Liquid Cooling loop issue

[nixgod]

Ok, so i have a liquid cooling loop, its not a high end one, to everyone's horror here i'm sure... Its a bigwater 760 pro (modded push-pull)

And before everyone says its crap, hey it was cheap and up until now has served my needs well.

Recently, i embarked on the need to water-cool my GPUs they were running way to hot on air, 60c idle, 101-102c load (at this point they would lockup) of course before water-cooling i tried everything to cool them including taking them apart cleaning and applying new thermal compound (AS-5) this didn't make much if any difference they were already free of dust, and the AS-5 only dropped temps maybe 4-5c so we were still at about 55c idle and pushing 99c load.

Therefore i decided the best course would be to add water-blocks to them, for the most part this has worked, CPU and GPU1 (first) idle at 35c and 45c respectively and load temps are 62c and 65c respectively so there we're doing pretty decent.

But when we come to GPU2 (second/last) the last link in the loop its hot under load, it idles the same as GPU1 about 40-45c but when its put under heavy load like a benchmark or high end game the temps shoot up to 99c in less than 120 seconds *EDIT* i should also say here too that after stopping the load the card returns to about 60c within 60sec or so

I've been trying to troubleshoot this for about a week now, I've considered thermal paste (checked and reapplied, I've considered the block not being tightened down enough checked seams good, I've considered air bubbles, were a few but have worked them all out now i think, and still its just as hot as day 1.

So, i've moved on to other components, I've considered 1) the TT radiator which is know to be low-flow/restrictive and the TT p500 pump, which some say is bad quality/not strong enough

A few things should be noted though, as for the pump, it seams to have pretty good flow to me, i'd estimate it having about 250lph with the current restriction I've placed on it, so its doing "alright" i think at least.

And as for the radiator, it does get really hot under heavy load, i can touch it for about 60 seconds before it starts to get unconformable, so i'm wondering if it could be not enough RAD in the loop, the only thing that makes me believe otherwise is, if the rad wasn't cooling enough you would expect the temp increase to be uniform across all components, and it isn't its only on GPU2

I should also note, that low end games 40-50% load on both GPUs the temps never go above 65c on either card, which is slightly weird.

If it is a rad issue, i end up with a delima, if i get a second rad, will my pump push it, i figure i'm probably at the maximum output of this pump now, i'm not sure it will be able to take on another rad, maybe if i was to switch out the TT rad with a high flow rad, it would let me add another high flow rad at the back, and adding a new pump to a bigwater isn't so simple because the pump is kinda made to the res or at least its directly connected to it. and i really don't have the funds for a pump/res combo + a radiator



I would appreciate any useful comments and suggestions.

i will however please ask people to reframe from immediately clinging to the the answer "Your hardware sucks buy all new hardware, that is the only way"

Thanks

 

[ilal2ielli]

Try running GPUs in parallel. If in serial, 2nd GPU might not be getting enough flow to move heat away fast enough. If going parallel fixes it, I'd say upgrade the pump if you want to go back to serial.

 

[nixgod]

Try running GPUs in parallel. If in serial, 2nd GPU might not be getting enough flow to move heat away fast enough. If going parallel fixes it, I'd say upgrade the pump if you want to go back to serial.

Thanks, i'll have to try that if their is no other option, the problem with it is it will be awkward getting them into parallel, since i can't attach a bridge/terminal to them (the G 1/4 threads are facing the wrong way for a bridge) unless they make a G 1/4 coupler that would let liquid flow straight through i'd have to get two splinters one going into the cards and one coming out

I might be able to reverse the loop to test at least so the cold water hits GPU2 first. not sure though if this would give me a definitive answer.

Any more thoughts please feel free to answer.

 

[ilal2ielli]

What card and what blocks?

Any way for you to post a picture?

There are a lot of adapters that can get you parallel flow pretty easy even with what you're describing.

 

[nixgod]

This is the blocks i'm using,
http://www.bitspower.com.tw/index.php?main_page=product_info&cPath=173_268_270_281&products_id=109


Here is a picture of my setup, please forgive the mess in there, i've taken it apart so much this month haven't had time to reroute everything

P.S Sorry about the image size, the forum wouldn't let me scale them.

 

[ilal2ielli]

http://www.performance-pcs.com/fitt...le--180-t/fitting-type--compression/?limit=90

Try this T fitting.

Looks like it would work for your application.

Or any other T fitting really...

 

[nixgod]

Yeah that might work, i'm just wondering if i would just leave the jumper peice on the other side and it still work right.

If this is the correct way for parallel, it looks like i'd need a way to cap off one side of it or perhaps like i said if i just left the jumper that's already there on that side would it still work as expected.

Because if i just use a t-fitting on GPU1 cap off the second port in GPU1 i have a feeling the liquid would have a hard time coming back out of GPU1 with only the one enter/exit port

EDIT - Yeah i think leaving the jumper piece there is the only way to make it work and use the t-fitting on the flow side

 

[ilal2ielli]

Two of those T's attached to top card. Leave those elbows on the bottom card.

Parallel doesn't have to exit at the bottom card, you can still have it exit from the top card.

 

[Erasmus354]

Try running GPUs in parallel. If in serial, 2nd GPU might not be getting enough flow to move heat away fast enough. If going parallel fixes it, I'd say upgrade the pump if you want to go back to serial.

In series the second GPU is getting the same exact flow as every other component in the loop. It is a closed loop, the flow is the same everywhere. So flow is not the issue here.

Typically a problem like this is caused by either a bad mount or plumbing the loop incorrectly such that the block is mostly skipped. It looks like you set it up correctly for a series configuration. So I would look at something being wrong with the block. Either it is not mounting properly or there could be some other problem. The water temperature is not going to be that much hotter between the first and second GPU (at most a few degrees C hotter, typically no more than 1C), certainly not enough to cause the disparity you are seeing.

 

[Activate: AMD]

I would definitely check the mount. I had this issue on a block that lost good die contact (crappy XSPC 7970 block) GPU1 was fine, GPU2 would go off the charts really quickly. If temps are fine on all components other than the one GPU, its only logical to assume that the issue lies directly with that block, or as Erasmus said, the plumbing is such that its not getting any flow. If its truly a serial setup then the latter would be unlikely.

 

[Nenu]

Yeah that might work, i'm just wondering if i would just leave the jumper peice on the other side and it still work right.

If this is the correct way for parallel, it looks like i'd need a way to cap off one side of it or perhaps like i said if i just left the jumper that's already there on that side would it still work as expected.

Because if i just use a t-fitting on GPU1 cap off the second port in GPU1 i have a feeling the liquid would have a hard time coming back out of GPU1 with only the one enter/exit port

EDIT - Yeah i think leaving the jumper piece there is the only way to make it work and use the t-fitting on the flow side

Both of those images are series.
Parallel is when both blocks receive a split from the main water flow.
The parallel in your image shows 2 flows from the first block into the second.
All this does is slightly reduce any restriction between the 2 blocks in a serial flow.

 

[Kardonxt]

Like the others have said something is wrong with your block mount or you may have a bad card.

I have seen this on cpus where they just run way too hot for no apparent reason, getting better TP, heatsinks, lowering voltage, all helps a bit but some magic inside just causes them to run way hotter than they should.

 

[Erasmus354]

Both of those images are series.
Parallel is when both blocks receive a split from the main water flow.
The parallel in your image shows 2 flows from the first block into the second.
All this does is slightly reduce any restriction between the 2 blocks in a serial flow.

Wrong, the first one is indeed parallel. Both blocks have an input on one side and output on the other side.

Think of it this way, if the second block was receiving 2 flows from the first one, where does the water go? It meets in a blackhole in the middle of the block that teleports it back to the output? Water can only travel through the block in one direction. If you assume that the left side is the input, then the water can only move from left to right through the block. So the tube on the right side cannot send water into the second block, because water is exiting it there.

 

[Nenu]

Serial also has an input on one side and an output on the other, its sort of required if you want water to enter then exit.
True parallel doesnt feed the heat from one hot device to the other, both cards are fed the same temperature water.

The "parallel" version shown is serial with effectively a fatter pipe between the 2 hot devices.
The hot water from the first device is fed into the second.
It is not parallel.

 

[Erasmus354]

Serial also has an input on one side and an output on the other, its sort of required if you want water to enter then exit.
True parallel doesnt feed the heat from one hot device to the other, both cards are fed the same temperature water.

The "parallel" version shown is serial with effectively a fatter pipe between the 2 hot devices.
The hot water from the first device is fed into the second.
It is not parallel.

Again, you are wrong.

A simple illustration of how the water flows in the parallel loop. The hot water from the first device cannot go through the second device. Water is exiting the block on the right side in the second device, the hot water can't enter the block there.

You have to keep in mind that image is designed with blocks like this in mind:

Where there is no path to get from the left side of the diagram to the right without going through the block. Also the water can pass straight through from top to bottom without going through a block.

 

[Nenu]

The original diagram did not show any independent routing.
When backed up by other meaningful information on how water will route through such as that you have provided, it shows promise.

 

[Tsumi]

The original diagram did not show any independent routing.
When backed up by other meaningful information on how water will route through such as that you have provided, it shows promise.

... really?

You just don't want to admit you were wrong.

 

[Nenu]

You havent shown the block that performs that task in the diagram.
The block you showed only has one input and output, not 2 inputs and 2 outputs.

If the path of water going to the 2nd block is in contact with copper from the 1st block, there will be a transfer of heat to the 2nd block.
The larger the temperature delta, the more heat will transfer.
So with cooler water entering the first block, there could be an appreciable amount of heat passing from the first block to the 2nd block.

Its not truly parallel unless there is no heat tranfer from one block into the other.

 

[Tsumi]

You havent shown the block that performs that task in the diagram.
The block you showed only has one input and output, not 2 inputs and 2 outputs.

If the path of water going to the 2nd block is in contact with copper from the 1st block, there will be a transfer of heat to the 2nd block.
The larger the temperature delta, the more heat will transfer.
So with cooler water entering the first block, there could be an appreciable amount of heat passing from the first block to the 2nd block.

Its not truly parallel unless there is no heat tranfer from one block into the other.

If you don't know what you're talking about, stop posting, and stop spreading misinformation.

If you are truly that ignorant, get yourself a full cover GPU block and study how it's designed. Then add in the fact that all full cover GPU blocks are built the same basic way. Then realize why that first diagram shows parallel flow.

 

[Nenu]

Not if there is heat transfer from the first block to the second.
If that has been addressed, fair enough.
But the block diagram does not illustrate that.

 

[Tsumi]

Not if there is heat transfer from the first block to the second.
If that has been addressed, fair enough.
But the block diagram does not illustrate that.

It has been addressed, right here:

Wrong, the first one is indeed parallel. Both blocks have an input on one side and output on the other side.

Think of it this way, if the second block was receiving 2 flows from the first one, where does the water go? It meets in a blackhole in the middle of the block that teleports it back to the output? Water can only travel through the block in one direction. If you assume that the left side is the input, then the water can only move from left to right through the block. So the tube on the right side cannot send water into the second block, because water is exiting it there.

Nor does it take a rocket scientist to figure out the flow in that block diagram can only go in one direction. How about a simple application of logic? If you're unable to read and comprehend, use your logic to figure it out.

 

[Nenu]

Blimey.
I can say the same to you.
If water touches the copper in the first block before it touches the second block, its not truly parallel.

There hasnt been a picture of an actual block that has 2 inputs and 2 outputs.

 

[Tsumi]

...

All the facepalms in the world doesn't even cut it...

 

[Kardonxt]

Serial vs parallel debate aside..... it doesn't make a big difference and isn't causing your problem. It's not like the water sits long enough in the first block to hit a high temp and flow hot water into your second gpu. The water moves so quickly it only picks up a few degrees.

 

[Erasmus354]

Blimey.
I can say the same to you.
If water touches the copper in the first block before it touches the second block, its not truly parallel.

There hasnt been a picture of an actual block that has 2 inputs and 2 outputs.

You do realize that even if it touches the tiny bit of copper in the fitting that the heat transfer is going to be too small to even measure right? We are talking about potentially 0.001C or something on that magnitude of heat transfer. You might as well cling to the fact that heat can transfer from the block to the air through the tubing and into the water!

 

[nixgod]

Thanks guys, I think what I'm going to have to do at this point, is add another RAD with a 148CFM delta pushing out the back.

More and more it seams like a RAD/Dissipation issue, because i noticed yesterday that if i play a high end game (BF4) on medium settings frame locked to 60fps putting like 50% load on each card, then that second card stays around 69C but if i raise the settings to ultra which takes them up to about 80% load the second card starts jumping back up to 85+C

So best i can tell i just need another RAD

 

[ilal2ielli]

What temperature do you get on the first card when only at 50% load?

Water temps will stabilize so if your water temp is too high, it still doesn't explain the fact that the other GPU is much higher than the first.

So yes a second radiator will help, but it still doesn't explain the high temps for GPU 2.

 

[Kardonxt]

Ilal2ielli is correct. If it was a rad issue both of you cards would be hot. Something is wrong with the mount or the card itself. You could take out the first card and just run the second if you want to prove it, or try flipping the order of the cards.

 

[nixgod]

Well, it depends on the ambient temperature which fluctuates a lot here.

But at 50% load the gap between the two is about 10-19c so GPU1 is about 49c while GPU2 is about 60-70c

Hard to believe its a mount issue because I have taken it off twice to check compound. coverage

I guess I'm just going to have to tinker with it might remount it just to be sure I think I have enough compound

EDIT:: I did notice something interestin, running seti@home at full throttle, so each component in the loop at max, I did see a much more uniform heat up CPU got to about 75c and GPU1 66C but still GPU2 hit 88c

 

[Tsumi]

Well, it depends on the ambient temperature which fluctuates a lot here.

But at 50% load the gap between the two is about 10-19c so GPU1 is about 49c while GPU2 is about 60-70c

Hard to believe its a mount issue because I have taken it off twice to check compound. coverage

I guess I'm just going to have to tinker with it might remount it just to be sure I think I have enough compound

EDIT:: I did notice something interestin, running seti@home at full throttle, so each component in the loop at max, I did see a much more uniform heat up CPU got to about 75c and GPU1 66C but still GPU2 hit 88c

Still not normal. The difference between GPU1 and GPU2 should be on the order of 5 C at 100% load, max. It might be the case that GPU2's block is defective. The best way to find out, as said, is to swap the two cards and see what the temperatures are like. CPU temps are irrelevant as just about everything is different.

 

[nixgod]

If in fact is is a defective block, i don't know if bitspower is going to replace it or not, it is one of their older models.

If its a defective card, well unfortunately, i don't have much choice there but to let it run warm until it kills itself, simply don't have enough spare money atm to buy new cards, that was kinda the reason i opted for water blocks, it was less than half the cost of new cards.

I did notice something today, perhaps someone could shed some light, possible defects aside, i noticed today while playing a game, non-sli GPU1 at 99% load GPU2 at 0-1% load GPU1 had a temp of about 60-61C now with GPU2 being at 0% load you'd expect it to be near a idle temp, it wasn't it was at the same 60c of GPU1 so their appears to be some heat transfer there.

 

[Tsumi]

In that case, it is likely your pump is far too weak for your setup.

Edit: If that is the case, you can partially alleviate the problem by running your GPUs in parallel. A high flow radiator won't really fix your problem. But it will fix the problem of having aluminum in your loop. As I recall, all Thermaltake radiators are aluminum.

Watercooling is one of those things where you're far better off spending the money upfront and doing it right the first time. Otherwise, you end up spending more upgrading to be at the same level.

 

[Kardonxt]

I would still swap the cards to verify what the actual problem is instead of speculating.

There may not be enough pressure on the 4 screws around the gpu die? Maybe try seeing if you can get another 1/2 or even 1/4 screw on there. I have had issues like that with cpu blocks that use thumb screws. It works well enough to prevent overheating but nearly as well as it should be. Then a few more cranks and it works fine.

 

[Erasmus354]

In that case, it is likely your pump is far too weak for your setup.

Edit: If that is the case, you can partially alleviate the problem by running your GPUs in parallel. A high flow radiator won't really fix your problem. But it will fix the problem of having aluminum in your loop. As I recall, all Thermaltake radiators are aluminum.

Watercooling is one of those things where you're far better off spending the money upfront and doing it right the first time. Otherwise, you end up spending more upgrading to be at the same level.

I don't think the pump is the problem. Closed serial loop means all components are getting the exact same flow rate. So this would not explain a temperature discrepancy. You make a very good point about the radiator being aluminum. This could cause all sorts of problems, there may be some corrosion issues that have generated gunk that is clogging up that GPU block.

I would still swap the cards to verify what the actual problem is instead of speculating.

There may not be enough pressure on the 4 screws around the gpu die? Maybe try seeing if you can get another 1/2 or even 1/4 screw on there. I have had issues like that with cpu blocks that use thumb screws. It works well enough to prevent overheating but nearly as well as it should be. Then a few more cranks and it works fine.

This is what I suggest you do. Just because it looks like it is all covered doesn't necessarily means that you have a good mount. You also could be using too much compound, which makes thermal transfer worse. I think you need to take it apart, open up your GPU blocks to see if they need cleaning, and remount the blocks to different cards and see if anything changes.

 

[Tsumi]

I don't think the pump is the problem. Closed serial loop means all components are getting the exact same flow rate. So this would not explain a temperature discrepancy. You make a very good point about the radiator being aluminum. This could cause all sorts of problems, there may be some corrosion issues that have generated gunk that is clogging up that GPU block.



This is what I suggest you do. Just because it looks like it is all covered doesn't necessarily means that you have a good mount. You also could be using too much compound, which makes thermal transfer worse. I think you need to take it apart, open up your GPU blocks to see if they need cleaning, and remount the blocks to different cards and see if anything changes.

Yes, but a low flow rate means that the water spends more time in each component, giving it a longer time to heat up before going to the next component. A high flow rate is what allows normalization of temps within a loop. A low flow rate causes temps to differ.

I think the suggestion to swap both cards is the best troubleshooting course of action.

 

[Erasmus354]

Yes, but a low flow rate means that the water spends more time in each component, giving it a longer time to heat up before going to the next component. A high flow rate is what allows normalization of temps within a loop. A low flow rate causes temps to differ.

I think the suggestion to swap both cards is the best troubleshooting course of action.

Even an exceedingly low flow rate will only result in a water temperature increase of ~2C, maybe 3C. That isn't nearly enough to explain the 20C temperature difference.

 

[Activate: AMD]

But at 50% load the gap between the two is about 10-19c so GPU1 is about 49c while GPU2 is about 60-70c

With 50% load on both cards? This is absolutely abnormal. I'm running 2x 290X's in parallel, used to run them in series (change was based purely on aesthetics rather than an expectation of improved performance) and the difference between the two at any point where the load is even load (whether its idle, 50% or 100% on both cards) is never more than 2-3C off, and its usually the first card that actually runs hotter since the GPU utilization on the display card is usually slightly higher

 

[nixgod]

Welp, I guess what i am going to do today, is take the block off triple check the mount, this time i'm going to pop the heat spreader off the chip and replace its thermal paste too. its the only thing i know left aside from it being just broken.

Because i can come up with nothing else which would cause such a dramatic temperature differential.

 

[Erasmus354]

Welp, I guess what i am going to do today, is take the block off triple check the mount, this time i'm going to pop the heat spreader off the chip and replace its thermal paste too. its the only thing i know left aside from it being just broken.

Because i can come up with nothing else which would cause such a dramatic temperature differential.

If you aren't comfortable removing the heat spreader I would just swap the cards and blocks around. It could be a defective block that you aren't quite noticing. Also I would absolutely open up the block and inspect the inside.