Lowest C/W straight Water?

[schoolie]

Hi,

I was wondering what the lowest C/W possible could be in a straight water cooled system, measuring the delta using the inlet water temp, and measuring the core using the thermal diode. Do you think 0.10 C/W is possible?

Thanks!

Edit: I meant lowest. Sorry

[redleader]

Highest? Thats easy, just don't use a waterblock and you'll come damn near infinity, for a few moments anyway

Lowest, I'm not sure. Most results use the practically worthless external probes. Once onboard measurements start to come in it would be easier to say.

[#Rotor]

I think it should be possible....

[Brad]

if you're using the ondie temp probe .1c/w is possible, but using a proper temp probe, I haven't seen lower than .23c/w

[mkosem]

my current system is pulling .063 C/W. I have a maze1c-1, heatercore, and eheim 1048. I just have some panaflow h1a's at 6v flowing over my core. my processor is a celeron 1.2gig@1.6gig producing 62w

--MAtt

[schoolie]

Interesting. The Celerons have the heat spreaders, right? Maybe the extra surface area helps more than the extra interface hurts.

[Brad]

it will be helping, but that temp probe is seriously ****ed. mk, do you get the same c/w at differant heatlods?

[Butcher]

C/W between different systems is a meaningless comparison. You can only compair if using same equipment etc.

[Dix Dogfight]

C/W IS a valid way of comparing between different systems.
IF you can measure the temperature and wattage correctly.
However many tests don't measure the temp in a correct way and therefore any comparison is meaningless.

[derraa]

I could see someone getting .1c/w. I get .15 with my externaml probe.

[schoolie]

There must be a minimum C/W. If you could model the core with the top held at a constant temp, and the power constant, then I wonder what the equilibrium thermal diode temp would be.

[Butcher]

Given the accuracy (or lack thereof) of measuring of termperature, and the complete lack of a way of measuring wattage, C/W will never be valid when taken from an installed system. The only way to get comparable C/W is to use a specially developed test bed to test the equipment in.

[schoolie]

Well, the die simulator has inaccuracies too. It does produce consistent results though. I'd like to see a test that eliminated the mounting of the water block as a factor.

[mkosem]

that C/w is at 100% load for 4 hours

--MAtt

[Butcher]

100% load varies a lot depending on the programs running, what windows decides to do etc. etc. There is no way of guaranteeing wattage for a cpu, hence the use of simulators.

[mkosem]

that is with seti and toast both running high priority. It's as close to 100% as I can get!

--MAtt

[EMC2]

With realistic water flows, a pure copper waterblock that fits within AMD's space guidelines, and the XP processor you could get down in the 0.09 C/W range. To get below that, you would need to use some type of heat pipe base or a material with a higher thermal conductivity than copper ( say, pyrolitic graphite ).

"Regular" heat spreaders on processors do not improve matters, in fact they make things worse, by virtue of the fact that they add another heat transfer boundary. The "good" of a heat spreader is that it tends to be a bit more forgiving in the case of incorrectly mounted blocks and helps to protect the core from damage. There is nothing "magical" about them, they are just copper and thermal paste.

[schoolie]

Thanks EMC2. It's nice to have a number for sanity checking. Still waiting on my Maxim chips. I may lose my patience and buy the ABIT NV7-133R board which is supposed to have a working thermal- diode reader.

Maybe I'll steal the Hope Diamond and make the worlds most expensive, and best performing water block

[Brad]

I'm thinking of a #rotor style base, 1" inlet, 2x 3/4" outlets, then running 4 heatercores in parrallel. I bet that'd do better than .1c/w

[schoolie]

I've always liked #rotor's blocks. I've never seen any test results on them, but they look great! What kind of flow rates would be needed to get that kind of performance from his twin outlet block?

[Brad]

I don't know, I'd run 2x 400gph or bigger though