Sign in with
Sign up | Sign in
Your question

High GPU temp

Last response: in Overclocking
Share
March 25, 2012 8:29:43 PM

My top card goes towards 80c wery quickly in IDLE in windows.

I put finger on waterblock and it burns my finger so the block in properly installed.

It has to be airpocket problem right?
When I tilt the PC on side so that the motherboard is against table I can see air go out of the gpu's into the cpu block and to reservoir.
However seems like anytime I do this air seems to go from res to pump and then finally back to the gpu block?!

I'm now running just the pump at level 5 and hoping for best in bleeding air.

More about : high gpu temp

a c 324 K Overclocking
March 25, 2012 10:15:04 PM

You'd be correct- you have an issue with an air pocket. What is your current loop order? I'm trying to visualize how your flow goes so I can try to determine how to address where the air is moving from and towards.
a b V Motherboard
a c 78 K Overclocking
March 25, 2012 10:29:01 PM

air being fed into, block... puzzling, you have the loop order a lil messed up i think.

front rad>cpu block>GPU blocks>bottom rad>pump>res> front rad

or so thats what it seems from pic.
Related resources
a b V Motherboard
a b K Overclocking
March 25, 2012 11:56:34 PM

It looks like it goes:

pump>bottom rad>GPUs>CPU>front rad>res>pump

Even though your pump is pushing water, you can still get air in the GPU blocks. There's a slight flaw in full cover block designs where you can still have air in the blocks and water flowing at the same time. It took me a while to bleed out all the air in my EK FC6970s because of that...
a b V Motherboard
a c 78 K Overclocking
March 26, 2012 12:10:18 AM

^ yeah that'd be the order. my bad:/
March 26, 2012 10:16:39 AM

boiler1990 said:
It looks like it goes:

pump>bottom rad>GPUs>CPU>front rad>res>pump

Even though your pump is pushing water, you can still get air in the GPU blocks. There's a slight flaw in full cover block designs where you can still have air in the blocks and water flowing at the same time. It took me a while to bleed out all the air in my EK FC6970s because of that...


This is how I have it setup.

I just realized that the EK bridge I have the inlet and outlet wrong, so I guess I need to bleed system completely and re-arrange tubing?
March 26, 2012 10:32:45 AM

edit-->

According manual ports are interchangable, so I think the problem still is airpockets?

a b V Motherboard
a c 78 K Overclocking
March 26, 2012 12:48:29 PM

what boiler was trying to say is that you have some air inside the gpu blocks, not the bridge.

by bleed, it means to remove the air bubbles while pump is running. Not bleed the loop to death and redo tubing :) 
March 26, 2012 2:11:34 PM

I think I got the air bubbles out of gpu block.
in idle windows 7 the core is 75c and VRM 45c when I turned off pc so it is still bad.

Not sure if the problem is air pockets?
a b V Motherboard
a b K Overclocking
March 26, 2012 2:38:49 PM

Lutfij said:
what boiler was trying to say is that you have some air inside the gpu blocks, not the bridge.

Well the bridge is directional (AFAIK), so that may help.

I think there are still some bubbles. Getting them out is quite a pain sometimes.

If you can, try bleeding the loop with your case at an angle; that way the bubbles have more of a tendency to leave the blocks and move through the loop. I also try to shake the case (gently, of course) while bleeding, since you just need to get the bubbles moving.
March 26, 2012 3:32:48 PM

So tiny bubbles cause temperatures reaching 75c+ within minutes?

I'm suspecting bad installation of the top block.
a c 324 K Overclocking
March 26, 2012 3:42:53 PM

In the image above, it mentions intlet can be used as outlet, and vice versa.

So you are running the serial bridge, just in reverse...this should work. However, when it comes to multi-GPU blocks, parallel is a little better for flow. I don't know that this really should be an issue for you, but if you have the ability to switch these, you might try that. (nevermind, the bridge is likely milled either serial or parallel, so not something you could change). Otherwise, I agree with boiler- you're going to have to bleed these out a bit more. If you can lay your case on it's side, you might have better luck...or as much as you can without pulling air through your res.
a b V Motherboard
a b K Overclocking
March 26, 2012 4:45:39 PM

Quote:
In the image above, it mentions intlet can be used as outlet, and vice versa.

LOL If only I could read... :) 

grandpatzer said:
So tiny bubbles cause temperatures reaching 75c+ within minutes?

I'm suspecting bad installation of the top block.


I'm suspecting it's a big bubble that's trapped (which happened in my EK FC blocks). Otherwise it might be a bad mount.

Either would cause a pretty rapid increase in temps, since GPUs produce a lot of heat quite quickly. The heat will conduct very quickly into the block, but if it hits a bubble it's the same scenario as if the block weren't touching the GPU chip at all - heat transfer to air is terrible unless you have convection with another medium.
March 26, 2012 7:36:59 PM

Update: the reservoir is 100% full and I can't see any air so now I need to drain the system and remount the top card as temperatures increase still.
a c 324 K Overclocking
March 26, 2012 7:51:47 PM

Are you actually getting flow? Can you tell if there is movement in your reservoir and how much?
March 26, 2012 7:56:24 PM

that's a good question, I have the pump at maximum speed, when I press tubing where there is tiny bubbles they flow correct direction.

The pump makes noise and vibrates, are you saying maybe there is air in pump and no flow in system?
a c 324 K Overclocking
March 26, 2012 8:05:52 PM

I doubt there is air in your pump, I'm more thinking there is air in your bottom rad that is contributing to the issue...although I would expect it to move out with your pump and res situated as they are. Your pump shouldn't be drawing any air from the res, which is good, this is what you want- it should always be pushing water and air out.

I just think that overall, you have air in your loop that is getting compressed in the loop as your pump is running and not effectively displacing the air. I'd pull your ATX plug, jumper it, and run your pump until you get the air out by tipping your case and rotating it as you can. There isn't any reason to run your PC turned on until you have your flow issues corrected.

If your SLI bridge is correct, and I think it is as it's pretty straight-forward, you simply have an airlock in your loop. If your SLI bridge isn't correctly configured, address this first and foremost. You might actually have a blockage if you have one or more ports configured incorrectly. But, as it looks, I don't think this is possible with it, but I've never used one.
a c 716 V Motherboard
a b K Overclocking
March 26, 2012 8:13:34 PM

This is a messed-up loop and running backwards. You're probably sucking-in air from the reservoir. Obviously, you need to bleed the air out. I would also recommend a flow meter/indicator to the loop.

I'd run: Pump - > CPU -> GPU -> RAD1 (bottom) -> (flipped) RAD2 (front) -> RES -> Pump and double check the reservoir installation.

Don't mean to butt-in, but it all looked odd.
a c 324 K Overclocking
March 26, 2012 8:24:27 PM

Loop order doesn't matter- and his pump inlet is coming from the res, which is also higher. As long as the pump isn't sucking air from the res, the can be bled.

I do agree that res > pump > CPU > GPUs > bottom rad > front rad is a better traditional flow order, but technically shouldn't make much difference. Just odd that it is routed how it is. I'm still betting the issue is from the pump > bottom rad > SLI bridge area, with air in the bottom rad being the main problem.

Jump the ATX plug and tilt your case front to back as much as you can without so you aren't pulling any air from your res (if any remains).
March 26, 2012 8:58:24 PM

rubix_1011 said:
Loop order doesn't matter- and his pump inlet is coming from the res, which is also higher. As long as the pump isn't sucking air from the res, the can be bled.

I do agree that res > pump > CPU > GPUs > bottom rad > front rad is a better traditional flow order, but technically shouldn't make much difference. Just odd that it is routed how it is. I'm still betting the issue is from the pump > bottom rad > SLI bridge area, with air in the bottom rad being the main problem.

Jump the ATX plug and tilt your case front to back as much as you can without so you aren't pulling any air from your res (if any remains).


yes I have been for 1-2 days now jump plugging an external PSU to bleed out air pockets, I no longer can remove airpockets with tilting the case to any direction.

I will one last time turn on PC but I think the block is wrongly installed as I no longer can see air anyway to be removed and visible.
a c 324 K Overclocking
March 26, 2012 9:24:57 PM

Again, I don't necessarily think the airlock is in your blocks, but in your rad. The GPU blocks are a bit more restrictive, so having 3 of them in series would be more so. Having air in your rad that isn't able to push through these would likely be where your airlock is originating.

I noticed on Overclock.net forums, you are asking the same questions, and getting the same answers. I think you need to address the airlock issue.
March 26, 2012 9:42:21 PM

ok number 1 your rad is upside down your intake and exhaust should be on bottom. loop should go from rad to pump to cpu to video to res and back to rad.. try that and should solve your problems.... oh and check your rez isnt useing the wrong inlet port cause if you are it will suck bubbles into the flow....
March 26, 2012 9:43:22 PM

Well I will try bleed one last time this time having the rear of case against table so bottom rad plugs point upwards thus making the air have to go upwards out of radiator.

as for the tilting the case with external PSU I have done alot, incase the bleeding of bottom rad is not improving temperatures I will either rearrange system or move it to another case.

My understanding rubix is that the bottom rad has air and not allowing flow in the loop?
a c 716 V Motherboard
a b K Overclocking
March 26, 2012 9:46:49 PM

Kinda all of what I've been (trying) to say, and IMO a flow indicator of some kind is essential.

Hey 'traditional' works otherwise it's like reinventing the wheel as a square and wondering why the ride's so rough.
March 26, 2012 9:51:34 PM

robustus64 said:
ok number 1 your rad is upside down your intake and exhaust should be on bottom. loop should go from rad to pump to cpu to video to res and back to rad.. try that and should solve your problems.... oh and check your rez isnt useing the wrong inlet port cause if you are it will suck bubbles into the flow....


jaquith said:
Kinda all of what I've been (trying) to say, and IMO a flow indicator of some kind is essential.

Hey 'traditional' works otherwise it's like reinventing the wheel as a square and wondering why the ride's so rough.


So are you guys saing it is better to have: PUMP --> CPU --> GPU --> 240 --> 360 -- RES --> back to PUMP?

I might try this before moving the stuff to another case.
a c 324 K Overclocking
March 26, 2012 9:53:02 PM

Quote:
Kinda all of what I've been (trying) to say, and IMO a flow indicator of some kind is essential.

Hey 'traditional' works otherwise it's like reinventing the wheel as a square and wondering why the ride's so rough.


I agree- and I think that would be the simplest method. I'm not a fan of flow indicators, but for beginners, they are useful.

That all being said, if there aren't any blockages, there shouldn't be any problems with loop order. However, loops like this can be a real pain to bleed correctly.

Quote:
So are you guys saing it is better to have: PUMP --> CPU --> GPU --> 240 --> 360 -- RES --> back to PUMP?


Yes, this is the most 'traditional' way of running loop order.
a b V Motherboard
a b K Overclocking
March 26, 2012 9:56:17 PM

robustus64 said:
ok number 1 your rad is upside down your intake and exhaust should be on bottom. loop should go from rad to pump to cpu to video to res and back to rad.. try that and should solve your problems.... oh and check your rez isnt useing the wrong inlet port cause if you are it will suck bubbles into the flow....


That might make a 1-2C difference...not a 40C difference...
March 26, 2012 10:05:48 PM

Ok thank rubix, I think I'll temporary run the system with just the front 360 and see if temperatures get better in IDLE.


I'm considering building putting all stuff in another case.
a c 324 K Overclocking
March 26, 2012 10:08:22 PM

Radiator orientation shouldn't have anything to do with it, but then again, loop order shouldn't either. I think the problem is that he's pushing through the lowest resistance first, and most restrictive last (radiator then 3 GPU blocks and lastly CPU block). The only thing that bothers me is that his res is full and feeding his pump correctly, so that should really push through any air that he is encountering unless it's within his radiators and there is too much restriction to break the airlock. Without knowing how much flow he actually has, it's difficult to determine if this is a restriction issue due to blockage, or if it's an airlock causing decreased flow.

Something just tells me there is an issue with that SLI bridge...something isn't right...can't be. The way his pump and res are setup- that D5 should have no problems emptying a res like that in a matter of seconds. I know for a fact- mine does that when I fill my loop. I would be almost certain that the EK SLI bridge is setup correctly, or there would likely be a leak, unless for some reason, those blocks aren't connected via the bridge connector, but again, highly doubt that is the case, either.

Quote:
Ok thank rubix, I think I'll temporary run the system with just the front 360 and see if temperatures get better in IDLE.


You can try bypassing the bottom rad first, and then maybe the GPUs to help isolate where the problem is. If you are able to do this, it would likely hep us identify the issue.
a b V Motherboard
a c 78 K Overclocking
March 26, 2012 10:10:08 PM

wow, i missed alot!
Quote:
Hey 'traditional' works otherwise it's like reinventing the wheel as a square and wondering why the ride's so rough.
:lol: 

Quote:
I'm considering building putting all stuff in another case.
building in another case won't remove this issue, its related to the components you have now.

A Ferrari is definitely better than a Prius but when both of them have all their tires flat, they're differences account for nothing. So changing the car doesn't help as the issue is with the tires to begin with.
March 26, 2012 10:12:00 PM

well I guess it could be bad installed waterblock?
as for the pump it is running maximum speed.

I'll reroute tubing and if no help test with just one gpu= if one gpu low temperature then I will reinstall top card.
March 26, 2012 10:16:20 PM

oh, also is it bad running PUMP --> VGA -- CPU versus

PUMP -- CPU -- VGA?
a c 716 V Motherboard
a b K Overclocking
March 26, 2012 10:27:45 PM

Again, I'd run: Pump - > CPU -> GPU -> RAD1 (bottom) -> (flipped) RAD2 (front) -> RES -> Pump and double check the reservoir installation.

Every build I have has either a visual (cheapo) or digital (expensive) flow meter(s); one per loop. I often use opaque tubing but even clear I want to 'know the flow.'
a c 716 V Motherboard
a b K Overclocking
March 26, 2012 10:30:35 PM

You know this could be from improper pad/block installation if only (1) GPU is running hot.
March 26, 2012 10:36:31 PM

yes only one gpu is running hot but it's also the main card in eyefinity other cards are 30c while the top card always goes up to 80c when I turn off PC.

I was wery tired when installed the top card block.

I'm also considering removing the bottom rad and it's easier to run PUMP --> VGA, but if the PUMP - CPU is better I'll do that instead.
a c 324 K Overclocking
March 26, 2012 11:32:22 PM

Ahhh...if this is only one single block, it's the installation of that block then. If all other components are cooled fine, yet one is hot, this would be your culprit.

My fault for the misunderstanding. It isn't an airlock then...it's an incorrect mount of the GPU block.
a c 716 V Motherboard
a b K Overclocking
March 27, 2012 12:11:19 AM

Yeah, it's fun time. You'll need to drain and pull the cards. I'd still reverse the loop as I mentioned and pretty much do all what I recommend. However, more cooling is better to reduce fan noise, so IMO leave all of exchanger; tubing is cheap and none of that loop is hard on the pump.
a b V Motherboard
a c 78 K Overclocking
March 27, 2012 12:29:02 AM

interesting :) 
a b V Motherboard
a b K Overclocking
March 27, 2012 12:29:39 AM

It's amazing how easily we all come to a consensus even though we've all been arguing completely different causes ;) 
a b V Motherboard
a c 78 K Overclocking
March 27, 2012 12:33:58 AM

thas jaquith for ya mate, a man with eagle vision :D 

oooor maybe we were all stuck on the problem at one view, while jaquith looked at it from a different perspective with a glass of lemonade :) 
a c 324 K Overclocking
March 27, 2012 12:46:31 AM

I think it was a matter of not knowing what really was the issue- I was under the impression that all GPUs were running hot, when it really was only 1. This would have been a lot simpler to isolate if this were known sooner.

OP needs to drain the loop, pull the culprit card and remount the GPU block- taking careful notes how well the block was seated upon disassembly and making sure it re-seats correctly.

This might be a good time to ask- are all the cards and blocks exactly the same, or was this card or block different in any way from the others?

Kudos to Jaquith on catching that with an eagle-eye.
March 27, 2012 1:20:57 AM

So I removed 2 cards and run single card(not the hot one!), I did not change loop order just one card instead of 3.

Now I'm still running eyefinity but now the temperature is only 32c core and VRM is 35c(?!).

So because the bottom 2 cards where 30c and the main card(hot one) was wery fast going towards 80c it is bad mount.

when I put finger on the hot card it would burn wery badly, why is that if it's bad install on core?!?!

a c 324 K Overclocking
March 27, 2012 1:30:18 AM

Quote:
This might be a good time to ask- are all the cards and blocks exactly the same, or was this card or block different in any way from the others?


^ This?
a c 716 V Motherboard
a b K Overclocking
March 27, 2012 1:45:18 AM

It's a bad mount/pad or the card is bad; one of the two.

Before blocking I validate the GPU(s) before breaking it down and blocking.
March 27, 2012 1:46:59 AM

The top card was 7970 and the 2 bottom 7950 I asked EKWB and they told me they would align 100% with the bridge.

So I'm not entirely sure what the problem is, the GPU block was burning hot to the finger so not sure why the GPU core went fast to 80c, seems almost as if no water went through it?
a c 716 V Motherboard
a b K Overclocking
March 27, 2012 1:58:26 AM

In a Serial its an all or none. I havent checked if those GPUs can CF together. I am on my iPhone so one of the guys needs to check. Generally I never mix-match GPUs.

Edit except SLI + PhysX ;) 
a c 716 V Motherboard
a b K Overclocking
March 27, 2012 2:19:05 AM

I guess you can, but it's not what I would do by choice. Also, you need to install your bridge connections properly, in the top image it's not.

!