My radiator has a C/W of 0.01?
 

I'm gonna get shot here with my 'precision' themal probes but still...... ;)

OK, I'm using cheapo themal probes typical of PC shops:

http://store.over-clock.com/Miscellaneous2.html

BUT, I do have one accurate (well, +/- 0.1'C) digital thermometer which I used to calibrate them against.

OK, so, my ambient temp is 23'C, water at 26'C and total heat load of 290 watts.

3 / 290 = 0.01 C/W

Is this possible? :shrug:

The radiator is HUGE (40cm x 30cm) and double-pass. I also have 4 x 120mm fans pulling air through (shroud depth approx. 5cm). Each one is rated at 44cfm, giving a total (unrestricted) flow of 176 cfm, although I have overvolted my PSU so the 12v line is actually 13v, so the (unrestricted) flow will be somewhat higher, perhaps 190cfm.

Flow rate is approx. 5 lpm (approx. 1.3 - 1.4 US gpm).

Looking here:

http://thermal-management-testing.com/ThermoChill.htm

It seems the largest rad is 0.02 C/W, but my rad has an approximate surface area of 1200 cm^2 :drool: , whereas the 120.2 only has 288 cm^2 ;)

If it makes any difference, the radiator I use seems to match quite closely Cathar's suggested 'ideal', although it is thicker, at 1.75" rather than the 0.75" -> 1" suggested by Cathar. The fin density is quite low though, so it really shouldn't be very air-restrictive.

Thanks for any suggestions as to how realistic this 0.01 C/W is. Given the high airflow and huge surface area, it may seem reasonable?

Yes its possible.

How do you calculate your 290 watt heatload ?

I thought someone would ask that ;)

290 watts is good huh? :drool: :D





CPU = Venice 3000+ @ 2.916 Ghz, 1.7 volts

67 watts x (2916/1800) x (1.7/1.4)^2 = 160 watts



GPU = X800XT - default = 60 watts (probably more like 40 for GPU itself). Running at 575 / 550 = 69 watts (maybe more like 46 watts for GPU alone).


Both Harddrives WCed = 18 watts

Two 50Z pumps @ 13v = 23 watt heat dump

NF4 chipset = 20 watts (guess).

Total = 160 + 46 + 23 + 18 + 20

= 267 watts (290 if 69 watts used for gfx card).

3 / 267 = 0.011 C/W

I don't think you can make a proper estimation of CPU heat by clock and voltage ratio like that... 160 watts seems way too high

Heat ouput is linear with clock speed and increases with the square of the voltage.

It is correct.

Hey Incoherent!

I am going to Sweden (Gothenburg) in June for a few days for business, anything to see?

I should have a Monday afternoon and a Thursday morning free.

One would actually want to quantify the heat load entering the cooling loop, without estimates, before making a statement as to the C/W.

Would need approximately ~165CFM at the least to reach a 0.01 C/W, and that presumes 100% radiator efficiency.

The fans may be 44CFM each, but unlikely that they'd be pushing that much air-flow through the 1.75" thick radiator.

Still, what you're proposing is potentially possible, but I'd still want to know what the actual heat load is before I believe it to be anything more than a guess.

Still, when it comes to radiators, surface area rules all, and you've definitely shown that to be true.

Gothenburg is a long way from me unfortunately.
I don't know that area of the country at all actually so I'm not much use, not being a native here.

man, thats a big Rad!

Heat Load - Agreed. Do you think my estimates are off?

Airflow - Agreed. As I stated, the airflow of 190cfm is unrestricted airflow.


OK, no worries!


It's an actual car radiator from, I believe, an 1980's Renault 18. I can't stand the price / performance ratio of PC WCing rads and this one was £10 from Ebay. I just needed to fit barbs to it and make a shroud. Luckily, I didn't need it to fit in a case!

Where do you have it mounted? External box? On a case panel? I've always wanted to make a case with car radiators for panels. The whole computer could be passively cooled.

It's mounted externally on a frame which also supports my two AQX 50Z pumps (in series). They are housed inside an aluminium box which contains 1" foam to muffle them.

I ought to get some pictures 'cos the whole system is, erm, interesting....

Not all the watts go into the water, especially when they run significantly hotter than the ambient in the case.
The CPU socket, motherboard, and electrical traces all get warmer, and then are all kept at a stable temperature via air currents within the case.
However, I'd not be suprised a full sized radiator outperforms a large heatercore.

:p EURICA!

I have found it!
The perfect radiator!

Pics

:p

:p EURICA!

I have found it!
The perfect radiator!

Pics

:p

Thats just a rough approximation. Besides, theres no way your Venice actually dissipates 67w. Thats just the TDP, not dissipated.

Good work! :D

Good work! :D

OK, fair enough.

Good work! :D

OK, fair enough.

Good point.

A picture (including volt-modded X800XT):

http://homepage.ntlworld.com/james.w...0003_small.jpg

Pumps are in a black aluminium enclosure behind the speaker.

Rad is mounted on top of the frame, fans sucking through and blowing air upwards.

Strange big 'L' shaped tube is res / airtrap which feeds the pumps using a 5/8" connection. Rest of the system is 1/2".

Front panel of PC case shows DIY fan controller (LM317 chip) allowing voltage control from 1.25 volts -> 10 volts or switching to direct power (13 volts due to overvolted PSU).

Molex on front of case (above fan controller) is connected to the X800XT and allows measurement of core (1.70 in this case) and adjustment using small blue potentiomenter.

Bloody hell! Very stealth ;)

I would say your wattage calculations are a bit out but the C/W is probably in the ballpark.

Cool setup.

Heh thanks d00d!

Stealth?

Re: Wattage - I wouldn't say 260 watts fully loaded is THAT far off, but I get the idea.

I have been volt-modding my gfx card to 1.8 volts (up from 1.40) and upto 670 Mhz (from 500 Mhz) so that should add some heat ;) I don't run it like that all the time though!

3DMarks (for those who are interested):

2k01 = 32k
2k03 = 15.5k
2k05 = 7.4k

I like getting the best scores I can from medium value hardware.

Venice 3000+
X800XT
512MB RAM

I tell you what, I think my calculations are spot on!

Here is a very different heat load:

CPU undervolted to 1.3:

(1.3/1.4)^2 x 67 = 58 watts

HDDs = 18 watts

Chipset = 20 watts (estimated)

(Graphics card out of loop)

Total heat load = 96 watts



Air temp = 21.6
Water temp = 22.6

1/96 = 0.01 C/W

;) :cool:


CPU temp = 29'c

Water temp = 22.6 'C

6.4/57 = 0.112 C/W

Estimated C/W of NexXxos block, given it's current flowrate of 5lpm (1.4 US gpm) = 0.11 - 0.12

;)

I'll be trying different heatloads............

Not bad at all. Here's all I could manage from my ~$200 (cpu + mobo) Socket A system:

2k01 => 26035
2k03 => 14493
2k05 => 7165

That's pretty good!