DeltaT of 11c!!!!

[Player0]

Thats right! 1940mhz at 2.15vCore! CopperTop Maze3 is now running at a 11c deltaT!

Wooo! I managed this by doubling the number of springs on my motherboard standoffs, suppling about 2x the amount of pressure. I also used the absolute minimal amount of Arctic Silver 3 and cut away some of the neoprene gaskets that are used for water proofing.

And it worked! A new record for delta T at my test MHZ :) Wooo!

I guess pressure really makes a big difference :) It was a bear getting two springs on tho.

[mdzcpa]

Good job Player0:D

[DexterHolland91]

Looking forward to your review, I can't decide on which waterblock to go with as of now but maybe that'll change? :D

[MrP]

deltaT?????????

someone care to explain to us with less than one brain cell

toodles :drinky:

[kms]

differential temp MrP, as in change in .. :)

[Player0]

Delta T is the most important factor in water cooled systems. Delta T just means temperature differential, and could be comparing any temperature. What most water coolers mean by this (or at least me) is the difference in temperature between the CPU and the water.

So, if my water temperatures are 8c degrees, and I have a delta T of 11c, then my CPU temperature is 19c degrees.

The lower the delta T the better, since its directly related to the efficiency (C/W rating) of the waterblock, as well as other factors such as mounting, thermal interface material, etc.

Delta T will change radically at different CPU voltages and frequencies. I have a personal, standardized way of measuring the performance of different water cooling setups. 1800mhz at about 2.0-2.1v. Its not an exact science, and Delta T is usually the same if i am a few tenths of a volt off or upwards of 150mhz off.

I can compare the sucess of different waterblocks using these test parameters, MCW462 gets a dT of 22c, Maze2 of 16c, Maze3 of 14c, etc.

Of course, when I change the parameters by changing the type of mounting system (now double spring loaded), this affects my test setup. The Maze3 used to read 13-14c, now reads 11c-12c.

What this does to final temperatures isn't so important. Im now running at 2.1vCore at 1960mhz, and my dT is about 20c at this point. Maze2 would probably have been about 25c or so. MCW462 about 30c right now.

C/W is used as a rating index as well, but I've found dT to be a better guide. C/W is hard to measure accurately, and I've never gotten consistant results with it. You really need to accurately know the wattage of the athlon, and...thats not something easy to determine. AMD measures this under extremely controlled environments on multiple chips. I havent seen them release any numbers on overclocked Athlons though. So measuring C/W with an athlon is pretty much impossible.

[jaani]

Originally posted by Player0
Thats right! 1940mhz at 2.15vCore! CopperTop Maze3 is now running at a 11c deltaT!

Hi Player0,

Just wondering how the temp is measured, seeing as how socket probes have a habit of being terribly innacurate 'n all. :)

What heat exchanger/fans are you using to cool the heat load?

[DexterHolland91]

Player0 - Regards to that c/w rating, using a stock athlon can be half-way accurate, but a much better way is to use a peltier where you know how much heat is being given off on the hot side. This gives quite an accurate reading, and assuming you test all the other blocks in the setup it is comparible. :) I'd also like to know about your waterchiller man, so brag away hehe. Details? :D

[Player0]

Jaani,

I use the Abits socket probe (unless the KX7 reads the XP diode, which i dont think it does). I have used external temperature probes mounted right next to the CPU before, and Ive only ever seen a 1-2c degree difference from what the Abit sais. So I dont bother anymore.

I use a 450w peltier chiller. You can see pics in the Gallery. Anyway, its three 156w peltiers sandwhiched between two very large aluminum waterblocks built by 72510 of ForumOC fame. It makes a nice little heat exchanger unit for the dual loop cooling. The hot water loop is cooled by two BlackIce Extreme radiators hooked together in serial. Each one has a panaflow L1a and a Sunon 68cfm fan attatched, for extra quiet cooling with a medium air flow.

I may add a third fan to each, or add 1-2 more BIX radiators i to the system, to help cool things down during the summer months. Right now, with ambient about 26c, my hot water temperatures stay steady at about 40c. Cold water is about 8c right now.

My old dual loop system was able to get cold water down to 4c, using the same heat exchanger, but was much much more loud, and bulky, and heavy. This new system is much more streamlined and efficient, and the slight loss of power is because of less fannage on the hotside and the fact that Im cooling the Ti4600 and Northbridge with the cold water now.

As for C/W. Stock athlons are anything but stock when it comes to wattage. Each chip has different thermal characteristics, as every overclocker knows. Some run much hotter than others, depending on what part of the silicon wafer it came from. Edge CPUs can be slightly more 'fuzzy' and run at a different wattage. The numbers AMD give are a general rating, but aren't accurate for every CPU...so you cant use this to get a very accurate C/W rating. You can however get a good idea of the dT tho.

As for using a peltier, yes, this can be a very accurate source of heat. The problem is, it doesnt act like a CPU. Even if you create a plate for the TEC that resembles the top of the CPU, its not going to have the same thermal characteristics. So this, although more accurate, is still not the best solution.

No, to measure the C/W...you need to use an Athlon, and you have to know its exact, specific, wattage. The only way to measure the Athlons exact specific wattage is to connect it to a heatsink which you know the exact c/w for already. Like a solid block of almost-pure copper. With lots and lots of heat probes all mounted on it. And then you form a thermal average and can find the wattage. Which is similar to how AMD actually finds their numbers :)

Ive considered building a test bed like this for waterblocks, but I havent had much need. Simply measuring the dT between my water temperatures and my CPU gives fairly consistant results, and the accurasy isnt as important.

[DexterHolland91]

Hehe, I get what you are saying Player0, but a C/W rating isn't only for athlons. ;) It's merely how much the temperature of the medium increases for each watt added... correct? But I understand your point that it won't have any relavence to CPU cooling if the die size is not that of an athlon. Still, a rating is a rating, and a good way to compare blocks or heatsinks.

[Player0]

Right Dextor, its just a standard measurements, but if not for athlons, who would care about it? ;)

Anyway...i understand what you mean, that even if C/W isnt accurate, its still a good way to compare different heatsinks. But heres the problem...no one agrees on C/W. For example, many manufacturers now list a c/w rating, but these can be incredibly innaccurate and false. So someone looking at these numbers when comparing heatsinks could be in for a big surprise.

And then take that one step further, many websites that will remain nameless here, have tried to create test beds to measure this. The results? Well one website sais heatsink #1 has a lower c/w than heatsink #2, and another website sais #2 is better than #1.

Why? Because c/w is a measurement thats hard to get right. Its very sensetive to so many factors that its simply unreliable for people to go by, imho.

For waterblocks, delta T is just a better measurement, since it seems to be more true to life. I can duplicate the same results over and over using my method, and it seems to hold true to other temperature results that other people get :)

[DexterHolland91]

True true, the problem is that ANY testing of any sort, be it even standard testing on a heatsink is not comparible across sites, it just isn't and that won't work. Even deltaT is not comparible on different test beds.

[Player0]

No dT cant be compared between test beds.

But, the key is...no mater what the test bed is, the better waterblock will have the better delta T with all things being the same.

Where as the c/w rating, being more susceptable to slight changes in the test bed, may not be as standard. So one testbed #1, one heatsink does better than another, but on testbed #2, the results could be reversed.

Measuring delta T is usually more true-to-life, so its easier to determine which waterblock is better :)

[AFroDYme]

its like Col. John "Hannibal" Smith says "I love it when a plan comes together"

( I think that maybe true for a WC setup too :D )


nice work Player0

[DexterHolland91]

Player0 - The problem is that all factors will never be the same. You give a simple example, where clamping pressure gave a 3 degree drop in temps. But it is still much easier to measure and stuff. And still is accurate as long as the test conditions stay the same.

[LiquidRulez]

Can someone PLEASE explain to me the best way to measure my Delta t??:confused:

[DexterHolland91]

Temp of CPU minus the temp of the water. Did you not read the first page? P0 explains it in more detail.