[SOLVED] 1080ti with kraken g12 crashing

Mar 10, 2020
10
1
15
Hello ! So this past few days I have encountered quite a problem with my gpu.

Got a Zotac 1080ti extreme amp and everything was fine.
Saw a chance and got the kraken g12 with an evga 120 water cooler for it , installed it and the temps were great. The problem appear while gaming , after like a minute or so , the screen freezes and the game is closed or I'm force to close it. In the event viewer I got the nvlddmkm 4101 error.

So I decided to put the original block just in case and it still happens ..

I did a clean installation of drivers and didn't work .. did a new windows installation and didn't work either .. disconnected a few things and still the same ..

Sorry for the bad English and thank you !
 
Solution
Remember, the VRAM in a 'stock' configuration is being cooled by part of the overall cooler - so there is quite the surface being used.

So, any decent heatsinks will be pretty decent in height to create as much surface area as possible.
Example: https://www.amazon.com/Alphacool-Co...d_r=0P7GHQE40JFBYPFNNN44&psc=1&qid=1583954882
On top of that, passively cooling (ie just the heatsink) is probably not sufficient. You'll want to come up with some way to...

Barty1884

Retired Moderator
The G12 is only fully compatible with reference PCBs. I don't see any major differences between the FE and the AMP Extreme, bar the heatsink on the power delivery.
Did you retain the heatsink? Or did you remove? Perhaps airflow alone (from the G12) is not sufficient to cool the power delivery. ....... I'm assuming you've connected the G12's fan to a motherboard fan header?

The other potential problem with the G12 (and others like it), is uneven mounting pressure - potentially warping the PCB itself - or worse, cracking the GPU die itself.

The fact the issue persists with the stock cooler is concerning, and may be an indication that lasting damage has been caused.
 
Mar 10, 2020
10
1
15
The heatsink went of completely , only the backplate is on because is not removable I think. The fan is in the cpu header at full speed. I have another fans blowing direct cool air into the gpu Wich is mount in a vertical stand (I tried mounting it without the stand and is the same result always). The problem is the auto Boost of the gpu that makes it crash mostly , the temperature is always within range . I have make some adjustments with afterburn with the clock and the speed -100 both and it kinda works but I get less performance and is kinda annoying. I checked the mounting and didn't found anything weird but maybe I have made a mistake doing all this sadly
 

Barty1884

Retired Moderator
the temperature is always within range

Temperature likely has nothing to do with it (at least, in terms of what you can monitor in software).
With an AIO attached, the GPU die temp should be fine - it's the temps of other components (power delivery and VRAM) that are more likely to present a problem.

As a test, leave your GPU clock untouched in Afterburner, but downclock your VRAM. See if the issue persists.
In the stock configuration, the VRAM has a thermal pad to the overall heatsink - and active airflow as a result.

With the G12, the airflow is focussed on the power delivery, and the VRAM is left pretty much to bake - which has been a fairly common occurrence on recent GPUs in their stock configuration, let alone leaving it completely uncooled.



12345.png


View: https://youtu.be/gqK7XqTuPwU?t=174
 
Mar 10, 2020
10
1
15
As a test, leave your GPU clock untouched in Afterburner, but downclock your VRAM. See if the issue persists.
In the stock configuration, the VRAM has a thermal pad to the overall heatsink - and active airflow as a result.

how do i downclock the vram ? i think i activated the option in afterburner but it doesn't show it anywhere
 

Barty1884

Retired Moderator
There's no independently VRAM voltage control. As far as overall "core voltage", that can be 'unlocked', but for most modern cards it only allows an increase, not a decrease.

Lowering "memory clock" in Afterburner will reduce VRAM clocks (and voltage by default).
 
Mar 10, 2020
10
1
15
So I lowered the memory clock by 100 just to check and open the division (Wich it crash after few seconds ) and everything seems fine , didn't stay long , didn't crash for a few minutes. I'm gonna do a longer test later just in case
 

Barty1884

Retired Moderator
I'd expect the VRAM will eventually heat-soak regardless, so for a longer-term test, you might want to downclock a little more (say -200).

By the sounds of things though, the lack of any real cooling on the VRAM is where the problem lies.

Normally, you could just point a case fan directly at a component that's overheating - but with the mounting plate used for the G12, it'll block direct airflow over the VRAM.

Not sure if there's any room (with the G12 installed) to install small heatsinks on the VRAM at all...
 
Mar 10, 2020
10
1
15
I was thinking about getting some heatsinks for the vrams today , the only problem is the size , I'll have to check how much space is left between the g12 but I think is quite enough
 

Barty1884

Retired Moderator
Remember, the VRAM in a 'stock' configuration is being cooled by part of the overall cooler - so there is quite the surface being used.

So, any decent heatsinks will be pretty decent in height to create as much surface area as possible.
Example: https://www.amazon.com/Alphacool-Co...d_r=0P7GHQE40JFBYPFNNN44&psc=1&qid=1583954882
On top of that, passively cooling (ie just the heatsink) is probably not sufficient. You'll want to come up with some way to direct airflow over those heatsinks too....
 
  • Like
Reactions: Syberyandog
Solution
Mar 10, 2020
10
1
15
Sadly in my country is quite hard to find this type of things , the ones that are close to be are for raspberry type , some bronze ones with 12-13-5 and aluminum ones of 15-15-15 ..
 
Mar 10, 2020
10
1
15
So I got the heatsinks and install them already. I put every setting in normal parameters and try to test The division . Everything goes perfectly, I played for around 40 minutes without any error or crash so the vram cooling was definitely the problem .. thank you very much for the help !
 
  • Like
Reactions: Barty1884