 |
Ern Worthman,
Editorial Director |
Moore's Law, the infamous knell of death for the unprepared, which states microprocessor complexity will grow exponentially without an increase in price, has held for four decades. But I wonder if Gordon Moore saw the issues about power when he coined the phrase.
In the mPC business, and not just computers chips, the component density of the latest chips, and the power they consume, causes them run temperatures of over 100° C if not properly cooled — literally at self-emolating temperatures for most semiconductor platforms.
We all are pretty much aware of the fact that most of the energy consumed by modern ICs ends up as heat. That’s why there is a frantic movement within the industry to control power, thereby its by-product, heat. This is being developed along a number of parallel paths — micro-managing power consumption, developing power-miser/power-intelligent devices, developing intelligent power networks and so on.
In fact, in the mPC biz for example, the issue is becoming critical. Intel killed a recent high-speed-CPU project, in part because it found no practical way to keep the million plus gate chips cool. And, as many of us have learned firsthand if we were ever involved in life testing, heat is the most common method applied to induce failures and erratic operation. Chips operating outside of their temperature window just aren’t reliable.
Well, technology to the rescue again — sort of. One potential solution, commonly associated with nuclear reactors, is liquid metal cooling. Spearheaded by NanoCoolers, a startup in Austin, TX, the technology takes advantage of an unusual compound of metals that remains liquid at room temperature. Currently, their mixture of gallium and indium (and a pinch of tin) flows freely at temperatures above 7° C. And a new formula could go as low as minus 10° C. This technology is one among a number of new and promising cooling solutions that are being evaluated.
Well, I like to think sometimes I’m ahead of the curve. I have to admit that about circa 2000, I was Mr. CPU overclocker. I’d buy the fastest CPUs, crank ‘em up and have a heck of a game of Star Wars, but there were always crashes. After a few third degree burns from the CPU I got hooked on trying to get these puppies running cooler. So, as only a true technogeek would find, I discovered water cooling. Yeah! I bought a few dozen feet of tubing, some clamps, a radiator (yeah, a 6"× 6" radiator) and a fan for the radiator. Then I dug up my old external aquarium filter and ordered the heat transfer unit that fits on the CPU.
I constructed this cooling system it in my office, filled it up and attached it to my computer’s CPU. One of my friends wanted to know if I was trying to build a new human being (very funny).
I fired up the cooling system, checked for leaks and wala! It worked. I cranked up the computer, kept my fingers crossed and when I checked the Mobo’s temperature monitor, sure enough, the CPU was running at a cool 32° C or so (from typically 55° C with fan cooling).
Well here I was thinking I’d finally stabilized the system and was getting ready to take out unlimited barrage of Tie Fighters when I smelled that unmistakable odor of frying semiconductor substrate... a few puffs of smoke rise from the board and it’s over...fried chip city. The cooling system sprung a leak and the rest is history (as was a good chunk of my budget).
Anyway, the moral of the story is that heat has long been the bane of speed, density and power, and there have been countless attempts to control it (some more successful that others).
This attempt, liquid metal is right out of Terminator 3... I love it. But it smacks a bit of simply a "better" mouse trap. It has some strange similarities to the water cooling system (The liquid metal flows in a loop around a PC or graphics card and picks up heat from the top of the heated chip. Then the liquid gets pumped through pipes to a radiator, usually with a fan blowing on it, where the heat is released into the air. Finally, the cooled fluid circulates back to the chip.)
However, this system uses no moving parts, which is really its only notable asset. The pump moves the liquid metal around in the system electromagnetically. Unlike water cooling, it consumes little power, and is silent. In a patent application, NanoCoolers even suggests powering the pump solely from the waste heat produced by the computer. And the liquid metal is non-toxic, but not non-corrosive to some metals, notably aluminum. So there goes that smell again.
Frankly, rather than developing band-aid technologies, I’m of the school that we develop core technologies with heat characteristics in mind. Easier said than done, however and I know that, but we gotta put our development dollars somewhere. We need to get these low-voltage, low-current technologies (< 1 V) to market. We also need to get more optical technologies integrated into chip design.
Power has been the Achilles’ heel of the semiconductor industry for some time now. When I hear of technologies, like this light metal cooling, I think these monies may be better spent at the front end (development) rather than the back end (control).
And while it may be practical for the nuke industry, I’m not convinced it will economically scale to the semi industry.
Wireless Design & Development
Advantage Business Media
Rockaway, NJ, 07866
© 2008 Advantage Business Media
