Incoming air must be filtered, dehumidified, and cooled. One of the most disappointing aspects of the problem is that the heat energy in exhaust air is just "dumped" by conventional HVAC: wasted.
I agree with the comment that the future probably lies in variations on liquid cooling.
A possible benefit of liquid cooling is that heat energy can be more easily collected and re-used, perhaps to drive mechanical components in the cooling system or in some other ways. With some redesign, using expansion of a circulating low-pressure chassis refrigerant could perhaps achieve greater efficiencies. Not only in cooling, either. As every overclocker knows, lowering the temperature at which silicon operates can result in improved processor performance.
Silicon power-management techniques such as clock gating and power gating, now being applied to mobile processors could be adapted. Dynamic allocation of processing (ie., only powering up as much system processing as you need) could help.
Smaller circuits (we're heading into nano territory) will also help.
Solid-state memory is approaching cost-effective usability, and looks like it will offer significant gains in efficiency and speed.
There's no single Killer Solution, but after engineers find metrics for data center efficiency, they'll probably begin to think in terms of closed-system solutions, each designed to be part of an integrated concept.
Quantum computing still lies ahead of us, and will theoretically improve processing efficiencies by several orders of magnitude. But increased processing demand will probably extinguish the gains, over time, and everything I've read so far suggests that such computers will need supercooling, at least at the processing core.
Speaking of which, a primitive form of quantum computer has been operating in Vancouver lately, though some question whether it is, in fact, a "quantum computer". It performs only a limited type of processing, but does not lack for customers.
backbonemag.com
cbc.ca
Jim |
