Quote of the day … from The Inquirer

Ok, this was perfect for a sleepy wednesday … I haven’t said much about Barcelona. Yeah, it released. Yeah, I want to play with some. No, I don’t have any (big hint there AMD….)
The Inquirer has an article on power, and some of the press (mis)interactions currently going on with AMD. And a little commentary. This gem was in the commentary …

What do you do with the old chip? Whatever you like, Alberto Macchi helpfully suggests. He doesn’t suggest wiring up a small African village with it and running a few virtual machines off it. Of course, it’ll be useless without the board you just pulled it out of, so you’ll probably chuck it in the bin, to end up as landfill. Very green. Very 50/15.
It’s when these technology firms bang on about how green they are that the whole fa??ade falls over. They don’t get it at all. The whole issue of squeezing extra performance per watt is perhaps a useful exercise, but they don’t do it to be green, they do it because their customers sitting in their data centres are toasting their nuts off. Their energy bills are going through the roof and not just because their kit draws so much current. It’s also because they have to turn up the air con to fight the heat their chips are giving off.

Yeah, well, thats about the size of it. Carbon footprint? Nah. “Chestnuts” roasting over an open server? Yeah….
Heat costs money, to generate and remove. If it (the heat) could be stored somewhere, we might be able to get some more use out of it later, limited by inefficiencies of the pumping, storage, and retrieval process). The problem is that heat pumps also consume energy, and are not 100% efficient. That darned thermodynamics … just can’t seem to get away from it .

1 thought on “Quote of the day … from The Inquirer”

  1. I’d have thought you’d be the first to get a new AMD chip, what with your white paper for them on Accelerated Computing.
    We do have to fundamentally revise performance metrics for the reality that heat is more critical than speed. For all the interesting parallelizable problems, speed is just a way to make up for what we lack in computational surface area. As a first order approximation, as we increase the frequency of our computers we increase the power they consume cubically. So my answer to the data center energy problems is this: buy more chips and run them slower. A data center with 1,000 times more chips can perform a parallel operation at the same operational throughput at 1/1000 frequency for 1 millionth of the power cost
    Since it is likely that the exponentially decreasing the cost per transistor will continue long beyond the exponentially increasing transistor density, the cost of energy will likely dominate the fixed cost of a chip so it will be good economics to just use a lot more of them. Of course to a layman: “to decrease the power cost of computing you need to buy more computers” sounds ludicrous.

Comments are closed.