Will water cooling be better than air cooling (PLEASE READ)

[Blackmyst]

Hi.

I want to watercool my PC. I've never done it before, and thats the main reason I want to, but the other reason is because my system is very loud. Because of the expense of setting up a good system, I want to hear what you guys have to say about what you'd expect from my system.

Here's my current setup:
Hardware
~ Epox EP-8RDA3+
~ XP1800 (can't remember stepping, but its a good one)
~ 2 Western Digital Raptors in RAID0
~ Vantec Iceberg on NB
~ Zalman Gold Heatsink on SB
~ Heatsinks and 60mm fan on MOSFETS
~ 5 Thermaltake Smartfan 2's (3 in, 2 out)
~ 92mm Vantec Tornado Fan on an SLK-900
~ 2 Vantec NXP-201 Fan Controllers

Settings & Voltages
~ Vdd = 1.6v
~ Vagp = 1.5v
~ Vdimm = 2.63v
~ Vcore = 1.825v
~ 215mhz x 11 = 2375mhz

With all fans at slowest under full load
CPU Temp = 60oC
*remember that the 8RDA3+ reads about 8oC higher than the 8RDA+*
Case Temp = 30oC

With all fans on maximum under full load
CPU Temp = 50oC
Case Temp = 29oC

Purposed watercooling hardware
~ Dangerden Maze 4 CPU block
~ Danderden Z-chip NB Block
~ Eheim 1250 pump
~ Black Ice radiator w/ 120mm fan
~ Tygon tubing
~ Voyeurmods reservior
~ Waterwetter/anti-freeze

So, my questions are:
1) Is my current setup good, and are the temps good?
2) With a good watercooled setup, what temps would I be expecting? I realise that this differs from setup to setup, but educated guesses are what I'm after.

Thanks for all your time, and if I've missed anything out, just ask...
Blackmyst

[Kaitain]

Looks like a reasonable plan for a watercooled setup. You'll hear all kinds of wild claims about how effective watercooling is and how low it's pushed the temperatures. Ignore them all - they all have a whiff of fisherman's yarn

Taken from a purely thermodynamic perspective, there are some rules of thumb for minimum temperature differences between air and water, and water and CPU, at equilibrium:

Bare minimum difference between radiator fins and air is 10ºC for a normally-flowing radiator. There's a couple more ºC between the radiator and the water. Realistically expect the water to be 15ºC warmer than the room.

Bare minimum temperature difference between the water and the waterblock, taken at the outlet is about 5ºC. The CPU will thus be 5ºC hotter than a perfectly effective waterblock. The residence time of water in the waterblock is small, so a block-volume of water will not heat up much inside the block, so you can neglect this.

Which brings us to the rough-guide for what CPU temps you should be expecting. About 20ºC warmer than your room. So in a nicely air conditioned office (18ºC), you'd be looking at sub-40ºC temperatures.

I see you've gone for a combined reservoir/radiator setup. I personally don't favour large reservoirs, preferring to actually dissipate waste heat rather than just storing it up for a warmer future. The effect of the reservoir will be to cause your system to take a long time - possibly hours, possibly even days - to reach thermal equilibrium where the above rule holds true. Until that time, you'll see lower temperatures, rising slowly over time. These pre-equilibrium temperatures are what a lot of people report, which is where they get thermodynamically impossible performance out of their cooling loops.

Though he's left active posting, Random Nonsense built a cooling loop with obscenely large radiators that brought the water back to room temperature on each pass. This system WAS able to improve on the approximate guide above.

I expect to see a lot of people posting up results claiming stable CPU temperatures of only 5ºC above room temperature. I'd imagine those computers are powered by a perpetual-motion generator in the back yard.

Watercooling's definitely a fun thing to try - enjoy playing

[Blackmyst]

Bump

[whatever]

watercooling can be good.

if not setup well and fail to keep the water cool, your overclockability will not get much of improvement. Water flow helps also. having good pump with good loop setup is good too. Need a good waterblock and do a good installation. keep the water ful, no bubbles.. things like these will help you overclock higher. I use chilled water so i get much higher clockspeed and my water stays cold because i have a big freezer. If u maintain *cool* water, you can get some nice boost in overclocking. Your challenge will be to keep that water temperature down because your cpu will generate a lot of heat.

[cloasters]

I thought that water cooling was more effective until I read Kaitain's learned post. Is it safe to posit that even at 15C above room temperature, water still removes heat noticeably better than air cooling? I'm asking about the huge gain in efficiency by using dense water as opposed to nebulous air.

[Postaldave]

hay wutever.... you use a freezer?... can u explain and or take pics?.... i'm curious as to what you mean... i live out in the desert where temps are getting as high as 126 with up to 80% humidity... .its literally hell here

i would really like to keep my system cold as i run it all day.... also is your system portable or what... thanks for letting me know...late

and no i don't want to do phase change cooling lol... i need something that will cool my nb.... proc... and card.... possibly even hard drives

postal

[Lazgoat]

Hey postal, check out watever's thread in the AMD section. Xp1700+ @ 2.7Ghz!

[Kaitain]

Originally Posted by cloasters

I thought that water cooling was more effective until I read Kaitain's learned post. Is it safe to posit that even at 15C above room temperature, water still removes heat noticeably better than air cooling? I'm asking about the huge gain in efficiency by using dense water as opposed to nebulous air.

Yes. Using water is hugely advantageous over using air, mostly due to the fact that water can "absorb" (sorry engineers) more energy per degree celcius it warms up than can air. A lot more. Which is handy

With water cooling you can also be a lot more flexible. The approximations I have outlined above are only approximations, and rather conservative ones at that. I have actually based them on the API standard "short cut" designs for air- and water- cooled heat exchangers. As you're aware, from the first law of thermodynamics, heat "flows" from a hot place to a cold place, but not from a cold place to a hot place (unless energy is provided to drive it - 2nd law). The rate at which heat is "driven", so the number of Watts transferred is directly proportional to the difference in temperature between hot and cold side.

There are better performing blocks that manage approach temperatures of only a couple of degrees celsius, and there are ludicrously oversized, overfinned and overfanned radiators which hold the water resident for so long that it returns to almost ambient temperature. These are at the extreme end of the "normal" scale. You'll find you get very far beyond the point of diminishing returns if you aim for one of these.

For reasons I'm happy to explain later, the point that makes the most difference in a cooling loop is the transfer of heat to air (radiator). It's the slowest step and so determines the rate at which the cooling loop ultimately transfers heat, so the temperature of your CPU.

For people who live in the desert, this causes a problem as your CPU will roast, toast and generally fry eggs unless you help your cooling loop out a bit. What you need is to increase the temperature difference between the water and the air to a huge number, without making the water too hot for the CPU. Two ways have been traditionally used here:
1) Peltiers,
2) Refrigeration

These exploit the principle of heat engines - pumping heat from the cold side to the hot side. Consumes energy so the 2nd law isn't violated.

With a peltier, the hot side is cooled by the water, raising its temperature to something stupid (60ºC e.g.) Then you've got a temperature difference between water and air of more like 35ºC in a normal temperate climate, and a reasonably well designed radiator should dissipate 25ºC of those. In reward, the cold side of the pelt can easily achieve temperatures closer to 0ºC

With refrigeration, such as whatever's using, exactly the same system works. Wave your hand near the radiator of your fridge. Don't touch it, because they can be VERY hot, and you'll see what I mean.

The difference is this: with the first set up - a standard water loop, heat is being transferred simply by allowing it to flow from the hottest point to the coldest point. With the latter set ups, heat is being actively pumped from a colder point to a hotter point, the better to dissipate it to air.

If you're expecting miracles out of the first, you'll be disappointed, however it is still more effective than relying on a lump of copper, air and an 80mm fan.

[Soul99]

I built a TEC set up, I have a 226watt tec cooled with a waterblock with outside air ducked to a 120mm Radiator.
This takes most of the heat out of the water, but before the water goes back to the CPU block i have it passing through about 40ft of 10mm copper piping coiled and immersed in a chiller box full of water. there is also the cold plate from a fridge/freezer In a U shape which is sitting inside the chiller box. This gives extra heat capacity because of the extra volume of water, plus the ability to leave the fridge on over night and do some super cooling with -25 water temps.
I usually dont have the chiller on when the pc's off cus it doesnt need the extra capacity, it cools fine jsut turning it on when i turn the pc on, all the water starts at room temp and rises from there. i get coldside temps of about -36c with it off the cpu.

I live in the attic of a large house and its freezing in the winter and cooking in the summer, air cooling is nigh on impossible in the summer which is why i built the cooling. I have seen max CPU temps of 17c with my pelt cooler during a 20 hour up time on one of the hottest brittish days for 30 years.my other pc is at 67c with an AX-7 with 80mm fan Xp2400 at default speeds, all case sides off and extra cooling.
>
My room is 33c and i'm cooking!!!!

I wish i was as cold as my cpu.
LOL.

cant wait for winter temps, that frosty outside air will chill the bugger out of the water and i wont have to turn on the fridge bit at all!

[Daniel ~]

Thanks to you all! But it was Kaitain's postage that got us to the front page! ":O}

[phiphen]

[Taken from a purely thermodynamic perspective, there are some rules of thumb for minimum temperature differences between air and water, and water and CPU, at equilibrium:

Bare minimum difference between radiator fins and air is 10ºC for a normally-flowing radiator. There's a couple more ºC between the radiator and the water. Realistically expect the water to be 15ºC warmer than the room.]

I've reached equilibrium with my air cooled system... it's 14c difference at full load...

slk900u delta ehe 80mm

[Kaitain]

Originally Posted by phiphen

I've reached equilibrium with my air cooled system... it's 14c difference at full load...

slk900u delta ehe 80mm

Can you state what tests you did to bring it to full load, on what hardware, how long it was maintained at this workload before you started monitoring the temperature, and what equipment you used to measure the temperatures?

It's not enough just to offer the result - you could be running an underclocked Duron 750 for all I know

14ºC is very low for an air-cooled heatsink, so either you're running a fairly slow and old chip with that heatsink (in which case it will not be asking the lump to dissipate anything like as much heat as it can) or you're using the onboard temperature monitoring, rather than properly placed thermocouples, and are looking at an error of +/- 50% in that result.

[Kaitain]

Originally Posted by Daniel ~

Thanks to you all! But it was Kaitain's postage that got us to the front page! ":O}

Yikes! You mean.... people are gonna... read this?

[Soul99]

ohhh front page! where my pc pic's for BUMPAGE!!!! damn.

[phiphen]

oh jeez... I have a barton 2500+ running at 2324mhz. I have 3 thermal probes that came with my lcd rheostat. I put one at the back of my case, one on my hd, and one right next to the di casing on my heatsink.... THAT probe reads less than 14 degree difference.. I took 14 from the probe that's underneith my chip. (that came on the mobo) my computer is on all the time, so temp monitoring is very important. (also cause it's anywhere inbetween 85-95F in my apt. no AC) the temp in my case (witch is open, and has 4-80mm fans and 1 92 mm fan) is anywhere from 33-40C, the probe underneath the cpu reads anywhere between 43-54C. this is all reading done by motherboard monitor... that's after 3 days of up time.. and varying stress cond.(ie. the games I play, and my initial stress test.. to test the overclock. done by prime 95 for 1hr) as well as room temp's..

does anyone else notice that after about 20 minutes of prime.. it goes into a seemingly random stress's? definatly not as tough as the initial 20 mins.

[Kaitain]

Originally Posted by phiphen

I took 14 from the probe that's underneith my chip. (that came on the mobo)

Well you realise of course that I'm going to dispute your result then

I have argued repeatedly against reporting the results fed by motherboard monitor. Whether the board's using an in-socket thermistor or on-chip diode, the temperature monitoring hardware has not been calibrated, has not been tested and is not proven accurate.

To give you an idea, I had MBM reporting a temperature of only 55ºC for a rather poor heatsink using an in-socket thermistor bonded to the bottom of the (then ceramic) CPU base substrate. A properly calibrated thermocouple mounted to exactly the same point on the ceramic base read a true value of 62ºC (That's an error of about 13%mv). That was on a board using the relatively popular WinBond monitoring chips for hardware diagnostic.

I'm not especially familiar with the Barton core, having not had the time to pay a lot of attention to hardware in the last few months. Could also be that it generates less heat output than earlier cores. Anyone have a link to the spec on these?

Originally Posted by phiphen

I put one at the back of my case, one on my hd, and one right next to the di casing on my heatsink.... THAT probe reads less than 14 degree difference..

As it should! Fin surface temperature should be between 8 and 10ºC for fins in the middle, directly above the die, and probably only a degree celsius or two for fins at the very edges, which are largely wasted.

Unfortunately, reading fin surface temperatures only tells you that your heatsink hasn't broken the laws of thermodynamics...

[phiphen]

what are you talking about.. nothing is in the fins...like I said ... right next to the die casing.. UNDERNEATH the fins... also.. the temps I was talking about.. are from the probe thats underneath the chip itself...

[Kaitain]

Originally Posted by phiphen

what are you talking about.. nothing is in the fins...like I said ... right next to the die casing.. UNDERNEATH the fins... also.. the temps I was talking about.. are from the probe thats underneath the chip itself...

You're trying to tell me that you have lodged a probe next to your CPU core and it's reading the core edge temperature as less than 14ºC above case ambient? You are trying to claim that your CPU, which does actually put out a lot of heat (I checked) is being cooled by only air and manages that delta at the core yet a second measurement device (even one with a huge error margin) measures a higher delta?

Doesn't this tell you that, far from defying the laws on which this universe rests, your equipment's faulty?

[phiphen]

I said di casing... it's pinkish? well... which ever... those are my stats... read my temp logs if you like..that is my current temp. log... not my 3 day temp. log...

[Kaitain]

Originally Posted by phiphen

I said di casing... it's pinkish? well... which ever... those are my stats... read my temp logs if you like..

Die = core = die, and yes, that's a very pretty picture. If you post up your temperature logs, I will gladly look at them. I fail to see how parts of the world that should be colder are hotter in your computer, and vice versa. Perhaps given a decent set of logs I'll be able to work out just which instrument is going wrong...