Articles: Cooling

Bookmark and Share

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 ]

Thermal and Acoustic Performance

My today’s test program opens non-traditionally with a noise level test. I want to show its results first so that you can see what operating mode of the Koolance Exos-2 LX system is the most practical, i.e. comfortable for constant use. In this test the fan speed was being changed manually along the 10 available grades and the noise level was measured from a distance of 1 meter. The pump was turned off during the test. Here are the results:

As you can see, the third and fourth speeds of the two 120mm fans are subjectively comfortable – the difference between the two speeds is negligible in decibels. The further increase of the fan speed leads to a considerable growth of noise. At speeds higher than the sixth one it is quite uncomfortable to be near the operating Exos-2 LX for a long time. As for the maximum speed, you may only want to use it for benchmarking and setting new records. The noise produced by the pump is also shown in the diagram, but the pump isn’t noisy even at its max performance although audible when the fans of the Exos-2 LX are working at the fourth or fifth speed.

The next diagram shows the temperature of the overclocked quad-core CPU (the liquid cooling system has only one water-block in the circuit – for the CPU).

I should first note the low efficiency of the CPU-330 water-block. The Zalman ZM-WB5 keeps the temperature as many as 8°C lower than the CPU-330 in the quiet mode. Frankly speaking, I had expected the water-block from Koolance to prove more efficient than Zalman’s, but it’s just the opposite in reality. The gap between the water-block diminished when I increased the pump performance and the fan speed, yet the Zalman ZM-WB5 was ahead anyway.

Comparing the two liquid cooling systems with the same water-block (Zalman ZM-WB5), the Koolance Exos-2 LX cools the CPU better by 2°C in the quiet mode. The gap grows to 8°C at the max performance, but the Exos-2 LX is considerably noisier than the Zalman Reserator XT (53.3dBA against 43.8dBA). Note that it is the increase of the fan speed, not of the pump performance, that contributes the most to the reduction of the CPU temperature. Moreover, if the two fans of the Exos-2 LX system are set at 1080rpm and the pump performance is changed from comfortable to maximum noise, the CPU temperature doesn’t change under load and even rises up by 3°C in idle mode. Running a little ahead, I can tell you that the efficiency of the liquid cooling system depends more on the pump performance if the hydrodynamic resistance is increased, which is quite logical.

The maximum stable frequency of the CPU when cooled by the Exos-2 LX in the quiet mode with the Zalman ZM-WB5 water-block was 3610MHz at a peak temperature of 67°C and a voltage of 1.6125V. I performed this test the last of all when the mainboard’s BIOS had been updated to version 0601, the CPU Voltage Reference was set at 0.63x and the CPU Voltage Damper was set at Enabled. As a result, the CPU was stable at a lower voltage, but a higher temperature (by 3°C). When I selected the highest fan speed, the CPU was stable at 3708MHz and had a peak temperature of 61°C, which is the absolute record for this CPU in our tests. Unfortunately, the Exos-2 LX is too loud in that mode.

Then I added the chipset water-block into the circuit. I had to remove the heatsinks with heat pipes from the mainboard and install copper heatsinks onto the power elements. The ASUS P5K Deluxe cannot monitor the chipset temperature, so I measured it using a thermal sensor included into the Exos-2 LX kit. I first attached it to the base of the standard copper heatsink that was installed on the mainboard originally and then I glued it to the base of the CHC-120 water-block. The chipset proved to be 60°C hot with its standard passive heatsink but only 39°C hot with the water-block (there can be a minor inaccuracy about the numbers). I guess that’s a superb result but it is only going to be practically valuable if you want to overclock the CPU by increasing the system bus frequency above 500MHz. I can also add that the CPU temperature grew by 2°C under peak load (the CPU was overclocked to 3610MHz at 1.6125V in this and subsequent tests) when I had included the chipset water-block into the circuit.

Next I added the graphics card water-block into the circuit to work along with the CPU and chipset water-blocks. The VID-282 was cooling the overclocked GeForce 8800 GTX in this test. The diagram shows the temperatures of the graphics card’s GPU and PCB components.

The performance of the liquid cooling system is impressive in contrast with that of the GeForce 8800 GTX’s stock cooler. Such a hot graphics card proves to be no hotter than 60°C (GPU temperature) and 50°C (PCB temperature) with the liquid cooling solution from Koolance. Moreover, the thermal sensor of the Exos-2 LX system attached inside the closed system case reported a decrease from 50°C to 34°C in the air temperature when the graphics card’s cooler was replaced with the water-block and the system was running Unreal Tournament 3 for an hour. And to remind you, the blower of the GeForce 8800 GTX’s stock cooler exhausts the hot air out of the system case! I wonder what the difference would be in comparison with graphics cards whose coolers do not exhaust the air.

But while the graphics card temperature decreased dramatically on my adding the VID-282 water-block into the circuit, the temperature of the CPU grew up after that:

The Firefly Forest test from 3DMark06 doesn’t load the CPU fully (as you can see by the difference in the CPU temperature with and without the graphics card water-block). On the contrary, the CPU temperature grows suddenly in Unreal Tournament 3 even when the graphics card water-block is removed from the circuit and the GeForce 8800 GTX works with its stock cooler. You can only reduce all the temperatures by increasing the fan speed and the performance of the pump.

The only water-block I did not test was the Koolance HD-55-L06. The temperature of the Samsung HD501LJ hard disk drive installed in our testbed is not higher than 39°C even during long write/seek operations, so there is no sense in installing a water-block on such a cool HDD. Liquid cooling may come in handy for Western Digital’s Raptor drives, but I didn’t have to such HDDs at my disposal.  

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 ]


Comments currently: 8
Discussion started: 11/30/07 03:49:16 PM
Latest comment: 03/31/08 02:14:16 AM

View comments

Add your Comment