Puget Custom Pentium-M Rig: A Silent WC System

Complete|Mobile Systems
Viewing page 5 of 6 pages. Previous 1 2 3 4 5 6 Next


The following tools were used during testing:

Because the system noise is more or less constant — there are no fans to change speed, and drive seeks were not noticeable above the idle noise — most of our testing consisted of an examination of power and thermal dissipation. The CPU thermal diode was not calibrated, so thermal measurements should only be evaluated relative to each other, not in absolute terms. Unfortunately, the VGA card in the system did not report temperatures, so we were unable to monitor this temperature.

Ambient temperature during testing was 23°C. Ambient noise level was 16 dBA/1m. SpeedStep was not supported by the motherboard, so it was not tested.

Puget Custom Computer
Activity State
CPU Temperature
AC Power Draw
3DMark05 + large file
copy from CD to HDD


At idle, the CPU sits at a reasonable 31°C — below the reported "system temperature" of 34°C, and only 8°C above the ambient temperature outside the case! This is what happens when you use watercooling to cool a laptop processor. If Speedstep was enabled on the motherboard the temperature would drop even more, but there's little practical benefit to this other than bragging rights.

The 66W that the system draws at idle is surprisingly high for a Pentium M system. By comparison, The AOpen EY855-II SFF system that we recently reviewed drew only 34W in idle (Speedstep disabled), half what our system draws. Obviously, that system did not need to power a water pump, but the Aquastream 3.0 is rated for 4.5W typical usage, only marginally higher than the combined power draw of the fans in the SFF system. The culprit is probably the power hungry X800XL. We could not find detailed information about the power consumption of this card, but most current mid-to-high end graphics cards draw in the range of 60-70W under load. It's not hard to believe that it draws ~30W at idle.


The thermal curve of the Puget system under load was unlike any other we've seen before. Instead of gradually increasing in temperature before stabilizing, the temperature of the CPU jumped to its stable temperature — 42° — as soon as CPUBurn was enabled, and then stayed there indefinitely. This phase of our testing took much longer than usual because we wanted to be sure that the CPU temperature would not increase as the heat capacity of the radiator was reached, but even an overnight session running CPUBurn did not manage to push the temperature above 42°C when it was checked the following morning. Turning off CPUBurn at this point immediately dropped the temperature down to 32°C, presumably close to the temperature of the water in the radiator.

Because we did not have any way of reliably determining the ambient temperature inside the case (the motherboard sensors are unreliable because they can be affected by heat-producing components of the motherboard), we were unable to calculate the cooling efficiency of the water cooling system. Suffice to say that a 42°C peak temperature provides more than enough overhead for warmer climates and overclocking.

The low 27W Thermal Design Power of the Pentium M 760 is very noticeable when running CPUBurn. AC power draw only rises 10W above idle, much less than the ~80W increase that is typical of the latest P4 chips. Assuming a 70% power efficiency for the Antec Phantom 350 at this load, the total DC load on the power supply is only 53W ? less than a sixth of its total capacity.


3DMark05 is primarily a graphical benchmark. Although it also stresses the CPU, the primary load is on the VGA card. This is reflected in the marginally cooler CPU performance and the much higher power draw compared to CPUBurn.

The high power draw of the X800XL is again apparent in this benchmark, as the AC power draw is 35W higher than the CPUBurn test, for a total of 111W AC. Assuming 76% efficiency, this translates into 84W DC load, still only a quarter of the Phantom's rated capacity. However, the fanless design of the system means that the Phantom does not have any airflow to keep it cool, so it's probably a good thing that the peak draw is so low.

3DMark05 + large file copy from CD to HDD

In an attempt to discover the highest possible power draw of the system, we also ran 3DMark05 while copying a large chunk of data from a CD to the hard drive. Not surprisingly, CPU temperature was not significantly different from the other tests we did, but the power consumption did increase by almost 10W beyond what was seen for 3DMark05. Even in this artificial stress test, the total power draw peaked at only 120W AC.

Previous 1 2 3 4 5 6 Next

Complete|Mobile Systems - Article Index
Help support this site, buy from one of our affiliate retailers!