As for the thermal conditions, there are two values you should take your bearings from: the CPU temperature (according to the built-in sensor) and the evaporator temperature (the number displayed on the front LCD display of the cooler unit).

Let’s start with the evaporator temperature. Unfortunately, there is no linear correlation between it and the real temperature of the CPU, so you can’t use the data from the Evaporator Temp sensor for continuous thermal monitoring. Under a small load (for example, a Pentium M processor with a TDP of about 30W), the difference between the ambient and evaporator temperatures is about 80°C. This is a strict correlation: suppose there’s -55°C on the evaporator at +25°C in the room, then the evaporation temperature is going to be -65° if the room temperature is lowered to +15°C. It means that the cooling efficiency is going to be somewhat lower in summer, but on the other hand, the enthusiasts will be able to squeeze an extra hundred megahertz in winter by ensuring a low ambient temperature.

The highest temperature of the evaporator we observed during our tests was -33°C – that was with a Pentium 4 660 overclocked to 5.1GHz at 1.65Vcore. On average, when the extremely overclocked processor is under a high load, the temperature of the evaporator varies from -36°C (an Athlon 64 FX-57 @ 3550MHz at an air temperature of +35°C) to -45°C (a Pentium 4 Extreme Edition 3.2GHz @ 4300MHz). When the CPU’s subunits are mostly idle (like in typical office applications), the temperature goes lower still by 5-10°C. For comparison: the minimal evaporator temperature of the earlier tested Asetek VapoChill XE was -32°C and the average range was -10 to -20°C (at +20°C air temperature). Well, the Prometeia Mach II GT is a direct rival to another cooling monster, the VapoChill LightSpeed [AC], rather than to the mentioned XE.

By the way, the picture drawn above does not refer to the Pentium M. Thanks to the very low heat dissipation of this processor (in comparison with the desktop monsters) the evaporator temperature always remained at -54-55°C during our tests irrespective of the load, clock rate or Vcore! The ECT Prometeia Mach II GT just doesn’t seem to feel the difference between taking off 10W of heat (in the idle mode) and 45W of heat (at 2700MHz clock rate and 1.6V voltage) – this cooler was developed with much higher loads in mind!

The manufacturer supplies detailed data about the “reference” testing of the Mach II and Mach II GT on a special testbed with an accurate control over the thermal load. The temperatures given are very close to what we got in our own tests, which does credit to the vendor: ECT publishes real-life data!

The CPU temperature is somewhat harder to work with. Many mainboards (particularly, from AOpen and Gigabyte) cannot correctly recognize below-zero CPU temperatures which makes any monitoring impossible. It is not ECT’s problem, of course, but the mainboard developers’ who did not take the trouble of making the thermal sensor operable under such conditions. The problem is partially solved in mainboards from other vendors, for example in those from ASUS (which we mostly used to test the Mach II GT). ASUS mainboards give you a more or less correct temperature down to -36°, but at the lower temperatures they just show you -111.5°C at once, irrespective of the real temperature of the CPU. Well, even when the chip that processes the data from the CPU-integrated thermal diode behaves reasonably, you should not trust it too much: the measurement error range is larger at lower-than-zero temperatures.

Thus, the minimum CPU temperature we could be sure of was -36°C, although the real temperature might have been a few degrees lower. The temperature of the Pentium M might even be as low as -45°C, but we could not make sure about that due to the problems with the AOpen mainboard. The average temperature of a processor working under normal conditions was usually -25-30°C, depending on the frequency and voltage, if we don’t count in the extraordinary result of the test of one of the earliest samples of the Prescott (3.2GHz, Socket 478), a product of the “thermonuclear oven” class, for which -10°C was the normal temperature. We don’t even publish the result of this test in the table above because, fortunately, few people bought this “CPU marvel” for use.

Under full load and at the maximum frequency, the Pentium 4 660 proved to be the hardest trial for the Prometeia (quite naturally, considering its impressive frequency of 5100MHz and high heat dissipation). Its temperature was -8°C. The temperature of the other processors varied at overclocking from -10°C to -25°C.

During all the time we were using the Prometeia we never had any technical troubles. With any PC configurations and under any temperature conditions (including long operation in summer heat and under high humidity) our Prometeia Mach II GT worked without a hitch, ensuing high performance and giving us no problems at all. This behavior was a nice surprise after the rather capricious Asetek VapoChill XE.

The fasteners are sufficiently robust for an average user: the evaporator and other details are in ideal condition even after a hundred install/uninstall cycles, except for the screws – their cross-like dents wear off after 15-20 cycles, so you should still be careful during installation.