High WC Loop temps, what gives?

[Daemas]

Just set up my first WC loop and the temps are insane.

2500k @ 4.7Ghz @ 1.35v

Old setup:

H100 with default fans in push config, highest pump speed
24.5 Ambient
31-33 Idle
75-85-85-81 highest load temp via IBT maximum.

New Setup:

2x XSPC EX240 rads each with 2x Scythe Kama 3000RPM 120x38mm fans in push config
1x Swiftech Apogee HD Waterblock
1x Swiftech MCP35x
1x Swiftech Mini res v2
7/16 ID 5/8 OD Primochill Tubing
Various compression fittings, only 1 90deg and 1 45deg
Distilled Water and Killcoil

Pump > Rad > Rad > CPU > Res > Pump (Crudely drawl picture of loop)

28.5c Ambient
37-40 Idle
Upon starting IBt, temps jump immediatly to 55 and 2-3 seconds later immediatly to 68-77-78-73. Temps will eventually reach 82-85 ish after 2-3 passes of IBT.


The temps from the new loop are at BEST 5deg cooler than the h100. From what I have been reading, they should be roughly 12-15deg cooler.



What gives? bad CPU block mount? pump broken (tach reads 2000+ RPMs in BIOS)? Too lofty of goals??


thanks in advance

 

[WoodButcher]

I'd be looking at the TIM and the mounting. Accounting for the 4 deg higher ambient your idle speed temps are worse but the load temps are better. Tells me the loop can remove the heat but the transfer from chip to loop ain't happnin',,,
Did you remove the protective sheet from the bottom of the block before applying the TIM?

 

[dajeepster]

You could have air bubbles in your system.. did you bleed the air bubbles out first?

also i would have done

pump > CPU > Rad > Rad > Res > Pump... that way you're not sending hot water to the pump, which could reduce the life expectancy of the pump.

 

[Rubycon]

Air bubbles - unless the pump is noticeably noisy - are going to have marginal impact on cooling performance.

Most likely your mount is questionable. For the best transfer the TIM must be properly applied AND the waterblock mounting hardware MUST be properly tightened down. Miss something in these two areas and the probability of heat transfer efficiency going down is certainly high.

As for loop order, it does NOT matter at all. As long as your reservoir comes right before the pump you are fine. It's a common myth amongst water cooling that you're sending "hot" water somewhere. Unless you have a very high restriction or completely inadequate pump, the water temperature everywhere in your loop is very close to being the same. Even with the miniscule flow encountered with the Corsair H series one is challenged to find a difference between the inlet side and outlet side of the radiator - temperature wise.

That being said, it's always a good idea to monitor your circuit temperature. A simple tee type fitting with a well mounted thermistor connected to a fan controller or motherboard header is all you need. Your water temperature should be no more than a degree or two above the ambient temperature. If it is then your radiator area, radiator fans, are simply inadequate to bring temperatures down. Some kits are sized by load in watts and these may be based on delta (difference between ambient and water temp) as high as 20 degrees C! That's absurd IMHO. Think about it for a minute. In the summer if your room is 26 degrees C and your system at full load is close to this "fantasy" rating your coolant temp could be as high as 46 degrees C! That's quite warm, even hot to some hands and your blocks are going to be that temp, MINIMUM! If there's a 30 degree delta (pretty common) between your coolant/block surface temperature and what core temp indicates then your CPU (core) temp will read 76 degrees C MINIMUM! That's decent for AIR but not liquid and that's a best case temp for a worst case cooling design!

Using the same delta for a proper sub 5 degree delta (coolant vs. ambient) you have a 61 degree C core temp which is much more reasonable. Remember these are just examples and you really cannot compare and calculate until you know what your coolant temperature actually is. Additionally on die sensors are not always accurate at temperatures encountered, they are accurate at one point and deviate accordingly. However once you know your starting point you can gauge improvements from a relative standpoint, generally.

A good tool to have is a non contact infrared thermometer. Cheap ones have fixed emissivity so it's important to learn that some readings (like bare metal) will not be accurate. If you need to take temperatures of such surface place a small piece of masking tape at the measurement point and after an hour point the device's laser at that spot. You will find this produces reasonable readings.

Oh and for pump resistance to higher temperatures, no need to worry here as well. The coolant temperatures encountered on a continuous basis in any PC loop will be far under the upper limit for any decent pump. The biggest enemy to your pump is running out of liquid. NEVER allow the pump to run dry, even for a second! If the system is low on coolant and you can hear the pump becoming audibly louder (slurping) shut it down IMMEDIATELY and correct the problem (add coolant!) BEFORE running it again. Do NOT underestimate how easily the bearings can be damaged by even partially running the pump dry or with insufficient submersion of the suction side. (this is why you ALWAYS have an adequate reservoir right before the pump!) Following this practice, the pump will last a long time unless it's a really shoddy model.

 

[WoodButcher]

Thanks ruby, you always explain things so well.
Glad you mentioned the tape for the IR, is a glossy painted rad OK? or should I paint a target on the rad flat?
Yes, my IR thermo is as cheap as they come. :whiste:

 

[BrightCandle]

I agree with the others its likely a bad mount. The Swiftech Apogee unfortunately is famous for producing bad mounts. You have to get the bow of the motherboard and the CPU pressure just right to get the bend of the CPU block to perfectly match. Remove it and check the paste pattern. There is a decent chance its only touching in the centre or on the edges properly.

Also remember - a grain of dried rice is all the paste you need, its to fill the gaps not make a contact between the entire CPU and the block.

 

[Rubycon]

Thanks ruby, you always explain things so well.
Glad you mentioned the tape for the IR, is a glossy painted rad OK? or should I paint a target on the rad flat?
Yes, my IR thermo is as cheap as they come. :whiste:

Painted surfaces should be fine. Bare, shiny metals will cause problems with instruments using fixed emissivity.

 

[rgallant]

Ruby's the best.

-but to add this seemed out of place along with your super fast rising temps.
""pump broken (tach reads 2000+ RPMs in BIOS)?""

in bios meaning using the mb cpu fan header for pump power ?, could it be in windows the header is slowing the pump's rpm down with a cpu fan [pump in your case] quite program , try direct 12v to the pump[molex] for testing , unless you trust it and you can monitor the rpm when your in windows.

 

[Boulard83]

+1 on bad mounting and maybe too much paste ?

As for the new WC, your loop is way enough for a 2500k, im running 4.5ghz on air and my load are ~65°c

Also look for bubbles and proper flow in the loop.