Trying to repair an Nvidia Tesla, thermalpad thickness? heatsink?

[azigta]

Hi, reciently I found a Nvidia tesla C2050 on the outside of a building, on the garbage. It has no heatsink. Well it has, since this graphics card has 2 heatsinks. One that covers all the power chips and memories, and other one that goes above this. That one is the one that is missing.

The thing is that I disassembled it and found that some thermalpads were in very bad condition, all of them broken and messy. So I thew them to the garbage.

Now I need to buy thermalpads because I tried putting thermal paste but it doesn´t make contact with the heatsink.

I went to the electronics store but there are a variety of sizes for the thermalpads. 0.5mm to 3mm thick

Do I have to measure the distance? or can I just pick one and it´ll flatten when I push the heatsink?

Also, does anyone know where can I buy a heatsink for this Nvidia tesla C2050? I´ve connected it to a computer with a big fan facing the card and I got into the bios settings with no problem so I don´t know if it works at a 100% of its capacity but it´s not dead , so I want to run tests but without the heatsink I don´t want to conect it anymore... xD

Thanks

[Mephitus]

To clarify, which heatsinks are missing now? If its just the RAM, you might be able to get away with some arctic silver alumina adhesive and single DIMM heatsinks like these: http://www.newegg.com/Product/Product.aspx?Item=N82E16835708012

Actic silver Alumina: http://www.newegg.com/Product/Product.aspx?Item=N82E16835100013

The main processor might be a little more difficult. You may need to just mod a generic one to fit. Such as this: http://www.newegg.com/Product/Product.aspx?Item=N82E16835186063

[wraper]

Hold on with blowing money away for heatsink and thermal pads. Don't put anything at all on RAM, temporarily attach some heatsink on GPU and check (for a short time before it overheated) if it at least boots to windows and works after driver installed. Don't think someone tossed it without a reason, especially considering that person have gone to effort to keep the original cooler. Very often graphics cards with ram issues will work in bios but will start artifacting or crash once driver is loaded.

[wraper]

As of compatible cooler, probably cooler removed from consumer counterpart with fermi GPU will do (search on ebay) or some aftermarket cooler for consumer graphics cards.

[azigta]

Y found a 2 heatpipe gpu cooler on a non working card so I used that for cooling the gpu. I managed to get in bios but It gets too hot and starts to slow down and I´m afraid to fry it (that was before adding this heatsink) so I will leave it as that until the thermalpad that I have ordered arrives. I decided to spend less money and buy a cheap thermalpad, I will try with 0.5mm and I guess that if I need more I will put layers above each other xD

I think I will post a pic tomorrow, it´s a very non good looking way to cool it but I think it will work XDD the cooler from the other card didn´t match the screws so y just bent it a little bit and sticked it with metal tape, I checked that there are no short circuits.

I really think this card could work because the guy who trew this away didn´t change the thermal compound that was behind the gpu lower heatsink.  I explain.

This card has 3 heatsinks xD one it´s the big one, the one that´s missing, That one that keeps the gpu cool. The other one goes below that one, and from the other side of the card too , its mission it´s to keep the power components and vram cool. And then! there´s another one, just above the gpu, like what PS3´s RSX bga has, that metal thing that´s sticked onto the bga with some kind of black adhesive. I got that out and I found out that the thermal compound there was very dry. So I replaced it with fresh thermal compound. Maybe the card had temperature issues, this cards turn off theirselves in order to prevent overheating, so it might be possible that it stills working.

I will continue wrinting about the progress when the thermal pads arrive this week.

Oh, by the way! I almost screw it! I blew a capacitor next to a vram chip but I managed to resolder it... the pads where very small and they were bad , but I got it... xD still working.

Thank you by the way

[azigta]

Well, the thermal pads arrived this week and I found out that out of the bios it only lasts some seconds, so I managed to get into a hiren´s boot utility really fast and I saw some horizontal lines that to my opinion are a clear sign of a malfunction on the gpu. What I think is that at least is not fryed, so I will try to improve my reballing process and I will try to reball this card to make it work.

I used 0.5mm pads, if anyone is interested in that xD

I did a really bad looking but working mod to fit another gpu cooler on this card xD Works great, one little fan on the gpu and a 120mm fan on the heatsink lower the temperatures great. The heatpipes don´t even get hot , they did get hot with a 90mm fan on the top insead that 120mm corsair high airflow fan, so I think I will continue to do my tests with that bad looking but working configuration xD.

This thread can be closed since I will not progress any soon. Thanks.

[wraper]

I saw some horizontal lines that to my opinion are a clear sign of a malfunction on the gpu.
This thread can be closed since I will not progress any soon. Thanks.

Either broken connection between the GPU and RAM or faulty RAM. As of broken connection, it could be just broken balls under the GPU or RAM, or broken connection in the GPU itself (between the die and substrate = faulty GPU).

[OldSchoolTechCorner]

Crack solder balls under the BGA chip, or broken connection between the die and PCB substrate, due to thermal stress and flexing. I use to get them for repair all the time. Same as the 8800 and 9800 series and a bunch of laptops that had a 6600 to 9800, due to material they used couldn't handle the thermal stress, was a very common issue, even with game consoles.

People showing video using heat gun, or oven won't fix it, as only temporary at best. I use Rework station and just replaced the chip and put copper shim and good quality thermal paste, when I put the card back together.

[Mephitus]

OldSchoolTechCorner:

That hardware looks pretty skookum. (Pardon my newbie questions) Is it reballing the chip or just cleaning away the mess? After reworking, what method do you use to reattach the chip? Also, any chance we can get some video of it in operation? I am always facinated by precision machinery.

[OldSchoolTechCorner]

I have to do a video one day of it. It a process. you have to removed all the old solder balls and then use Stencil to align the new solder balls in place while holding the BGA with a jig/holder and carefuly reflow the BGA first while still in the holder, then use a IR/heat gun, or SMD oven to reflow the balls back on the IC. Then carefully align it on the PCB and set temperature profile for ramp rate and let the machine do the rest of the work. The reason for the machine is had zones for pre heater and ETC, to control temperature in steps. So you don't damage, or warped the board and moisture control for a successful reball process.   

You do have to change the BGA in most cases as reason for the failure was due to PCB substrate based, it’s thought by some that green paint used by one of the suppliers, Nay Ya Print Circuit Board (NPC), could have lowered the expected heat resistance or that a certain adhesive used - Ajinomoto Build-up Film (ABF) - might not have been robust enough. When under run failed, it caused excessive mechanical vibration on solder balls and reason they cracked in the first placed. 

[Rasz]

OldSchoolTechCorner:

That hardware looks pretty skookum. (Pardon my newbie questions) Is it reballing the chip or just cleaning away the mess? After reworking, what method do you use to reattach the chip? Also, any chance we can get some video of it in operation? I am always facinated by precision machinery.

here is random one:



and some playlists from repair centers
https://www.youtube.com/playlist?list=PLkVbIsAWN2lsHdY7ldAAgtJug50pRNQv0
https://www.youtube.com/playlist?list=PLbwhgauidRJIJfj0X9EmsS2cCTTJPLj85