We might all be plenty happy to ogle over the next gigahertz revolution, but CNet recently pointed out that rating processor performance alone is becoming a useless measure. We can make the processors fast, the quesiton is, can we do it without burning through the mother board (and the electric bill).
Unfortunately, the MPG of a CPU (theoretically measured in performance per watt) is not a simple thing to calculate. The measure of hertz, meaning processes per second, is now pretty much ignored. There's no perfect way to measure the performance of a CPU, and there is really no way at all to accurately measure performance per watt.
{mosimage}Sure, you could hold a thermometer to your CPU and run a billion calculations and divide one by the other, but AMD and Intel are going to start advertising these numbers, so they're going to have to figure out some industry standards.
Unfortunately the performance per watt of a processor tends to vary significantly as the processor is used more or less. Intel is measuring it's processors at full load, something very rarely seen by a processor, with encouraging results. AMD, on the other hand, says a more natural system for measuring performance should be tried.
Already the battle has begun, Intel claiming that its new server processor, Woodcrest, has overtaken AMD in terms of performance per watt, but AMD disagrees.
All of this, if you can believe it, is a big deal for IT industry power consumption. Fifteen percent of money spent in the operation of data centers goes to power and cooling down all of the hotness the processors create. And the fight over performance per watt is a good one for the industry, the customers and the environment.
Let's just hope they can agree on how to measure it.
Note: Since this article got Dugg there's been a bit of discussion about the lack of real, hearty, technical information here. Let me just say that I wrote this mostly as a redirect to the CNet article, which does have a lot of great information on it, and that I didn't expect this to get on Digg. It was nice surprise though. Thanks to everyone who's visiting EcoGeek for the first time, and check out the CNet article if you want more details.

written by a guest, June 13, 2006
written by a guest, June 18, 2006
written by a guest, June 18, 2006
written by a guest, June 18, 2006
written by a guest, June 18, 2006
Interesting article opener, but I wish it would have been more than that.
written by a guest, June 18, 2006
hotness? I think the word you're looking for is "heat"
written by a guest, June 18, 2006
written by a guest, June 18, 2006
written by a guest, June 18, 2006
written by a guest, June 18, 2006
written by a guest, June 19, 2006
I could be wrong, though.
written by a guest, June 20, 2006
written by a guest, June 21, 2006
To the "AMD sucks" guy: take a look at the data out there for the current CPU's and guess what you'll find. That's right, AMD CURRENTLY produces less heat. Again, if you want to go off of speculation or personal pref, thats fine, but I like to go off of hard data and statistics. I am a logical sort of guy, and if someone tells me something that they cannot back up, like what your saying, then I will tend to not trust what they say.
I am sorry you Inteliheads can't seem to understand the importance of data and real world statistics, but that is your fallacy not mine.
| < Prev | Next > |
|---|
JUN 11
"Ok to the first guy that commented on my first entry: I mentioned www...."
View all Comments