really want your card to last very long. 2nd hand high end cards can hurt sales of new lower end cards. The early death of higher end (and hotter running) cards is a good thing for their interests.
Customers also sometimes complain about fan noise, and modern gamers think it's fine for a video card to die after 1-2 years. So they have little incentive to address this in the video card market.

You want to reduce the severity of the temperature swings it experiences. So that means reducing the peak under load.
My suggestion is to find a utility to watch the operating temperature of your card. If it's getting very hot under load then find a utility to modify it's fan behavior. Set it to be more aggressive than the factory settings. I've set my GTX260 to peak at about 10-15C cooler than it did from factory, and I believe that will help it last longer.
Debris and dying fans also are a problem in the long run. Don't expect the card to stop, it will gladly broil itself.
Speedfan can be used for monitoring on most machines, and it can be set to play an audio alarm if it goes over a threshold. Setup of that program can be confusing though.

This was a massive problem on the HP dv6000 laptops with an NVidia chipset. I replaced a motherboard for that issue and the only "fix" I could come up with was to use a 3rd party utility to undervolt and underclock the CPU. Too bad it lost a bit of speed but the fan is just too lazy and I couldn't find a way to change that.

nforce4max wrote:

vlask wrote:

Got recently 2 Quadro cards based on GF4 Ti and both had typical broken image artefacts. These older cards just worked more years, before they died on this. Never ones usually die after 3 years. And doesnt care from whom you buy them. Card makers don't care and customer must buy new one.

Take off the coolers and toss in the oven on cookie sheets then bake for 10min at 200c then allow for a gradual cool for 30min. It doesn't fix bad dies but broken bonds it will. For US/UK you want an oven that can handle at the very least 385f and bake for the same time.

I see people mention this method quite often, but it could quite easily destroy the caps as they're usually rated for either 85c or 105c MAX temps! Not to mention the damage it could cause to the core and RAM baking them at 200c.... But i suppose it's worth a try as there's nothing to lose if the card is already buggered.

> as they're usually rated for either 85c or 105c MAX temps!

That's the temperature the capacitor will function 1000 hours at. Temporary soldering temperatures are higher.

i really think you risk contaminating your oven 😜 even without lead in the solder.

shamino wrote:

Lead free solder + large temperature changes = cracked solder joints.

Apple had this problem with white G3 and G4 iBooks where the brittle solder would break and render the machine non-functional.

shamino wrote:

This was a massive problem on the HP dv6000 laptops with an NVidia chipset. .

Apple experienced a problem with the Macbooks with nVidia 8400M graphics. It was an issue with the video chipset itself, something to do with the layers of the chip separating due to a faulty manufacturing process and it can't be repaired unless you replace the GPU chip itself. All the GeForce 8 based cards other than the 8800 had the problem and neither Apple nor nVidia ever fully acknowledged the issue even though independent testers caught it early on once the warranty claims started rolling in.

The caps shouldn't go over 105C but once you have cracked joints under a BGA, the caps are really the least of your concerns. Caps aren't too hard to replace, I would just plan on doing that if the affected device otherwise seems to have recovered. Or another approach would be to remove them ahead of time, then solder them back after the oven.

h-a-l-9000 wrote:

> as they're usually rated for either 85c or 105c MAX temps!

That's the temperature the capacitor will function 1000 hours at. Temporary soldering temperatures are higher.

But you only touch the capacitor legs with the soldering iron for 1 or 2 seconds... 10 minutes in an oven at 200c is a major difference and they could easily leak or burst! But they cost pennies and are easy to replace anyway, as long as nothing else has been damaged in the process.

Soldering iron? That's not how capacitors make it to the board nowadays.

http://www.sciencedirect.com/science/article/ … 957417404001721

Check out the temperature profile graph. Ok, it's less than 10 minutes.

h-a-l-9000 wrote:

Soldering iron? That's not how capacitors make it to the board nowadays.

It is with me when i replace them 😉.