BlueSmoke - Review : Arkua 6149

		Date	: May 7th, 2002
		Category	: Cooling
		Manufacturer	: Arkua
		Author	: Jin-Wei Tioh

A lot of people have heard of Arkua and/or Thermosonic, but what most people don't know is that this heatsink manufacturer also makes 1U heatsinks. What is somewhat mind boggling is that an entire 1U server case is only 1.75" thick - leaving just barely 30mm for a cooling solution. The average cost per hour of downtime (according to Strategic Research Corp) can range anywhere from $89,500 to $6,500,000! Therefore the efficiency and reliability of a HS/F design is even more critical in a 99.999999% uptime server environment where system failure due to an overcooked processor simply will not do. 1U heatsinks have other potential uses as well. A good example would be in Project Numenor as a low profile replacement of the Spire 5P53B3 if need be. People looking for a high-end solution to cool their graphics chip also comes to mind. Without further ado, let's take a closer look at the Arkua 6149 1U HS/F.

		Our evaluation unit came in a plain white box simply marked "Arkua 6149", but with the proper protective packaging inside. A small instruction booklet and packet of silicon grease are included.
		The 6149 came installed with a Bi-Sonic Technology BS601012H fan. 60x60x10mm (HxWxD), ball bearings, 3-pin tail, 4800RPM and pumps out 23.7CFM. Subjectively speaking, this fan pushes a surprising amount of air, which would account for part of the 6149's relatively high efficiency. And it does all this while remaining whisper quiet, definitely a big relief from the black label Deltas.
		The basic design of the entire Arkua line of heatsinks is exactly the same, the only difference being dimensions, presense of a copper core and fan type. Not that it's a bad thing, since the design is similar to the original Thermosonic Thermoengine that is still a viable HS/F today. Basically, Thermosonic and Arkua merged; Thermosonic manufactures the products, Arkua handles all the sales and marketing. If you compare the Arkua 6149 for its bigger brother, the 7528, you will notice that the fin design is the same - curved away from center and corrugated. This serves to decrease laminar airflow and thus increase turbulence (the air molecules come into contact with the fins more often), drawing out more heat. A full copper version of the 6149 would definitely have been interesting.
		Arkua opted for a generic clip, similar to the one used in Thermaltake's 1U HS/F (DTFCF014-1). While not quite as good as the original Thermoengine's clip, it is still fairly easy to engage with a screwdriver, holding the unit down firmly.
		The machined based is a pretty decent piece of work, with the mating surface being flat and even. However, it is a little rough as machining marks can be felt (just barely).

These graphs show the temperature delta between the CPU case temperature and the intake air temperature. To maximize CPU power dissipation, we ran StabilityTest, HotCPU Lite and the BlueSmoke RC5-64 Client simultaneously under Windows 2000. The test unit basically consisted of an AMD Duron 800, AOpen AK73 Pro (A) (courtesy of AOpen), a S3 Virge PCI sitting in the last PCI slot and the measuring apparatus (courtesy of HighSpeed PC). The unit was setup on an open tabletop, giving the cooler an ideal functioning environment (ie. unobstructed airflow, minimized recycling of hot air). Air turbulence in the room was virtually non-existant, ensuring that only the fan on the heatsink was doing all the work. Other heat sources (eg. lights, monitors) were also minimized to ensure a more accurate determination of performance results. About the only confounding factor is the heat generated by some of the nearby components, eg. RAM, voltage regulators and the north bridge. Lastly, the primary TIM (thermal interface material) used in the evaluations is Arctic Silver II, kindly supplied by Arctic Silver LLC.

This graph shows the thermal resistance (°C/W) for the heatsinks under each condition. °C/W is both a portable & reliable measurement due to it's dependence on only the temperature delta between the CPU case and the intake air. While absolute temperatures may vary from place to place, you would only need a pair of thermometers that agree with each other to obtain a reasonably accurate °C/W value. To calculate the °C/W of a cooling solution :

°C/W = [CPU Temperature (°C) - Intake Air Temperature (°C)] ÷ Watts

The result would usually fall in the range of 0.2 - 1.5, the lower the better as we'll shortly see why. A rough temperature measurement shouldn't be too difficult for anyone, but obtaining the power dissipation for your CPU (Watts) might present some difficulties). Fortunately, Chris Hare has compiled all the necessary information for nearly all CPU types. All fine if your CPU is running at stock speed and voltage. But what if you overclocked it? No problem, just use :

OC'ed Power (Watts) = Stock Power (Watts) * [(OC'ed Speed (MHz) ÷ Stock Speed (MHz)) * (OC'ed Voltage ÷ Stock Voltage)^2)]

The lower the °C/W, the better? Given the °C/W for a heatsink/fan, you can calculate how well it should work in your system via the following formula :

CPU Temperature (100% Load) = Intake Air Temperature + (°C/W x Watts)

The Arkua 6149 turns out be a rather capable cooler for its size. While we didn't have access to other 1U HS/Fs for comparison (the CoolJag JAC311C comes to mind), the 6149 still puts up a show, besting the Spire 5P53B3 and coming up on the Mini Copper Orb - both heatsinks more than twice its size. In light of that, the Spire 5P53B3 currently installed in Project Numenor will be replaced by the 6149. However in the severely confined space of a 1U system, the efficiency of a HS/F unit also depends critically on the overall ventilation of the case. While the Arkua 6149 is a rather good unit, sticking it in a badly ventilated 1U case will not alleviate thermal problems much. Thus it is important to test run the server before putting it into its production environment, something you'll definitely want to keep firmly in mind. Highly Recommended!