I think I killed my GTX 1070....

Deleted

 

Did you static shock anything? Are you sure you weren't working on carpet?
Are you using liquid metal anywhere near the GPU?

Get some Deoxit D5, spray it on the video card contacts, spray it into the MXM slot also, then reattach the card and resecure. I had a similar problem on my GT73VR (except it was just low resolution and Code 43, not a scrambled screen). When I used Deoxit, it fixed everything.

https://www.amazon.com/Hosa-D5S-6-DeoxIT-Contact-Cleaner/dp/B00006LVEU/

Seriously, buy this. It is SOOOO useful on electronic parts,

 

Yes, I was stepping on carpet but not working on it.
I did use LM and when I re-paste it with it everything was clean, no sights of spillage at all. I’ve done this many times before...
I will try what you suggest.

 

I did clean the contacts of the gpu and the gpu die with all of the surrounding elements with isopropyl alcohol :



Didn’t help....

I’ve done this so many times before and have never experienced such a problem...
Do u think it still has a chance to survive?

 

Are you absolutely 100% positively sure that you got no conductive balls of doom around any of the IC's or SMD's?
All it takes is one tiny little orb to do some work on that video card.

If you get the Deoxit, make sure you clean the MXM port itself. If the port has some foreign artifacts, arcticlean or alcohol isn't going to do much. Just spray it in there at lowest setting. (protip: always unplug the battery connector when servicing any laptop btw. I've seen youtube disassembly guides where they did that **LAST** instead of first thing...)

 

No, I did use foam dam on first place and there were not any visible trace of leakage...
Yes, I remove the battery first always.
The gpu worked perfectly today after I replaced the fan of the heatsink twice.
Then I thought “ok, not much of a difference with modded fans, I will replace the pads with Gelid ones, will re-paste (first tape around gpu die, the foam dam and then LM)”.
Could this be related with the fact I bought one of those unknown mxm-b from China (black pcb, P5000-layout). But it is two mounts old gpu...

 

I'm sorry, I have no idea.
But the fact that the card started off "okay" then got worse as it warmed up acts like it's a contact issue.
Hopefully Deoxit D5 will fix that.
If there is some other damage to the card, I don't know. I'm not the hardware guy. @Khenglish is. I only know how to put stuff together after repasting

 

That card has failed solder connections to the memory. When it's cold the broken connections are in contact, but as it heats up the card bends slightly, losing contact.

Usually the connections break on the core side, but if you're messing with thermal pads then maybe you broke them on the memory side.

 

I think this GPU can be fixed by resoldering process with special machine for that, or by Heating it up in an oven or heat gun. You can find a lot of videos on YouTube .

 

There's a chance it's just the GPU not touching the heatsink due to thicker thermal pads. Try and check that.

 

Actually the core is touching well the heatsink as the temps just before crashing are in the very comfortable range of 55-60C...

 

Do you think there is any chance for the card to be repaired? And how?

 

Reflowing all the memory chips, maybe excess pressure due to hard thermal pads.

 

Ah damn, that sucks.

Yeah I often worry about MXM cards bending (all of them do -_-). I place old thermal pads under them at the rear edge to keep them straighter. Haven't had any major issues either way.

 

Since the card mostly works it should be repairable by reflowing it in a toaster oven.

If the card has not worked right in the past few weeks you need to heat it to 100C for 5 hours to bake out moisture, or else the core's packaging will pop.

For the actual reflow, start the oven with the card in it on a flat surface. Set the oven to 230C. Turn the oven off 5 to 10 minutes after it reached temperature and crack the door a little to cool it.

You can also line the core and suspect memory chips with solder flux prior to heating. The heat will make the flux flow under the solder joints and help make a better connection.

 

Would you tell me if this is going to be a permanent repair or it will just work for several days/weeks?
I’ve read that the temp should be around 190C, is 230C still ok for a positive result?
Do I have to put the card on flat surface or I have to keep it “in the air” with four alu-folio balls?
Thanks so much for your help!

 

The issue you experience seems to be reminiscent of reflow issues as others described, but I have to ask, which thickness of thermal pads did you use?
If they are too thick, they can cause problems elsewhere with the heatsink not touching other crucial bits properly... which wouldn't be noticeable until you started stressing the GPU (because when you're not stressing it, it works on relatively low clocks).

 

It should be a permanent fix. Unleaded solder melts at 217C. Any temp less than that is a waste of time.

 

Sorry to hear that - Have you tried a UltraSonic cleaner filled with a Isopropyl alcohol only ? It's good for dirty electronics.
When using LM I put also regular thermal paste just around the die to surround the liquid metal - if it touches the paste - it will not touch the board and is easier to smear away with the paste. Try the ultrasonic washer - I rescued a Nokia phone mobo after heavy rain it's working till today. Gratulations of using double copper heatsinks (we talked about this idea in other forum). Good luck !!

 

Thanks!
I sent the GPU for repair and it was successfully repaired - faulty vRam chip was the culprit. They replaced it with a new one and now everything is perfect again.
And most important - now with all cooling mods the GPU doesn’t reach 79C on full max 99% load and +150Mhz core OC - most of the times it is max around 76C. Before on this settings it would easily hit 90-92C and the notebook would turn off.

 

Could you share How did you get those two extra heatpipes done ? Did U bend them manually or are they from another set ? What are their sizes please share

 

There is a guy who made this.
I don’t have all those technic needed to do this job. He is a master, no doubts about it.
As far as I know the heatpipes are bent by default and he soldered them above the other two. The small heatsink is soldered manually too.

 

Could you describe this a little more? My gpu doesnt go past 80c on average, so what other parts are possibly not being in correct contact? How do you determine that? For that matter, how would you tell that a gpu is going out? I previously had issues with crashes when over clocking past 150mhz, however that seems to have been fixed with increasing voltage. Not sure what else to look for.

 

When I had my GL702ZC, I experienced choppy performance on my RX 580 GPU when it was fully stressed. The clocks would fluctuate wildly as a result.
Sure, the temps would go up and be maintained at say 75-80 degrees C (which is what one expects out of a GPU like that when running a demanding game), but the gpu clocks jumped up and down every second producing choppy performance and something like this: /\/\/\/\/\/\/\/\/\/\

Thermal pad thickness can be important because if the pads are too thick, they may cause the cooling assembly to be uneven and not properly touch other components on other portions of the hw.

 

i saw video on aliexpress about bending these tubes with simple vice https://www.aliexpress.com/i/32576952111.html
it looks like it does not need any special equipment. I have read (in other page)that you can bend any heatpipe, but bend radius must be at least 3X tube diameter. so 8x3 tube can be bent around ~50mm diameter conductor safelly. i would preheat them till 100C before bending anyway...
I have second hand alienware gpu heatsink with 3 straight heatpipes, i will try to bend them by my self and test them after that

 

Thats kool!

thanks for the link

@TheReciever You finished that heatsink monster?