X-bit labs

Water-Cooling Systems Roundup: 16 Systems Reviewed! (page 2)

Category: Coolers

by Kirill "ALT-F13" Balalin

[ 03/30/2006 | 08:48 AM ]

Testbed and Methods

A special testbed is used to check the performance level of a PC cooling system. Most often it is just a regular top-end computer. Such a testbed yields rather stable and repeatable measurements. Yet another advantage is that real-life components with their real-life dimensions and heat dissipation are in use instead of some simulations. Such testbeds do have drawbacks, though:

• The testbed is expensive; the repeated installation and reinstallation of the components shortens the service life of the mainboard, a complicated microelectronics device it is. The tester is required to be very careful to avoid foreign elements, water, metal parts, etc. getting to the mainboard accidentally.
• A computer-based testbed is not suited for precise test measurements since the real heat dissipation of the processor is known only hypothetically and depends on the operating mode.
• A computer-based testbed is not suited for extreme test modes or for special experiments when there is a much higher risk of failure of the components.
• The maximum power consumption of a computer is limited by the power envelope of the employed processor whereas a specialized testbed, on the contrary, allows modeling a situation when a processor with a higher thermal power is being cooled.
• The computer stops working when the OS crashes or there are some accidental and unpredictable failures in the hardware.

Considering all these factors, we decided to develop an all-purpose specialized testbed for testing air- as well as water-based cooling solutions for the PC, but we met some problems in our early experiments with CPU emulators.

At first we built a testbed with a power transformer and a wire-wound resistor in a metal casing. The resistor proved to be the weak link as it burned down due to high density of the generated heat energy, but we couldn’t use a more powerful resistor since it would have been much bigger than a normal processor.

We didn’t try to make a heater element on nichrome or Peltier elements due to certain problems with such devices and their operation.

The fundamental solution was to use a high-power semiconductor device as a heater. We didn’t come to this solution by chance, but based our choice on the CPU analogy. The CPU die and the die of a high-power transistor have the same basic manufacturing technology. The high-power transistor, like the CPU, can have a high heat dissipation density at small dimensions and weight.

The first and cheap emulator of CPU heat dissipation was made by us out of a bipolar transistor from Toshiba, part number 2SC-5200. It is meant for low-frequency equipment like power amplifiers, output and driver stages of high-power drives.

We’ve learned the parameters of the transistor well enough, and our long tests in linear mode proved its reliability at peaks of critical power dissipation.

The transistor is not the only option here. Various chips, amplifiers or transistor packs, special semiconductors capable of linear operation at high power loads can be employed. When choosing the semiconductor for our operating mode, the main factors are small dimensions, high operating power and power dissipation, and easy implementation of control.

The first version of the stand was JudgeMARK-300.

The testbed is based on a mainboard PCB, which necessary components such as the power section and the measuring instruments were all laid out on.

The tested sample was installed right on the case of a 2SC 5200 transistor which played the role of a heater. The sensor was put right into the transistor’s case.