Overclock.net banner

CPU is hot, water is still cool? Help please! New very custom loop with motorcycle radiator.

2K views 23 replies 9 participants last post by  jsco 
#1 ·
UPDATE! Re-applied Thermal Paste and improved the situation greatly. Please see post #14

Hey guys, just built a custom loop. Its awesome! I have an old aquarium pump, a radiator from 80s Honda Motorcycle, an automotive fuel filter, 4 large 120mm fans, and a gallon reservoir that began life as a pickle jar. Spend the whole day putting it together. Aside from the one leak out at the radiator its working well, and that leak will be handled tomorrow.

Only problem is that my CPU is getting really hot despite the water being cool. Has anyone else had this trouble and what caused it? AMD 970 Black overclocked to 4.1, and I am getting the same results as I was when air cooled. Im using arctic silver thermal paste and this water block: LINK

Im getting photos of my custom loop together for my second post.
smile.gif


In this photo the returning water was still cool to the touch but my CPU was hot!



EDIT: I really expected it to make the CPU temps more stable, but it hasnt... CPU has crazy fluctuations in temp while surfing the net, going all the way to 46 C and then back under 40, so somehow I am not getting proper heat transfer from the CPU to the water or the water block.
 
See less See more
2
#6 ·
I wonder if that pump is powerful enough to get much flow though the filter? GPU blocks work well with very little flow. CPU blocks, not-so-much.

A filter in a watercooled loop is generally not a good idea, but in a loop where the water is in a bowl exposed to outside air well, you've probably got a ticking time bomb waiting to happen in any case.
 
#8 ·
Quote:
Originally Posted by TheJoe90 View Post

Old Aquarium pump with unknown specs. Im trying to wrap my head around the physics of how that would be the problem without me also seeing the issue on the GPUs.

My mind was more to the thermal paste.
Oh i didnt notice that. 32 avg on the GPUs but 50+avg on the CPU is indeed odd. Id check the cpu block, reapply thermal compound
Quote:
That pump flows 400lph while a D5 generally flows 1400lph. Im not saying its the worst pump ever, but, its definitely not a great choice for an automotive radiator.
 
#9 ·
Thank you so much for the feedback!
Quote:
Originally Posted by Jeemil89 View Post

Sounds like a bad CPU block mount or bad flow trough the block. But since your GPUs are not hot then it has to do something about your CPU block. Double check that the block has good contact with the CPU.
This is new info and very helpful, thank you so much!
Quote:
Originally Posted by Unicr0nhunter View Post

I wonder if that pump is powerful enough to get much flow though the filter? GPU blocks work well with very little flow. CPU blocks, not-so-much.

A filter in a watercooled loop is generally not a good idea, but in a loop where the water is in a bowl exposed to outside air well, you've probably got a ticking time bomb waiting to happen in any case.
Im working on getting the holes drilled into the lid of the jar to close the loop up, just hasnt happened yet. I figured the fuel filter would protect against any 30 year old junk coming out of the radiator. After I change the water next week I will remove it. Thank you for the input.

Quote:
Originally Posted by <CRACK@> (CS:S King) View Post

Oh i didnt notice that. 32 avg on the GPUs but 50+avg on the CPU is indeed odd. Id check the cpu block, reapply thermal compound
That pump flows 400lph while a D5 generally flows 1400lph. Im not saying its the worst pump ever, but, its definitely not a great choice for an automotive radiator.
Ok, so I am not alone in thinking that is WAY to big of a difference to simply be a sub par performance pump. This pump was sitting in the shed and free, its not the one I plan to run long term. Id like something that is a bit quieter and easier to turn on off.

Also, I dont trust this one for long term use. It was free and got me started though.
smile.gif


Looks like there is more indepth info on how to apply thermal paste to this CPU than I knew about when I was setting it up. I am off to do more reading, then some re-applying.
 
#11 ·
The tubing length is very long, what is your flow rate?

Looks like a car radiator? They're designed for heavy flow, so it's likely you're just not circulating fast enough to dissipate heat from your blocks.

that's assuming the blocks are mounted correctly and you've already re-seated and re-applied TP upon them.

In a rush, didn't get to read above messages, so appologies if these have already been covered.

Welcome to the Liquid club mate
smile.gif
 
#12 ·
shouldn't be a big deal to pull that CPU block and take a look. you'll be able to see what's up as soon as you take it off. look for good distribution of paste and evidence that the middle of the block was nearly touching the CPU's heat spreader (thin/transparent paste in the middle with a spread pattern around it that looks like the paste was pushed outward). with an off-brand block, i would make no assumptions about anything being correct, and double check that the spacers on the screw posts are not too tall, preventing it from making good contact, and not too short, allowing you to over-tension and bow the block upward in the middle; double check that the tensioning method causes even pressure and doesn't cause the block to tilt sideways. if the paste shows an uneven mount, bowing upwards in the middle, or no thin spots, you'll know where to focus. consider lapping the block if it bows upward in the middle and it's not due to extreme mounting pressure. take your time and do multiple mounts til you get it right.
 
#14 ·
Well, this was the back of my CPU block when I removed it.

I think I used way too much and may have over tightened the block. Either way, cleaned it up, tinted it as per the instructions, put just a tiny dot of paste, and the results are much much better!



I still surprised though. I expected that the water would absorb the heat (Even more so in a lower flow loop like mine), heat the next component in line, and that the water coming out would be very warm. I didnt expect results that make it look like the chip is a full 10-20 degree C hotter than other pieces on the same loop, or the water coming back into the reservoir to be cool to the touch. I then turned the Radiator fans off, continued testing with Prime 95 and let the heat build to just over 60C, then turned them back on.

Are these results typical or do I still need to do more work?
 
#15 ·


Its interesting how differently the GPU behaves than the CPU. Took me some research to learn why, but what I found is that when it comes to CPU waterblocks the quality of the parts makes a much bigger difference. Im going to be doing some hunting to find the best waterblock for my processor, that way I can shave a few degrees if it becomes necessary.

Thank you so much everyone for your positive statements! Makes me even more excited.

Now that heat is so much less concerning, I need to look into how much voltage I can safely run through my processor.
 
#16 ·
Those results look about right. I'm assuming that you're stress testing the CPU only. Despite being in the same loop, the chip under load is going to be a lot hotter than the parts idling. You're limited by the rate the heat is conducting through the heatspreader and block into the water. Also, it's perfectly normal for the water to be cool. In fact, if the water is significantly warmer than ambient, you either have too little radiator area or too little airflow. If the water entering the radiator is much hotter than the water exiting, you have too little coolant flow. A general rule is to keep your Delta T (temperature difference between ambient air and radiator coolant output) below 10C, with 5C being a high performance goal.
 
#17 ·
Thank you for the input!

In trying to keep the Man Cave quiet, I would like to keep the fans turned off till they are needed, then automatically turn on when needed.

I'm using the Rheosmart fan controller, which picks up the signal from the 4 pin fan controller, then passes that to six additional fans.

Because of this, looks like my fans will not be coming on till the cpu hits 40 degrees C, regardless of other temps.

When running Furmark that means the water heats up to 40 degrees, heating the CPU to 40, then the fans come on and cool it right back down. On the flip side, the CPU can get to 40 degrees while the water is still room temp, running the fans faster than required.

This will need more fussing to get it correctlly automated.

While the fans are on the water stays more than cool enough. =)
 
#18 ·
It would be best to control your fan speed based on water temperature. If you really want to have automated fans, there are threaded temperature probes that you can add to your loop. You should probably establish a minimum fan speed too. Without any airflow, you have basically no cooling other than the heat capacity of the water.
 
#19 ·
a 10* delta is high performance territory. it is something to aim for but not even close as to be needed or a general rule. a delta of 20*C is perfectly acceptable. the only thing you have to worry about is the temp of your cpu and gpu and at what temp your water cooling parts fail at. retail sold as pc water cooling gear has max limits of 50* water temps. i believe it is the res that is usually that 'low'.

get a fan blowing air on your cpu vrm. an msi board without air blowing directly on the vrm is a disaster waiting to burst into flames. what mobo is that? it looks to be a 4+1 phase as i only see 5 chokes.

if you can get a parallel set-up for your gpu blocks, youll drop a few more C off your temps. a parallel bridge will get your flow rate up. if youre worried about price for a bridge
http://www.performance-pcs.com/swiftech-dual-sli-crossfire-bridge-2-4-pci-spacing-white.html
perfpc doesnt show a black but they have the white version listed twice
doh.gif
 
#20 ·
Quote:
Originally Posted by u238 View Post

It would be best to control your fan speed based on water temperature. If you really want to have automated fans, there are threaded temperature probes that you can add to your loop. You should probably establish a minimum fan speed too. Without any airflow, you have basically no cooling other than the heat capacity of the water.
I have not seen any type of fan controller that functions based on the temp of the water. Do you have a link? I have seen some for automotive use, but they usually switch at about 200F.

During normal playing on the net and streaming shows the water temp doesn't seem to rise at all. I also am not sure why it would matter much what the temp of the water matters, as long the temp of the components is within my goals. My goals are not hard for me, not sure why it would matter if the CPU is 45 or 55 under extreme load, as long as it is sustainable.

Is there a reason to keep the water temps lower?
Quote:
Originally Posted by pc-illiterate View Post

a 10* delta is high performance territory. it is something to aim for but not even close as to be needed or a general rule. a delta of 20*C is perfectly acceptable. the only thing you have to worry about is the temp of your cpu and gpu and at what temp your water cooling parts fail at. retail sold as pc water cooling gear has max limits of 50* water temps. i believe it is the res that is usually that 'low'.

get a fan blowing air on your cpu vrm. an msi board without air blowing directly on the vrm is a disaster waiting to burst into flames. what mobo is that? it looks to be a 4+1 phase as i only see 5 chokes.

if you can get a parallel set-up for your gpu blocks, youll drop a few more C off your temps. a parallel bridge will get your flow rate up. if youre worried about price for a bridge
http://www.performance-pcs.com/swiftech-dual-sli-crossfire-bridge-2-4-pci-spacing-white.html
perfpc doesnt show a black but they have the white version listed twice
doh.gif
My Mobo MSI 870A-G46

When you say "Delta" are you referring to the difference in temp between ambient and water temp? Or something else?

I have seen no evidence that I have anything to worry about temperature wise, and I have a few ideas of ways to control the fans in a satisfactory manor. Right now all 4 fans are going off of the 4 Rheosmart, I think I am going to set up a 3 pin Y cord and power two of the fans directly from the 3 pin port on the Mobo. Once I do that, I can tell SpeedFan to turn the fans on any time either graphics card hit 35C or the CPU hits 45C, which should keep them off during TV watching and only on when they are really needed. I will adjust the 4 pin connector, which can only be controlled via BIOS (It doesnt play nice with speedfan for some reason) to come on at 45C, which is what the Rheosmart bases its commands from. This should work perfectly.
smile.gif


I have one case fan at the top of the tower to make sure the GPUs, Ram, and Northbridge all have air flow. Is this enough?

I had never considered a product like that, I would have Frankensteined something with brass fittings from Home Depot. At that price its cheaper to buy it though.

Isnt there concern about the cards receiving an uneven amount of flow?
 
#21 ·
Quote:
Originally Posted by TheJoe90 View Post

Quote:
Originally Posted by u238 View Post

It would be best to control your fan speed based on water temperature. If you really want to have automated fans, there are threaded temperature probes that you can add to your loop. You should probably establish a minimum fan speed too. Without any airflow, you have basically no cooling other than the heat capacity of the water.
I have not seen any type of fan controller that functions based on the temp of the water. Do you have a link? [...]
I can think of a few good ones.

Aquacomputer USB Poweradjust 3 - Ultra
Any model of Aquacomputer Aquaero.
Lamptron CW611
Lamptron FC Touch

There are those and many other fan controllers that will let you connect standard 10k ohm temp sensors (either air or water) for automated speed control.
Quote:
When you say "Delta" are you referring to the difference in temp between ambient and water temp? Or something else?
This.
 
#22 ·
The temperature delta measurement just gives you an idea of how effecting your cooling system is. Ultimately, the chip temperature is what matters.

A couple of Y fitting will work perfectly well for paralleling your graphics cards. bridges are just for looks.

I would avoid shutting off your fans completely. The water temperature will continue to rise when there is no airflow. You will experience dramatic temperature swings by making the fans all or nothing.
 
#23 ·
Well, got my overclock dialed back in, still fiddling the right spot for my graphics card though. I have observed the results of my lower flow rate, but I sure am happy with the setup!

1.547 Vcore voltage, 4120 Megahertz clock speed, FSB is at 2200 Megahertz (Wanted to go higher but couldn't get it stable) 1610 Megahertz Ram Speed (If I'm reading it right).

This test was run over about 90 mins with FurMark AND Prime95 running a "Blend" with my fans set to automatic. The water is kept nice and cool (Measured with the returning water running across my finger) but I do see the temp of the processor dragging up the temp of the graphics cards by a few degrees, which wouldn't happen if I had a more powerful pump. That said, for the money I have spent, this is AWESOME! I have never had temps this low before!

(Please note that all of these test were done with fluctuating ambient temp while also streaming The Simpsons.)


I then reset Prime95 to generate as much heat as possible, and here is the result! As you might notice, this is where my less efficient radiator and lower flowing pump start to be overwhelmed, things are getting warm. Still no where near a danger zone, and Id be shocked if any game I play manages to put such a complete load on things for this long. This has made me question if I should add another 4 fans to the back of the radiator as well, but I dont think my system will ever get this hot outside of this test.

(Still running Furmark and having Prime95 running to create max heat, and streaming Simpsons. The water feels about bathtub temperature after 40 mins)


Wanted to see how the system did with Furmark running and only a passive load on the CPU to see if the fans did.


I could set the fans to 100% all the time and keep the Delta lower, but Im happy enough with these results to call it "Done" and start mounting the Radiator. Overall? Room for improvement, but so much improved over what I had, I am so happy!
 
#24 ·
Quote:
Originally Posted by TheJoe90 View Post

Id be shocked if any game I play manages to put such a complete load on things for this long. This has made me question if I should add another 4 fans to the back of the radiator as well, but I dont think my system will ever get this hot outside of this test.
not a chance in hell. nice results, and nice followup.
smile.gif
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top