8800gts fixed from oven trick (pics)

[xplicid001]

SO I started a thread asking for a suitable suggestion for my fried 8800gts (found here: http://forums.anandtech.com/showthread.php?t=2079927)

i took some pictures of the process for shiets and giggles. I rushed to get it done before i had work. The whole process took about an hour. I had in the oven for 8 minutes at 400.

The broken 8800gts

removed all the screws to remove the heatsink

RUST? i used a razor blade to remove after i backed it

card before thermal paste removed

heatsink with paste

paste removed

into the oven

screwed back in and ready to go



RESULTS: SUCCESS!!

 

[TwinsenTacquito]

I wish I had something that has died for no reason that I could do this to. I got a motherboard with blown capacitors, dead PSUs, a motherboard with huge black burns from lighting... booo no science for me!

 

[BD231]

I wish I had something that has died for no reason that I could do this to. I got a motherboard with blown capacitors, dead PSUs, a motherboard with huge black burns from lighting... booo no science for me!

Yeah really, who cares about benchmarks when you can easy bake oven your hardware.

 

[nenforcer]

Is that a small philips head screwdriver or a torx screwdriver you used to remove the screws?

 

[taserbro]

Happy to hear that. I'm about to try with a couple old radeons that went down almost two years ago now though I probably won't be using them.

I do find it amazing that such ghetto reflow soldering has such a high success rate on these cards. It makes you think what causes cards to consistently fail due to smt solder points and what could be done to prevent that.

 

[EarthwormJim]

Happy to hear that. I'm about to try with a couple old radeons that went down almost two years ago now though I probably won't be using them.

I do find it amazing that such ghetto reflow soldering has such a high success rate on these cards. It makes you think what causes cards to consistently fail due to smt solder points and what could be done to prevent that.

Google bumpgate.

This is really a specific issue to Nvidia, and their poor choice of underflow material.

Basically the solder bumps crack joining Nvidia GPUs to their substrate, causing open circuits.

Since the substrate isolates the solder bumps, and as long as the chip is not physically moved while heated. The solder should remelt, removing the cracks for awhile.

The best solution is to not let affected Nvidia GPUs undergo extreme temperature gradients. The sudden and large changes in temperature causes the solder to undergo excessive strain, leading to cracks.

 

[xplicid001]

Is that a small philips head screwdriver or a torx screwdriver you used to remove the screws?

It was a philips head.

 

[Bl0cks]

This is great news to hear :thumbsup:

 

[pmv]

I don't suppose there's a possibility that that rust could have been shorting something out? it looks a bit as if the oxide has spread across the PCB - could that possibly have electrically connected things that shouldn't have been connected?

 

[GodisanAtheist]

I don't think that's rust on the underside of your PCB. Having used a heat gun to remove a number of processors from their PCBs, I can tell you it looks a lot more like heat damage than it does rust...

 

[TurtleCrusher]

Nice job. I hope my old 8800GTS 320mb did this. I need a cool homework project. I should probably start testing that card again.

 

[Scali]

This is really a specific issue to Nvidia, and their poor choice of underflow material.

It isn't actually.
It's a common problem with many companies when they switched to ROHS regulated materials.
Aside from that, nVidia doesn't build their own cards. They just outsource it to some big Chinese/Taiwanese companies. The same companies that also build various hardware for various other companies.
XBox360 is also a wellknown 'offender' of 'bumpgate', and quite a few laptops also suffer from it (mainly Dell, but also others).

Radeons suffer from this aswell. Humus (former ATi employee) has an article on how he baked his Radeon 3870 X2 back to life: http://www.humus.name/index.php?page=News&ID=283

 

[veri745]

It isn't actually.
It's a common problem with many companies when they switched to ROHS regulated materials.
Aside from that, nVidia doesn't build their own cards. They just outsource it to some big Chinese/Taiwanese companies. The same companies that also build various hardware for various other companies.
XBox360 is also a wellknown 'offender' of 'bumpgate', and quite a few laptops also suffer from it (mainly Dell, but also others).

Radeons suffer from this aswell. Humus (former ATi employee) has an article on how he baked his Radeon 3870 X2 back to life: http://www.humus.name/index.php?page=News&ID=283

I understand your point about defects like this affecting other companies; this is correct, but I would like to correct you on your point about "nVidia doesn't build their own cards" as a reason why it's not Nvidia-specific.

The "bumpgate" issue affecting their GPUs is the solder between the GPU die and the package, not between the package and the PCB. So it doesn't matter who builds the PCB, it's still an Nvidia defect.

 

[Scali]

I understand your point about defects like this affecting other companies; this is correct, but I would like to correct you on your point about "nVidia doesn't build their own cards" as a reason why it's not Nvidia-specific.

The "bumpgate" issue affecting their GPUs is the solder between the GPU die and the package, not between the package and the PCB. So it doesn't matter who builds the PCB, it's still an Nvidia defect.

Not really, as nVidia is a fabless company. That was my point (I never mentioned PCBs specifically, 'cards' means cards as a whole, PCB, GPU, memory, cooler, everything). Even if it were a problem that only affects nVidia products (which it isn't), it's not nVidia that chose the exact materials used in production, but the company they outsourced it to.

 

[veri745]

Not really, as nVidia is a fabless company. That was my point (I never mentioned PCBs specifically, 'cards' means cards as a whole, PCB, GPU, memory, cooler, everything). Even if it were a problem that only affects nVidia products (which it isn't), it's not nVidia that chose the exact materials used in production, but the company they outsourced it to.

When Nvidia ships their GPU's they are packaged. You can see the package, outlined by the white square, on the pics above. regardless of whether or not they are a fabless company, they still package the dies before they ship them. Pinning out a GPU die directly to the PCB would make the PCB ridiculously expensive, so they don't do that.

 

[EarthwormJim]

It isn't actually.
It's a common problem with many companies when they switched to ROHS regulated materials.
Aside from that, nVidia doesn't build their own cards. They just outsource it to some big Chinese/Taiwanese companies. The same companies that also build various hardware for various other companies.
XBox360 is also a wellknown 'offender' of 'bumpgate', and quite a few laptops also suffer from it (mainly Dell, but also others).

Radeons suffer from this aswell. Humus (former ATi employee) has an article on how he baked his Radeon 3870 X2 back to life: http://www.humus.name/index.php?page=News&ID=283

I'd still say that this is more of an Nvidia specific issue. At least such a wide spread, all encompassing defect. I believe around 3 generations of GPUs and chipsets are affected because of their underflow material of choice, and solder composition.

Manufacturing is outsourced, but ultimately design and packaging decisions are made by Nvidia.

Yes soldered parts are subject to stress related failures, but not really at the massive rates that the affected Nvidia GPUs are subject to.

 

[NoQuarter]

nVidia actually does choose the substrate material, it's like choosing cloth or leather seats - even if the fab is outsourced the materials choice and design layout is entirely nVidia's. The bumpgate issue was actually a combination of poor bump layout combined with poor choice of substrate material. Both are engineering decisions nVidia made.

Fwiw everyone I know who owns a 8600-8800 has had theirs die now and had to be RMA'd (my buddy's 8800GTS just died last week was the last one to go), and these are the only video cards we've had to RMA since we started with 3dfx Voodoo and Riva TNT cards.

 

[lopri]

Very nice pictorial. What's the ideal oven temperature and how long?

 

[EarthwormJim]

Very nice pictorial. What's the ideal oven temperature and how long?

The OP said in his post, 400F for 8 minutes. You need at least 400 degrees since it's lead free solder (200+ C melting point).

It would be more ideal to use a heatgun and a shield , rather than an oven. You want to heat the GPU up to melt the solder bumps. You don't really want to heat up the caps and other components.

 

[Unkle_Tar]

The OP said in his post, 400F for 8 minutes. You need at least 400 degrees since it's lead free solder (200+ C melting point).

It would be more ideal to use a heatgun and a shield , rather than an oven. You want to heat the GPU up to melt the solder bumps. You don't really want to heat up the caps and other components.

I got a HP laptop where the graphics crapped out(6150). I used a butane lighter to heat it up around the chip. Worked for about a month, then it crapped out again. Repeated the process, lasted for about 2 months the second time. I didn't do it a third time.

 

[EarthwormJim]

I got a HP laptop where the graphics crapped out(6150). I used a butane lighter to heat it up around the chip. Worked for about a month, then it crapped out again. Repeated the process, lasted for about 2 months the second time. I didn't do it a third time.

Large temperature gradients are what causes the cracks. GPUs do not heat up uniformly, and as things heat up they expand. Since they don't heat up uniformly, different parts expand at different rates; causing excessive strain on the solder.

Basically if you could upgrade the cooling or underclock the GPU to keep temperatures lower, it would last longer.

 

[lopri]

The OP said in his post, 400F for 8 minutes. You need at least 400 degrees since it's lead free solder (200+ C melting point).

Doh. Thank you I was distracted by all the pics!

 

[Scali]

I'd still say that this is more of an Nvidia specific issue. At least such a wide spread, all encompassing defect. I believe around 3 generations of GPUs and chipsets are affected because of their underflow material of choice, and solder composition.

Manufacturing is outsourced, but ultimately design and packaging decisions are made by Nvidia.

I really don't think that when you're outsourcing production, that you have control over details such as choice of underflow material and solder composition. Those things are very specific to the manufacturing process of the specific fab that you're outsourcing to.
In fact, I doubt that nVidia even has any kind of expertise in that area at all, and is even able to make such a decision.

Regardless, my point was that this is not a problem specific to nVidia. Radeons also suffer from it.
Yes, nVidia suffers from it more, but that's party a result of nVidia outselling AMD about 2:1 in those days. And partly a result of the media who love to hate nVidia and cuddle ATi.

 

[NoQuarter]

I really don't think that when you're outsourcing production, that you have control over details such as choice of underflow material and solder composition. Those things are very specific to the manufacturing process of the specific fab that you're outsourcing to.
In fact, I doubt that nVidia even has any kind of expertise in that area at all, and is even able to make such a decision.

Regardless, my point was that this is not a problem specific to nVidia. Radeons also suffer from it.
Yes, nVidia suffers from it more, but that's party a result of nVidia outselling AMD about 2:1 in those days. And partly a result of the media who love to hate nVidia and cuddle ATi.

They do, there are several materials to choose from when designing the GPU. nVidia used high lead instead of the more expensive eutectic bumps (what ATI used and nVidia switched to once they realized the engineering problem). They just underestimated the amount of stress that would be on the high lead bumps due to hot spots. The other contributing factor was their choice of underfill, which they also immediately changed after the issue was identified. They definitely have to have expertise in this area in order to account for how materials effect GPU design. You can't just make a design and expect it to work given an unknown material not chosen and accommodated for in your design.

I think the reason nVidia actually used the high lead bumps and Nomics underfill is just because they adopt new process technology slower than other companies, and this lead to higher risk of failure on those GPU's. And this *particular* increased risk is specific to nVidia GPU's of that era.

 

[Scali]

They definitely have to have expertise in this area in order to account for how materials effect GPU design. You can't just make a design and expect it to work given an unknown material not chosen and accommodated for in your design.

It's not a one-way street.
The nVidia engineers have to be informed and advised by the manufacturing company about the possibilities of their manufacturing process and an indication of the differences between materials.