GPU temp problems

Soldato
Joined
18 Jan 2007
Posts
19,845
Location
Land of the Scots
Hey guys, need a little advice on my loop. Bit of a long winded synopsis but the more info the better eh. I have 2x 290s in crossfire in my loop and usually their temps are around 50 - 60 in a long gaming session. Recently I have been playing the division a lot which does not support crossfire so I've only been using one card.

I popped out of game the other day to notice that my GPU had actually hit 85 degrees. Cause for alarm no doubt. I had a little look around and tipped the case a bit thinking maybe there was a bubble stuck, but I was running extra volts so I dropped them down too. After that temps seemed still higher than normal at 70+. I tried running another game (GTA) which showed the GPU temp at 60 degrees so I figure there was some quirk in the division running my GPU hotter and figured I would look into that.

However since then even with no extra volts the division was starting to run the card to 85 degrees again. So I tried GTA again, I noticed two things however, fristly I was running GTA in crossfire and the temps were actually showing on the first card as 60-65 and the second card 40-45, an exremely unusual difference. I tried GTA without crossfire and it was running the card to 70 degrees. Not normal at all.

I ran a few other games in crossfire and tried a few different drivers to troubleshoot and had a feel about in the case, the the first card's block felt cool to the touch the whole time, but the second card felt warm as I would expect.

For now I have switched the video output so the main card is the second one, back in GTA it was running 40-45 degrees. So I'm of the opinion that there is a problem with the first block but not sure what.

So here's my logic: If there was a problem with the block such as microchannels being blocked then the block would get hot and not dissipate heat into the water. Therefore the block would be very hot and not cool.

If there was a problem with restriction in the first block I would also expect this to manifest higher temps in both cards as the flow rate would be less to both.

My thinking is that it may be a problem with the TIM and it's not transferring heat effectively from the GPU to the block. I'll be doing some more tinkering tonight but won't really be able to drain and dismantle until the weekend so hoping other people can give their opinion in the meantime.

I previously used GELID GC-3 on the card and if I'm going to take the block apart I'll have to replace that anyway so what to people recommend for TIM on the waterblock? I bought the gelid before I watercooled and seem to remember that it's really better at higher temps so guessing there may be something better at watercooling temps.

Thanks for reading if you made it all the way through :)
 
Sounds like the block. Be interesting to see exactly what is is. Its more down to amount more than anything. Line down the GPU die 3mm from each end.


Gelid gc-3 has been the best stuff I've used for my gpu's.
 
Last edited:
Right guys, a little more help needed.

I replaced the blocks at the weekend and things seemed better, however I did notice that when I was refilling the loop the flow rate seemed low, I took it apart again and checked the gpu blocks to see if there was any restriction there but they were fine. Temps seemed fine for a bit but then started creeping up again.

I think the problem is the pump...

When I originally built the loop if I turned the pump up to 4/5 it would churn the water up with bubbles as if it was too high and going any higher than 1/2 would not affect temps by more than a degree or so.

I did notice before I took the loop apart that it was no longer doing this but thought nothing of it at the time. It still doesn't do this now but instead when I have the pump at 1/2 temps are getting up to around 60 degrees, If I turn it up to 4 or 5 the temps now drop back down to around 46ish degrees.

Is it worth taking it out to give it a clean or is it just on it's way out and time for a new one?
 
im not sure i understand what you mwean by "churn the water up with bubbles". when i got my pump res combo it had some things to insert into the res to stop vortex' (es or ' not sure lol) maybe you need one of them inserts? there was a little sponge thing and a "hd tube" plus a big EK logo insert.
 
When ya say churn the water up with air bubbles, are you sure its not just pulling the air bubbles out of a a block or rad, them bubbles can get pretty lodged in there and require that extra speed to drag them out, either that or a lot of tilting and knocking on stuff. If ya running a d5, and ya mount it on foam and rubber fixings its pretty inaudible unless ya put your head right next to it anyway (at least mine is)so ya might aswell keep it on full speed.

Btw i used to run the second card as the main one for a while as it was easier to position a cable to at the time because of rad mountings but it caused endless overclocking problems, it took me ages figuring it out what the problem was, it just didnt seem to like the 290x in the second slot running as the primary, even if i swopped the cards over. Just thought i better mention it incase you start getting random instability blackscreens etc and cant figure out whats causing it.
 
Last edited:
The bubbles didn't seem to be coming from anywhere, I think it was sucking them up from the top of the res maybe. Was difficult to say, but there weren't any bubbles coming up the tubing into the res (as far as I remember). As I said though it's not doing that any more. Currently have it on 4 and can't hear it so shouldn't be a problem.

I started using the top card as the main card again as the second card was actually pretty unstable, not sure if that's the same problem you had as when I bought it from the MM the original owner could not get it to boot into windows, I installed it as the second card no problem. When I used it as the main I had the same issues.

Did a good load of tipping this time to get rid of air pockets as there were plenty.

I am using this res:
https://www.overclockers.co.uk/xspc-acrylic-tank-reservoir-for-laing-d5-wc-172-xs.html
 
Do you have any temp sensors for the coolant?

The xspc LCD ones are really good if a little fragile.
 
I don't have a sensor for the loop, keep meaning to fit one when I clean or change but always forget.

I had another thought last night actually, I did add another rad to the loop a little while ago so not sure whether it's just requiring a bit more oomph to push the liquid through the loop. The new rad is mounted at the top so the liquid is going higher and through a rad at the top too.

I guess I'll just keep an eye on it for now, the pump does have to take the liquid through the cpu, through a 240 rad at the top, two gpu blocks and then a 360 rad which is mounted with the plug ports at the bottom. So in essence pumping the liquid to the top of the case twice.
 
A d5 can handle that just fine, mine goes through 240,360,gpus,cpu,360,res then pump. I was a little worried about one pump going through so much, i thought the speed might slow too much but i can see it coming back into the res and it hardly slows at all.

I tried dropping my pump speed when i rebuilt this one, i tried all the speed settings but it made virtually no difference to the temp, in the end i figured i might aswell leave it on full speed since its been running at that for 10 years, hopefully it can manage another 10 before i have to change it :)
 
I started using the top card as the main card again as the second card was actually pretty unstable, not sure if that's the same problem you had a

Yeah i had the same problem, thats why i thought i better mention it, i think it must be a 290 thing cause my 7950s never had the same problem.

Yeah like bluntwrapped says, what are your CPU temps, you could judge your loop temp from that.
 
I know what you mean about being able to see the additional water movement and bubbles when running your pump 100%. That's one of my bleeding methods when I've refilled my loop to run different speeds to try and clear air bubbles. This shouldn't however continue once the loop is air free! In my loop if I'm running the pump at 100% even with a 200mm tube res and vortex stop insert (its an EK X-Res D5), you still see a noticeable swirl.

Are you running your GPUs in serial or parallel? I prefer parallel as its less restrictive.

Have you taken the blocks apart and cleaned them? I use a soft tooth brush and DI water to clean all the fine areas of the block to clear any crud out, as my blocks are all nickel plated and it's apparently really easy to strip the plating if you use vinegar or the like.

Have you flushed the loop with DI water and replaced the coolant? Might even be worth considering using something like Mayhems Blitz cleaning system (just take note that the radiator cleaning fluid, must only be used on RADs as its too harsh for the blocks). BTW what coolant have you been using? Any sign of it degrading?

TIM wise, i use thermal grisly kryonaut, you need to be a bit more liberal with the TIM on GPUs than CPU's as it needs to completely cover the chip (whereas because of the the integrated heat-sink on CPUs, that's not required). As longs as it's non conductive TIM it doesn't really matter if it goes over the edge of the chip a little.

E-I
 
Thanks for the input guys, I have narrowed down the issue. After I cleaned the block last time the temps were fine but then began to rise from 45 to 60 again by the end of the week, the second card remained unchanged. So I took it apart at the weekend and drained it completely and replaced the Mayhems with DI water and so far it has been fine, holding at 45 degrees. Seems there was some gunk in my coolant I'm not entirely sure where it came from but the microchannels on the first card were trapping it all before it got to the second and causing restiction.

So I have ordered some more coolant, some tubing and some fittings that I couldn't get but really wanted when I added the new rad. I'll take everything out and give it a good clean. I've also got some new fans to add so it'll be a nice spring clean.


Are you running your GPUs in serial or parallel? I prefer parallel as its less restrictive.
I believe it's in parallel.


Have you flushed the loop with DI water and replaced the coolant?

Have you taken the blocks apart and cleaned them?
This is basically what I've done now, and it seems that was the problem.

As above not sure if it was the coolant was using Mayhems X1 and have ordered the same again.
 
Back
Top Bottom