R9 390 Nitro PCB, Kraken G10, Kraken X31

[Symcy94]

Recently I bought R9 390 nitro GPU with backplate, and decided to upgrade it with water cooling.
I bought Kraken G10 and Kraken X31 from NZXT, so install it on my GPU. But that G10 + X31 can only cool down GPU core, so I decided to buy few heatsinks for other components on my GPU PCB.
There is 80mm fan on G10, that's blowing cold air directly to VRM, but I just wanted to even increase cooling so I installed heatsinks there also.
After 5 minutes of GPU load on 100% my GPU crashed. Monitor lost singnal and screen went black.
I was monitoring temperatures in GPU-Z. Core temp was below 60°C, VRM1 on around 80, and VRM2 around 87 peak.

With stock cooler I had almost same temperatures on VRM. Core was on 75°C, but VRM were around 80-85.

So it seems that something else is overheating and causing GPU crash in my case I think.

http://s15.postimg.org/990ifcvkb/03_bigee.jpg
http://s15.postimg.org/is5lv2tor/2016_03_25_13_51_33.jpg

One picture is my GPU with heatsinks applied, and another is downloaded from web and I marked where I put my heatsinks.
Should I put them somewhere else also?
Or does anyone have answer to my question, what went wrong, and why is my set up not working?

 

[brandonmatic]

Did it crash at default voltage and without an overclock?

 

[Symcy94]

Did it crash at default voltage and without an overclock?

Yeah there wasn't any overclock.
It was running benchmark okay for 5minutes, then it crashed.
Not sure what it was :\

 

[thilanliyan]

Point a fan to blow over the entire card and see if that helps.

 

[someEEguy]

Not clear, does the card still function or is the monitor always black now? If the latter is true, was the thermal compound you used non-electricity conductive?

 

[Symcy94]

Point a fan to blow over the entire card and see if that helps.

Did that and still the same.

Not clear, does the card still function or is the monitor always black now? If the latter is true, was the thermal compound you used non-electricity conductive?

Well monitor is working fine, also cable was in as it should be.
After force restart PC, it was working fine. But only untill I ran game or benchmark. So if load was on 0% it was okay, but under load it crashed in few secounds or max 2min.

I used "Nano Grease" thermal paste for GPU core.
For VRMs and RAM cooling I used those stickers that I got together with heatsinks. Didn't apply any thermal paste on those because it would be messy.
I did put back stock cooler and card is working fine. Was running vally benchmark for 2 hours at 100% load ... didn't crash or anything.

Also I would like to mention that since normal stickers didn't do the work on VRM line, I used that black thingy it cam together with card.
https://youtu.be/q1TdiGU6Rdk?t=1m31s
Check that youtube video at 1m 31sec ... some black thingy that he pulls off, and then stick it to backplate. That's exactly what I put on VRM and then applied heatsinks on it. It was holding a bit better, because when I tried to use just stickers from heatsinks, it would just fall off, because VRM chips are so small, so it just wouldn't stick there.

 

[Jhatfie]

You might want to double check and make sure that the X31 is applying even pressure on the gpu die. I doubt that is the issue, but depending on how it was tightened down, maybe the die was getting uneven cooling?

 

[Symcy94]

You might want to double check and make sure that the X31 is applying even pressure on the gpu die. I doubt that is the issue, but depending on how it was tightened down, maybe the die was getting uneven cooling?

Well core temp was always below 60°C, and CPU core is acctually in the midle under that if I'm correct. I know that my GPU core on top wasn't fully covered, but temps were okay all the time on GPU core.
Kraken X31 was tight on it and was colling just as it should.

I think something else is overheating, I just don't know what. GPU-Z only monitor VRM1, VRM2 and GPU core ... all temps were normal, that's why I think something else might be causing problem.
Just no idea what exactly.

 

[thesmokingman]

That is not a good setup on your vrms! You can't get even distribution of force on the vrms. Get a heatsink block and cut it to fit. You have the hole support for a proper block so use them. You're vrm sensor readout of is belying the fact that an individual vrm phase or more is overheating.


Doing it like a pro:

http://www.overclock.net/t/1511914/sapphire-r9-290-arctic-accelero-xtreme-iv-vrm-mod

 

[Symcy94]

That is not a good setup on your vrms! You can't get even distribution of force on the vrms. Get a heatsink block and cut it to fit. You have the hole support for a proper block so use them. You're vrm sensor readout of is belying the fact that an individual vrm phase or more is overheating.


Doing it like a pro:

http://www.overclock.net/t/1511914/sapphire-r9-290-arctic-accelero-xtreme-iv-vrm-mod

Woow, that guy realy did a great work with it
I'm just not that handy with things like that custom making aluminum stuffs xD
But I do know someone with lathe, so he might be able to help me out with it.

 

[thesmokingman]

Woow, that guy realy did a great work with it
I'm just not that handy with things like that custom making aluminum stuffs xD
But I do know someone with lathe, so he might be able to help me out with it.

You don't need a lathe, you use a dremel. Those heatsink blocks can be bought off ebay or amazon. Get a sheet of paper cut it to fit the space, think template. Then transfer to heatsink, cut with dremel, cut fins down at the mounting holes, etc.

*for ex. for 7 bucks

http://smile.amazon.com/Omall-20MM-Aluminium-Diffusion-Cooling/dp/B00UVPT0CS/

 

[Symcy94]

You don't need a lathe, you use a dremel. Those heatsink blocks can be bought off ebay or amazon. Get a sheet of paper cut it to fit the space, think template. Then transfer to heatsink, cut with dremel, cut fins down at the mounting holes, etc.

*for ex. for 7 bucks

http://smile.amazon.com/Omall-20MM-Aluminium-Diffusion-Cooling/dp/B00UVPT0CS/

Well that's true.
But still need to figure out which areas shouldn't be touching aluminum directly since it could fry everything if I don't do it correctly.
And another problem is that I'm from Slovenia, so here it's hard to get all material needed for that project.
I'm already writing emails to companys that are custom editing aluminium upon my request. Will see what is estimated price for it. If it will be too high, I will just try to do something on my own

 

[thesmokingman]

Well that's true.
But still need to figure out which areas shouldn't be touching aluminum directly since it could fry everything if I don't do it correctly.
And another problem is that I'm from Slovenia, so here it's hard to get all material needed for that project.
I'm already writing emails to companys that are custom editing aluminium upon my request. Will see what is estimated price for it. If it will be too high, I will just try to do something on my own

Oh sorry, I didn't look at your location. Well, worse case you can keep using the modules you have. Just get some thermal tape, instead of that strip. Then you fashion a bolt plate fastener, like a sandwich like below but without the teeth of course.

 

[thilanliyan]

You don't need a lathe, you use a dremel. Those heatsink blocks can be bought off ebay or amazon. Get a sheet of paper cut it to fit the space, think template. Then transfer to heatsink, cut with dremel, cut fins down at the mounting holes, etc.

*for ex. for 7 bucks

http://smile.amazon.com/Omall-20MM-Aluminium-Diffusion-Cooling/dp/B00UVPT0CS/

That is cool but I'd be hesitant to do that. One small mistake with the aluminum touching a resistor you missed or anything like that and the card could be dead.

 

[Silverforce11]

Temps are fine for GPU and VRM, but the temp sensor is not universal on the GPU surface, take it apart and look at the TIM, it sounds like it may not fully cover the die properly.

I have done ghetto AIO mods where I had the same issue.

 

[thesmokingman]

That is cool but I'd be hesitant to do that. One small mistake with the aluminum touching a resistor you missed or anything like that and the card could be dead.

You're in the wrong thread then.