Last month, I had an unusual opportunity for a trade-press editor. I got to sit in on an all-day session as the key players in the power supply industry took their best guesses at where they were headed between now and 2010. Under normal circumstances, if you’re an editor, you get many briefings on new products, occasional chances for off-the-record backgrounding on issues like standards developments and lawsuits, and some even rarer peeks at R&D work in progress. Sitting-in while OEMs, power-supply makers, chip vendors and industry analysts share their expectations is to participate in the ultimate Last month, I had I had an unusual opportunity for a trade-press editor. I got to sit in on an all-day session as the key players in the power supply industry took their best guesses at where they were headed between now and 2010. Under normal circumstances, if you’re an editor, you get many briefings on new products, occasional chances for off-the-record backgrounding on issues like standards developments and lawsuits, and some even rarer peeks at R&D work in progress. Sitting-in while OEMs, power-supply makers, chip vendors and industry analysts share their expectations is to participate in the ultimate in cosmic convergences.
Of course all these participants are also competitors in one way or another, so it seemed to be also a bit of a poker game. Some cards are dealt face-up, some you have to pay to see, and others are just shuffled back into the next hand.
This all took place just before the annual Applied Power Engineering Conference and Exposition (APEC) at the end of March. The occasion was the Power Supply Manufacturers Association’s irregularly scheduled industry-forecast meeting. This year, for the first time, members of the electronic trade press were invited to participate as scribes. Our notes on the oral presentations will be included in the final report, which will be published in May and sold for $750 a copy. What follows is a high-level overview of the proceedings from one reporter’s perspective.
After introductions, the meeting kicked off with presentations by Cisco, HP, Intel, and Dell. I learned that distributed power for servers and routers are the applications where the power supply and chip suppliers focus most of their attention, followed by silver boxes and ac adapters for desktop and laptop PCs. The two most interesting things I gleaned from these presentations had to do with flattening: the power going into a rack of blades will hit a brick wall at 17 kW, and the next frontier in power-supply design will be raising efficiency specs at low output levels, essentially flattening the efficiency curve.
Here’s the background on that 17-kW figure for server racks: Today, a rack full of sever blades or routers has a peak power consumption of 8 to 12 kW. Soon it will be 15 kW. That’s being consumed by the silicon that does the work; all the front-ends, bricks, and POLs are running in the 90 percents in terms of efficiency. The 17 kW wall is thermal, dictated by the temperature of the cooling air as it exits the rack. As long as the cooling paradigm is forced-air, that’s where it will sit.
There are similar walls for silver boxes and laptop adapters. PC gamers place the most pressure on silver boxes because, while CPU loads are flattening out, graphics cards and memory continue to escalate their power demands. As in the server room, the limiting factor is cooling. With the silver box, however, there is no guarantee that the incoming air will even be air-conditioned. With so many unknowns, there’s no telling exactly where the wall will be, but it will not be too long before gamers hit it and have to do something to get around it.