Watercooling & C/W ratio

From which you measure the ambient? The case temperature or the water?

I think it should be measured from the water, as when you're calculating a HSF C/W, you measure the temperature of the medium used to remove the heat from the HSF. Which is air with heatsinks.

Ambient temp of the processor should be measured 1" from the processor inside the case. That is the temp of the air surrounding the processor.

Some people use the air flowing into the radiator as ambient since it's what is cooling the water, but I prefer the 1" from the processor method.

But the air 1" from the cpu has absolutely NOTHING to do with the C/W measurement of a watercooling setup! atleast measuring the air going into the rad. has some relevance for the C/W. But to get a descent C/W value, use the water in the tank.
Otherwise you would have some fun things going on if you use a bong cooler, and actually manage to get the water (and maybe the cpu) to below room temp, and still measuring the C/W 1" from the cpu. not very relevant =)

please tell me if I am wrong here.

I agree with BadManiac. I'm not using a radiator at the moment, so the air inside my case has no effect on the CPU temperature. Not much anyway.
My case ambient was 33C at one point today, usually it's been 30C during these hot days. No change in CPU temperature that could be accounted for the difference in the case temperatures.

I think that watercooling with a radiator needs somekind of new formula for C/W ratio. There's 3 temperatures you need: water temp, the air temp going to rad & the CPU temp. Some math-genius could propably devise a usable one.

Now that I think of it, watercooling actually has 2 ratios: the water/cpu one & the water/air. Maybe something to combine those. Propably some ratio does exist also that I can't think of now.

Bad Maniac (Jul 26, 2001 12:09 p.m.):
But the air 1" from the cpu has absolutely NOTHING to do with the C/W measurement of a watercooling setup! atleast measuring the air going into the rad. has some relevance for the C/W. But to get a descent C/W value, use the water in the tank.
Otherwise you would have some fun things going on if you use a bong cooler, and actually manage to get the water (and maybe the cpu) to below room temp, and still measuring the C/W 1" from the cpu. not very relevant =)

please tell me if I am wrong here.

Click to expand...

I disagree. The waterblock is still effected by the temp of the air around it. If the air around the waterblock is lower, then the processor will be lower.

I guess it's all up for debate, but my calculations come out about right with the standard on "GOOD" watercooling setups I have seen.

It seems to reason that the air around the processor is the closes correlation to ambient you can get. The water temp could be used, but remember you're then measuring the temp of the water after the CPU has heated it.

just to show off...if the temp 1 inch from the processor is right...heh

i had taken off my rear intake 120 mm fan, and was only running two 80mm exhaust fans on the front. well...i put my case back on fired up some 3d rendering and noted my case temp of 95 degrees...while my processor got up to 92 degrees.

heh, i guess i have a negative c/w ratio, altho radiate gives it an 0.010 c/w for some reason.

ambient temp for a Water cooling setup is going to be the temp of the air going into the Rad. In a good case, the air temp should be pretty similar throughout the case so for some people with good case designs it won't matter. Perhaps someone should ask Joe on how to measure it?

The C/W or thermal resistance of a 'thing' between A and B is a measure or how 'hard' heat transfers (noticeable by a higher deltaT) from A to B (or B to A). So for a watercooling rig, it's supposed to take the waste heat from the processor and dump it to the ambient. Where it dumps the waste heat to the ambient doesn't matter where the radiator or bong is installed -just the temp of the room your PC is in (hence 'ambient'). So, temp A should be that of the CPU ('heat source') while temp B should be that of the surrounding air ('heat sink'), but don't take measurements near the bong or radiator as the air there is still in the process of exchanging heat with the colder air further away. If you measure the CPU and water temps, your are actually just measuring the C/W of the waterblock and a portion of the watercooling rig downstream from the waterblock. The C/W of the watercooling system is a sum of the C/W of the contact resistance, waterblock, and radiator, but you can just measure everything as one lumped sum.

Oh, if your radiator is being fed air from inside the casing, ambient temperatures should be representative of the pre-heated air in the casing. It's not as high as the mobo temp. diode reports as the diode itself is subject from the electrical heating of the mobo.

Looks like my previous post never got here Must've had that "preview/attach" box checked. Oh well.

I have a quite direct correlation (is that the right word?) between my water temps & my CPU temps. I guess everyone else has too (???). Eg. If my water temp drops by 2C, so does my CPU temp.

Ambient = room temp, check a dictionary!

The temps in your case are case temps. The Delta T between the two tells you how good your case cooling is. If you really want to know what is going on, measure the ambient, case temp, air going into the radiator, air exiting the radiator, water temp and water block temp. If you run pelts, measure the cold plate temps too. Then figure Delta Ts to your hearts content. When analyzing a cooling system, I use them all.

This makes a good case for buying a handfull of the indoor outdoor thermometers at the Shack when they go on sale for $10 twice a year. I never seem to have enough. A dual Compunurse can handle the block and coldplate temps.

Colin, you wouldn't happen to have a formula for all those temps that gives out a number like C/W ratio?

The C/W needs to be calculated for the whole watercooling system, not just the waterblock, so like Colin says you need to take into account the temps at various points around the loop. However, I believe that to a good approximation you can just use the temp of the air immediately before it enters the radiator, and the cpu temp. Let me explain my reasoning.

The only other important temp as I see it would be the 'ambient' or case temp (whichever you prefer) near the waterblock - however, since most of the heat transfer will be between the waterblock and the water you can ignore (as an approximation) the heat transfer to the air. Remember, water has a thermal coefficient about 20 or 30 times (seen different figures, not sure which is right) greater than air, and so will absorb most of the heat. I expect then that you could change the air temp around the waterblock quite a bit (by a few degrees C at least) and not see a change in cpu temp.

If you wanted a more accurate C/W, you'd have to take into account the efficiency of the waterblock at transferring heat to the air - it'd be interesting to see a waterblock with a hsf built in on top. I don't know how you'd factor that into the overall C/W - it'd probably be best to use a weighted sum of the C/W for the waterblock to the air, and the C/W of the watercooling system to the rad. Something like:


I don't know whether that's even along the right lines - maybe in a couple of years, when I've done a bit more of my Materials Science and metallurgy degree, I'll be able to tell you

Hmm... This post is gathering more scientificism than I thought. I like it So I'll do this: BUMP

Just a little note here: I finally managed to get a filter for my front intake fans, allowing me to run them at 12v, when necessary.
My case temps dropped by a nice 2C, but the CPU temps stayed exactly the same. I think I'd need about 6C+ changes in case temperatures for it to be accountable for changes in CPU temperatures.

asmodean (Jul 30, 2001 08:46 a.m.):
Just a little note here: I finally managed to get a filter for my front intake fans, allowing me to run them at 12v, when necessary.
My case temps dropped by a nice 2C, but the CPU temps stayed exactly the same. I think I'd need about 6C+ changes in case temperatures for it to be accountable for changes in CPU temperatures.

Click to expand...

I noticed about the same thing. Running my fan full speed or half speed results in the same CPU temp. I guess the water and radiator are pretty efficient as is and it takes LARGE amounts of air for minimal gains in CPU temp. I'm sure glad I put the rheo on my fan, I run it at half speed and don't have any noise.

I use the intake temp at the radiator entrance for ambient and calculate C/Ws from there. There are many variables in the system, but for comparative purposes, I always use the same radiator and pump to minimize them.

Users are interested in what the whole system can do, so I think it's more meaningful to report on that basis, although there is no doubt that the other measurements mentioned are valid indicators for fine-tuning the system.

Joe Citarella

Rob Cork (Jul 28, 2001 06:05 a.m.):
The C/W needs to be calculated for the whole watercooling system, not just the waterblock, so like Colin says you need to take into account the temps at various points around the loop....

Click to expand...

Nope, for the WHOLE watercooler, you only need two temperatures -the heat source and the heat sink (CPU and ambient) by definition of C/W: 1/UA=deltaT/Q=thermal resistance. The only reason you'd want to measure several temperatures is to find out the individual thermal resistance of each component on the WC to try and improve upon any bottlenecks.

If you wanted a more accurate C/W, you'd have to take into account the efficiency of the waterblock at transferring heat to the air - it'd be interesting to see a waterblock with a hsf built in on top. I don't know how you'd factor that into the overall C/W - it'd probably be best to use a weighted sum of the C/W for the waterblock to the air, and the C/W of the watercooling system to the rad. Something like:

(0.1 x C/W of waterblock to air) + (0.9 x C/W of water system)​

Click to expand...

heat transferred:
Q_CPU=Q_HS+Q_WC

thermal resistances (C/W):
R_total=(R_HS*R_WC)/(R_HS+R_WC)
..due to two parallel heat transfer paths

R_HS=(T_CPU-T_case)/Q_HS

similarly,
R_WC=(T_CPU-T_ambient)/Q_WC

Q_CPU you can find out from Radiate.

Q_WC you can find out by measuring the temperature difference (deltaT_w) of the water entering the radiator/bong and exiting it, and knowing the mass flow rate (m') of water (kg/s): Q_WC=m'*c*deltaT_w; c is the specific heat capacity of water (use an average value between the two temperatures)

Find out Q_HS since you already know Q_WC and Q_CPU.

T_case will approach T_ambient in a well-ventilated casing.

You can now find out R_HS and R_WC to find out the total thermal resistance of your WC rig with the little HS on top of the waterblock . Hope this helps you fellas.

my primary language is English, however I speak a little Spanish. For whoever understood this, would you please translate it to a language I can understand? =)

here it goes

You stick your finger on the block, guess the temp, divide it by the guessed temp using a toe measuring the air coming into the radiator. Dived the first by the second, add 5 and eat a sandwhich, and you have your C/W ratio.


Well, thats what babblefish gave me.