Performance

Test Setup

CPU : - XP 2100+ (t-bred core, AUIHB stepping), P4 2.4B GHz
Motherboard : - Epox 8rda+, 2.03XP Nvidia drivers , IT7 MAX 2 V1.0
Memory : - 2x256MB Twinmos PC3200, dual channel,
Video Card : - Sapphire Radeon 9700 Pro, Cat 3.2's
Hard Drive : - IBM 120GB 180GXP 8MB cache
PSU : - Thermaltake 420W
Case : - Lian Li PC60, 2x intake, 2x exhaust fans
Windows XP + Service Pack 1, DX 9.0a

In testing I installed several programs that would put the CPU under heavy load, Seti, Folding @ Home, SiSoft Sandra Burn-in Wizard, 3DMark2001SE 330, 3DMark2003, and ran 2-3 programs at a time constantly looping for several hours then taking down a temperature reading 3 times over 5 minutes to check for abnormal spikes in temperature, either highs or lows.

Ambient air temperature is measured using two different thermal sensors. One is slightly more accurate to 1 decimal place the other is only accurate to whole numbers but helps to verify that the probe's are working correctly, CPU temp is being read by onboard motherboard monitoring, with a temp probe also used with the AMD CPU to verify the results are accurate. The probe could not be fitted to the P4 due to lack of space around the retention bracket.

I was hoping to add a C/W rating, efficiency rating for those of you who don't know, for an the XP 2100+ and the P4 2.4B GHz but due to inaccurate or conflicting information absolutely everywhere I looked, I couldn't find a specific set of numbers to work off. For instance, AMD's PDF white paper files are updated very regularly, so finding accurate information on a chip you brought more than a couple of months ago is almost impossible. The main problem is both companies prefer to only manufacture a few different CPU's at a time, instead of producing all of the different speeds separately. So they produce only 2-3 different quality cores at any time and during final packaging the multipliers are changed to adjust the final clock speed. This means that for any chip you have, you could often have an under clocked faster CPU which would have slightly, but importantly different thermal specifications.

There are also independent sites keeping records for us to use, but these appear to be inaccurate also, with most quoting the 3.06GHz P4 at 99W max output while at stock. This being the latest chip is very easy to find the rating for and it's 81.8W. As a result I won't be using information directly from those sites though the calculation script I have used should be fine.

I also want to state, before the emails flood in, that there are as many opinions on how to calculate the thermal efficiency of a water cooled system as there are hours in the day. Some seem to suggest the ambient temperature should be based on water temperatures, others claim it should be air ambient in the case and yet others say air ambient at the fan intake. Personally I go with the latter, but what about heat lost directly from the block or the tubing? The truth is it's quite a complex business to work out a genuinely accurate C/W rating for a water cooled system so I suggest you take as much notice of the actual temperatures as you do of our C/W rating.

So on to the results:-

Let's start with the AMD results. They are some pretty impressive numbers, 51C is regarded by some as high, and personally I wouldn't have a problem running a CPU at that temperature. However 51C is under full load, and at a 670 MHz overclock and under high voltage. Clearly we're seeing some inaccuracies here as in theory a C/W rating should remain constant across the scale even as the CPU wattage increases therefor I'm assuming it must be the increase in voltage that's screwing up the equation. Of note however is the fact the C/W rating improves from 0.389 to 0.285 as we crank up the voltage and the frequency and I'd suggest that 0.28 is nearer the actual C/W rating in normal use.

Stock temperatures are very good too, 35.5C is very impressive, even at stock this chip is fairly fast and churns out a reasonable amount of heat.
I was unable to hit 2.4Ghz with my SLK 800 which was previously installed in this system, the Hydrocool enabled me to overclock a little further at stable speeds and still at good temperature.

With the P4 results I've included the C/W results that just wouldn't have been possible with any accuracy on the XP 2100+.

At stock settings the results aren't fantastic, but the increase in heat when running a fairly high overclock (for this chip anyway) and using the maximum VCore that the IT7 MAX 2 will give me show that the Hydrocool is still very efficient raising just over 5C. As this unit is designed to dissipate 200W of heat, I think this is down to this individual P4 being a particularly hot running one. As the heat output has gone up by 50% or so and the temperature has increased only 1.5C, I think this shows the block and radiator can handle the heat.
The turbo mode again didn't make much of a difference and I can't see that you'll need to use it at all.

The C/W scores obtained for stock speeds aren't particularly good at all, though again I suspect the chip just won't run cool at all. The C/W scores while running overclocked are much more impressive, still not stunning, but more than good enough.
At just over 80W, a temperature of 46C in a quite warm room is very good. It shows this unit could more than handle a 3.06 GHz P4 with an 81.8W output. Given the temperature, I'd expect a 3.06 GHz to still be effectively cooled by the Hydrocool200 even at a reasonably large over clock and still quietly.

A Closer Look

Conclusion