Is my EVGA 1080Ti SC2 now broke? [again?]

[FlyingFish]

https://forums.evga.com/m/tm.aspx?m=2573413&fp=1

 

[Aretak]

To me, it appeared that the paste was missing on the top/right-hand side of the die?

is there something wrong with the paste? am I doing something wrong?

Are you following proper procedure when putting the mounting screws back in? You should be putting two in at diagonally opposing corners and tightening them just a little, then adding the other two screws and then slowly tightening each one bit by bit on opposing corners to ensure even contact. If you just stick one screw back in and tighten it all the way, then move onto the next, or put two in on the same edge and tighten both down, it's going to cause uneven pressure and make the thermal paste spread out unevenly.

 

[azazl187]

Put it in the oven

 

[BS Dave]

That would make more sense to me rather than the paste is bad. Comparing the card to ther EVGA 1080Ti SC2's I cant see anything out-of-place - The pads seem to be in the normal places (a picture from overclock3d.net);

If the cooler wasnt sitting correctly, that would explain a lot. I have never removed the base-place. The pad on the far right did tear in half once, but I put it back on (should be okay?). I did notice there is a screw directly right of the die, but that doesnt appear to be sticking up? The cooler has 3 cables (2 for fans, 1 for lighting) which get replugged in each time, im pretty sure none of those cables are trapped.

ps. here's how the cooler looks after cleaning;

[update] Ive been spent this morning looking at the pictures to see if I could see anything making the cooler not sit correctly. I was looking to the right of the card (towards the power connectors). Couldnt really see anything.. maybe the led cables... but there is a cut-out for those cables. None of the pads look wrong to me.

The heatsink fin you said about doesnt look as bad in the picture with MX4 on, could it be the flash on the camera?

I started to wonder if the paste was being pushed to the right, like the cooler isnt sitting correctly on the left (towards the bracket). The only thing I could see was some pads under the base-plate;

Would these be enough to make the cooler sit poorly? I guess they should be on-top of the memory chips, but they do look like they are nearly on-top of something else.

Does removing the base-plate mean removal of the back-plate too?

I dont have a clue, to me nothing looks majorly wrong.

[Update] Today I took some pictures of the card assembled;

It does appear to be a slight sag, it looks worse in the pictures (when you draw a straight line, it doesnt seem as bad). The card is installed vertical nowadays, so hopefully sag wont be so bad.

picture with the purple box. I was looking at that too. If you take off the backplate do it on a big bit of paper then draw boxes around the card where each screw goes-leave screws in boxes maybe use blutack too.

you could always try to pick up a compatible water block too 2nd hand. Is it a reference design?

 

[Roguey000]

hi guys, thanks for the advice. Today I took the card apart; cooler, base-place and back-plate and checked over the card. I couldnt see anything bad; although im not sure why the back of the PCB had be some kind-of oily stuff (around the gpu area). I did clean some of it off;

That wouldnt be the result of an old AIO leaking above it? I dont use the AIO any-more, when I changed cases I changed to a 360mm from a 280mm AIO. The 280mm does have some age on it - to be honest. The sticker on the right more or less fell off, as it wasnt sticky any-more (looked it like was soaked in whatever was on the rest of the board, dried and so on). Hopefully its not something more serious? I guess it could be the new AIO either? or its nothing?

I do think a couple of pads need replacing, I guess no-one know which pads I should use (this card uses 1.5/2mm pads)? Pad 2 ripped apart awhile back, and I should really replace it, and pad 1 could be better;

I did give the GPU a good clean and tested with small blobs of paste. You guys were right, it didnt need no-where near the amount I was putting on. I just was getting fed-up keep replacing thermal paste every few months. Hopefully this is 'fixed' this time.

 

[FlyingFish]

So are you back up and running? Good temps and clocks?
The oily marks on the pcb, are from the thermal pads and are quite harmless and normal. As with any cleaning of components, take care not to damage anything. Somethings don't need doing just for the sake of it, especially if it risks damage. I've heard of plenty of people that have 'knocked' components off trying to clean stuff.

 

[Roguey000]

So are you back up and running? Good temps and clocks?
The oily marks on the pcb, are from the thermal pads and are quite harmless and normal. As with any cleaning of components, take care not to damage anything. Somethings don't need doing just for the sake of it, especially if it risks damage. I've heard of plenty of people that have 'knocked' components off trying to clean stuff.

phew, I thought it might of been something leaking. Glad its normal due to the pads, and not the AIO.

Well the card is fine atm; within 1c of my last repaste and boosting to the same levels. Hopefully it will stay working, as its done this a few times (looks like it fixed, then messes around later on).

Do you think its worth renewing pad 2? and if so, changing the paste to something like Kryonaut over MX4? Kryonaut says max of 80c, surely a gpu would be getting up there? I guess we have no idea what the max of MX4 is?

thanks everyone for your advice!

 

[BS Dave]

when changing the pads maybe look at the edge of a pad that isn't being squashed

 

[Roguey000]

Well its been over a month now, and its now started crashing again.

I bet I can re-apply the paste again and it would last another month or two, but surely each repaste should last longer?

So what do you think I should do? I really dont want to keep re-pasting the gpu. Im still thinking something is bad with my MX-4.

I think I will move the card into another pc later and check if it does the same thing; I wouldnt be surprised - ive done all this before. When it does the same thing, ill disassemble the card and check the paste again. I dont want to simply re-apply the paste and get the same problem in a month/two time. I want to find a fix for it. I did look at pad before like;

Carbonaut Thermal Pad - 25 × 25
https://www.overclockers.co.uk/thermal-grizzly-carbonaut-thermal-pad-25-25-0.2-mm-th-02e-tg.html

Im not keen on a pad, but hopefully it would last longer?

Another option get some Kryonaut or something?

When I put the card back together last time, I did cleaned the gpu die and applied (a small dot in the middle). I did check to see how it spread and it was fine. I bet the die is missing some in a certain area now. Ill take pictures after the disassemble.

 

[Roguey000]

Okay, well today I removed the card and stuck it into another pc. Same thing happened, crashed to CTD. So that removes the possibilty of the MB, PSU, Windows etc. for me.

So being at this point ...again. I wanted to check under the cooler to see what the sitution is. I thought if its like last time, some of the die will not have paste on. This is how the die looked when I removed it in Jan, and today. In both instances the die is missing paste on the right-hand side;

Before Jan I had spread the paste out, covering the whole die (admittedly I used a lot of paste, but this was like my thrid or fourth repaste at this point - getting fed-up repasting it). In both cases you can see that thermal paste had been pushed out, there is a square outline on the cooler of paste;

When it was redone in Jan I put a pea shape in the middle and checked it after to see if it was spreading correctly. It created a nice circular pattern.

Any ideas why the paste keeps disappearing on the right-hand side of the die?

A big thanks for your help!! I really want to get this thing fixed for good!

 

[Benass]

the pressure between heat sink and gpu chip is uneven. Thus thermal paste. Were the pressure is strongest thermal paste gets squeezed out. look at both times its similar pattern. reading your issues i think you have cracked soldering joints under gpu. if im correct your best bet would be to reflow gpu chip.

complicated procedure. defo needs professional person with bga station.

Im learning about gpu repairs and had a fair amount of success thermal shocking few gpus. i managed to fix 7970 and r9 270. i simply heat up the chip for 400c with air gun for 1min 30 sec. and luck would have it, it worked. But it is temporary solution, similar as you repasting it.


In order to fix is permanently u need to reflow the chip. Saying that, the chip has two levels of solder balls, but only one level can be replaced. If u unlucky and its second level balls are cracked, then bye bye chip.


if u want to know more about bga reflow, let me know.

 

[Roguey000]

the pressure between heat sink and gpu chip is uneven. Thus thermal paste. Were the pressure is strongest thermal paste gets squeezed out. look at both times its similar pattern. reading your issues i think you have cracked soldering joints under gpu. if im correct your best bet would be to reflow gpu chip.

What would cause cracked solder joints? Ive had the card from new, and it was brought here.

When I re-apply the paste the card works fine for awhile (running for hours without any problems, boosting to 1949'ish, staying under 70c), until something happens to the paste (a month or two later). Wouldnt the card CTD all the time if the joints were cracked?

I dont want to come across disrespectful, but trying to understand.

So far making feel like the best option would be to try a pad?

 

[FlyingFish]

Looking at the photos, it looks like there is uneven contact between the chip and the heatsink. It looks like the heatsink is not parallel with the circuit board, hence the large amount of compound on one side of the chip.
You need to try and figure out why this is happening. Could it be that the heatsink block is sitting proud on something? Perhaps a thermal pad too thick? Or the heatsink not mounted perfectly over the circuit board on its fixings somehow?

 

[Roguey000]

You need to try and figure out why this is happening. Could it be that the heatsink block is sitting proud on something? Perhaps a thermal pad too thick?

Would I be looking for something proud on the left or right side of the die? the thermal pads on the card are the orignal. If something is proud, how much of a height am I looking for? 1mm? 0.5mm?

Why does look like no paste was added on the right-hand side? is the cooler coming down so low, its forcing the paste to the left? In the Jan picture, the paste was spread out across the whole die;

After that I did the pea method, and it looked similar after a month.

Many thanks all!

 

[FlyingFish]

Quick sketch on my phone. I'd be looking at something on the left (in your picture) that could be keeping that side of the heatsink slightly lifted off the core. If it was a nice good square contact you wouldn't have all that paste on the left hand side, itwould all be squeezed out evenly.

 

[FlyingFish]

Just had a thought, could it be how you're tightening the heatsink back on? Nip the screws up gently, then tighten each in turn by small amounts, so the heatsink is secured squarely. Normally done opposites.

 

[Roguey000]

Thanks for the diagram, that helps me a lot. I will take another look at it, how big would the obstruction have to be? 1mm? 0.5mm?

Another question I had was about the side with good contact. Are you saying because there is good contact on the right, it is forcing the paste left? or is the paste being squashed so much, it doesnt need much on the right-hand side? surely its bad to see the letters DIA after removing the cooler?

oh and btw, with the screws - yes I do them in a diagonal pattern, and do like a quarter turn on each.

many thanks,

 

[Aretak]

Some coolers just don't make good contact no matter what you do. Gamers Nexus has done a lot of mounting pressure testing over the past few years, as well as measuring the flatness of the coldplates on coolers, and there have been some real bad examples. It doesn't have to be that the cooler is necessarily being obstructed by anything. A further random element is that the die itself won't be exactly flat, and some are more flat than others. The gaps we're talking about here are are lot less than a millimetre. Gamers Nexus have been measuring in the tens of microns range, and even gaps that tiny are enough to make a difference. It may have worked perfectly fine from the factory as the stock thermal paste tends to be quite thick, both in terms of application and viscosity, but then as it broke down over the years the contact worsened. The thermal paste you're using looks quite runny, which would make the pump out effect worse with uneven contact.

 

[Roguey000]

Well today I spent a few hours looking over the whole card - taking it apart and checking it for any clearance problems. I couldnt really see anything - especially towards the back/bracket side of the card. The only thing I could see was a thermal pad at the top (above the memory), so I removed it for now (it makes contact with the base-plate and the cooler). After that I re-pasted the die and spread it;

I dont think that pad is really important to temps 'n that - after an hour or so the memory temps seem to be reaching around 76c acording to ICX. The pads under the base-plate are still in-place, and the memory is still being cooled;

Hopefully I wont be posting on this thread in a month time saying the card gone down again, but wouldnt be surprised (after the amount of problems ive had with this card). I wanted to show you how the paste was applied. The card is working fine; no problems with even +20% to the power-target. Before when I increased the power-limit it would CTD after minutes.

 

[Benass]

Solder balls cracking from thermal cycle. Its simple, metals tent to expand when heated up and contract when cooled down. Thats just how metals behave. Eifel tower is few meters shorter in winter comparing to summer.

So this thermal cycle sometimes breake solder balls. Its a lottery game can happen to any one any time. there is a history behind it starting with gtx 8800 many years back. since then both amd and nvidia improved solder technology, but it still can happen.

i can give u another video of showing the guy reflowing the chip. It will give you better understanding on whats involved.