Nvidia GPU chip problem

[linears4]

yes, the problem is the unleaded alloy itself, soldering alloy used originally in hp dv series is a very hard alloy, and that suffer a lot during thermal cycles during the working of the laptop, day by day, during these thermal cycles crack easily occurs due to the hard nature of the alloy. this happens also because the chip is not cooled properly due to badly designed contact with the cooling pipe.

leaded alloy is much softer, and less sensitve to thermal cycles and it does not crack easily. of course, improving the contact of the chip with the coling pipe will add more reliability too.

[ian8115]

Been reballing now for a while but frustratingly still getting returns. Not sure if it is inadequate cleaning of PCB tracks, wrong flux or poor quality solder. Using Lead free but what is the best leaded solder? Reflowing HP laptops I have found is a waste of time and 90% will come back within 12months. Most within 2 months.

Thanks

[linears4]

reflowing on that model is not a good thing.

you are experiencing returns on amd or intel model? or both?

if you use jovy machine for the work, absolutely avoid pre programmed thermal profiles... they don't let you to set a very important parameter, the TAL (time above liquid) , if the tal is too short, as in the pre programmed profile, the solder joint will not be reliable.

and also, the thermal ramp is gotta be adeguate, if the temp raises up less than 1,5 centigrade per second, the solder joint will be unreliable too.

and last but not least... i suggest you to use only new bga chip from factory.

why ?

i've done many researc, and in many cases, in the nvidia chips, the defective part is the flip chip itself . The flip chip is the little square at the center of the whole bga.

the flip chip is bonded with the substrate (the green part of the whole chip) exactly like a bga... yes, it has under itself many hard alloy spheres, about 200 micron diameter....

when you do reballing or reflow you are going to heat this small connections too, so, the chip will work for that reason... and even if you do a perfect rebballing and soldering on the board, thi chip will fail in the future cause the problem are not the balls between bga and board( second level connection) but are the small connection between the flip chip and the substrate(first level connection)


many times the problem is this, other time the problem is the second level connection, but it is impossible to figure wich kind of fault are, so replacing the chip with a new one it is the best deal.

[Dan Collins]

Hi,
linears4, what TAL do you use?
I must concur, that we are dealing with the pcb reflow only and the breakdown may be under the chip itself. That would most certainly explain the 50% success rate with these nVIDIA chips. But it is difficult finding replacements for some of these. I can find the 6150, but not many other models in new stock.

Anyone out there have a good source for other NEW nVIDIA chips like the NVS-110 or the G86-620?

I have another related question for the experts out there. Would it be better to strip the lead free solder off a new chip and replace with leaded balls? I assume the new ones have lead free.

Dan

[linears4]

i use 70 second for the small chip about 25x25, max 90 for the bigger chips about 35x35. avoid to go furter, too much high TAL will do the solder joint brittle.

i alaways take off the unleaded solder balls and put a new set of leaded ones. But just to avoid problem with unleaded soldering process. Making a good and reliable solder joint with unleaded balls is a very hard challenge. if you intend to use the original unleaded balls you have to bake the chips before the installation, if you fail to do so you will certainly have some internal delamination due to the higher temp. The moisture trapped inside the new chip will damage the chip itself, even if it appear perfect after the soldering process.

from the other hand, changing the original solder balls will add a thermal cycle to the chip, increasing the chance to damage it or to make it unreliable in the future.

bga and mcm rework it is a very hard thing...


good result is the combination of many many factors

[linears4]

i have 100% welding success, with any kind of chip until 35mm, but only at zero time, zero time means after the welding process.

my biggest problem is the long term reliability.

in some cases the chip fails again after about 6 months. in other cases i have some chips working even after one year.

i'm trying to figure the cause, but it is very hard.

i know too that if i fail to do baking i will encounter some problem, but baking process is very time consuming.

a must is to bake the new chips, for 8 hours at 85-95 degrees at least, and then make a moisture barrier package. the theorical floor life of this kind of components is maybe 72 hours, after that a new baking is needed.

but very time consuming process.

even the whole board has to be beaked, to avoid cracks in it during welding.

but all this is a crazy thing...

but a thing is shure, only reflow will never give a long term reliability, cause you don't have flux on the pads, and a reflow can't change the penetration of the soldering alloy to the metal pad, it can only adjust a crack, but if the problem is on the interface a reflow can't change the things too much

[Dan Collins]

linears4, are you suggesting that baking is important for any embedded chip due to moisture? Interesting conclusion. Sound like there are a lot of factors to take into consideration. I will start doing some research on my own. My Brother-in-law is a solder paste chemist. I will run some of this past him and post any findings

Tnx

[linears4]

yes, expecially with multi module chip and lead free, all the graphics chips today are multi module chip, whit flip chip on it. Bga's are completly different thing, on the construction side.. for example, southbrige intel 82801 series are bga... nvidia gf6150 are multi chip module's. all the black plastic chips are bga... the green ones like nvidia and ati, but even intel northbridge are multi chips.

baking is not my conclusion, there are ipc standards that say that... j std- 033 and so on..

bga are less sensitive by moisture, multi chips are very sensitive to it... if you store the chip in a moistured enviroment the moisture will damage internally the chip during the soldering process. maybe the defect will be latent and will come out after many months.

[Dan Collins]

Oh, I see then, for me here in a very dry desert climate and all components stored in AC enviro, the moisture concern is minimal.

My RE-7500 arrives in 3 weeks from Jovy rep in Flordia. I have been doing BGA work for years with other methods. Really looking forward to using the infrared. I have spent many hours reviewing posts and videos and have a pile of junk boards to do much experiment on before I put system in production. That is why many of my questions get into specifics.

I do have a question regarding use of RE-7500 that is not clear in posts I have read, and that is the height of the upper heater. It would seem to me that the goal is to supply heat that comes as close as possible to the setting and T/C readings as possible.
What is the optimal height setting for the RE-7500? Does it vary for different tasks?

TNX
Dan

[laptoppartes]

hi, i own a small laptop repair shop in el salvador, central america. i have been fixing this issues with the classic hot air gun. i recently read a GREAT story from the inquirer explaining the cause of the nvidia bga problem. i must say that u guys are doing a great job giving your techiques to many of us who are interested in fixing this motherboards. my technique right now is the usual heating of the chip for about 90 seconds at aprox 150F ata a distance of maybe 2cms. then we place a copper piece for it to make contact with the heatsink. but as u may imagine the customers come back sometimes after 3 weeks, sometimes after 2 months. i have been web-searching a system to cool down the gpus and i found a thermoelectric cooling system wich uses a small plate with one hot side and one cold side (peltier cooling system), this morning i bought my first thermoelectric system at a price of $6 on ebay and will receive it in 7 days and i cant wait to receive it...i will certainly post the results here.
someone was asking for a website to buy the chips i found some in alibaba.com...i purchased 2 f500/f700 motherboards from them and im satisfied.

[linears4]

no, no absolutely no! if you want to do reflow only, you have to see if you have done it correctly! you have done only a chip heat... if you want to do a reflow, you have to rework the chip , and you will have to check if the reflow is done properly. for doing so you will check with a tweezer if the chip become loose during the reflow process..

in other words you have to check if the solder balls will liquifiyng during the reflow, for doing that you must to tap the chip on the side, to see if the solder balls reached the liquifing zone.

but as i sayed , reflow is a waste of time, it does not add reliablity

there is no way to do so with an hot air gun, believe me or not... you need power, and at least you need a good infrared or hot air station. you need preheat, there is no chances to melt the unleaded solder balls with a simple hot air gun

[laptoppartes]

thx a lot for your help.
my question is now why does heating chip temporarily fixes the problem, if the solder isnt liquified with the heat gun, why does work? and i also have another question and is if the cracked solder balls are between the silicon chip and the substrate( not between the substrate and the motherboard) why does everybody focus on the motherboard-chip area?

and one last question...why does the wireless card stops working before the video goes out?

thx a lot for your time and responses

[Dan Collins]

laptoppartes - I have used alibaba, but I never get a reply from the sellers.

Sometimes the USB ports also goes south before the video.

And I think you answered your own question as to why the short life on repairs because you do not reball. I would be interested on your conclusions from the thermoelectric heat sink mod.

Thanks

[laptoppartes]

thx for your response. what i actually meant with my first question was to know what effect does heating the chip has, cuz it does help a little, when i do this repairs i only give the cust one month warranty on the repair, i charge between $30-$50 depending on how good is the customer (regular customer or stanger). if they dont answer you in alibaba is probably because they are in china and they are all sleeping right now. the msn of my seller is [email protected], she is a nice salesgirl. as i was saying, i purchased 2 f500/f700 motherboards from them at about $105 (shipping not included) , they are used but they work

[linears4]

Originally Posted by laptoppartes

thx a lot for your help.
my question is now why does heating chip temporarily fixes the problem, if the solder isnt liquified with the heat gun, why does work? and i also have another question and is if the cracked solder balls are between the silicon chip and the substrate( not between the substrate and the motherboard) why does everybody focus on the motherboard-chip area?

and one last question...why does the wireless card stops working before the video goes out?

thx a lot for your time and responses

solder carcks doesn't mean a gap of miles...it means micron level gaps... when you heat the whole chip, thermal dilatation of the balls occurs so these micron gaps are temporary gone away... and the connection is almost restabilished due to the termal dilatation of the heated balls.

but, successives thermal cycles of the chip during the working create various dilatations of the solder balls. the balls increase it size, and decrease it size day after day, hour after hour.... but the crack is there, so when the gap become newly too big... game over... no working motherboard.

the cracks are not alaways between the chip and the substrate, there are no simple method to figure where is the crack, only a scanning acoustic microscope can tell the true, but we all human mortal can't even touch those kind of diagnostic machine...

so we can only think that the crack is between the chip and the motherboard and do reballing...

this is the reason why it is advisable to use new chip.

or, if we have various similar fault whe can have a statistical approac... trying to reball the faulty chip and look if the work last


response to the last question :

the wireless card stops working when the faulty chip is the northbridge, not the video card. for example on hp dv series with amd processors.

when the northbridge is going out, the first sign is the lack of the wifi functionality or other types of faults not correlated with the video section.

when the northbridge is totally out , black screen occurs, but not due to the video chip, but only due to the northbridge chip, that controls the video chips too.