Apples and Oranges

Intel recently made some additional claims about spectacular reductions in power leakage.

Or did they?

Here’s what they actually said:

“Test chips made on Intel’s ultra-low power 65nm process technology have shown transistor leakage reduction roughly 1000 times from our standard process (our emphasis). This translates into significant power savings for people who will use devices based on this technology.”

What Intel is doing here is comparing apples and oranges. They are comparing leakage from their ultra low voltage processors to Prescott just to come up with a more impressive number.

Nor do they indicate the relative speeds of the chips. If you aren’t running a chip quickly, it’s not going to leak as much compared to one that is.

It should be noted that an ultra low voltage Celeron M is being produced today which has a TDP of just 5 watts. That certainly can’t be leaking much power. Let’s say for argument’s sake that it leaks 10%, or a half-a-watt (regular PIIIs didn’t leak much more than that).

Let’s be nice and assume a Prescott leaks about half its power. That’s 50-60 watts.

What the difference between a half-watt and 50-60 watts? That’s a factor of over 100X, now (and this is probably an underestimate).

So if you compare apples to apples, the improvement is as most a factor of 10 (and probably less).

More importantly, how much actual power is being saved? Let’s assume we’re right about the half-watt. If power leakage is being reduced by a factor of 10, this means .45 watts are being saved. In this case, that would mean a 9% reduction in CPU power (and much less than that in overall battery life since there are plenty of other powered items in a notebook besides a CPU).

So the thousand-fold reduction in leakage is really only about a ten-fold reduction, which will probably increase battery life a couple percent.

Not so earthshaking, is it?

Of course, what is being learned in reducing leakage for tomorrow’s ultra-low-voltage processors will no doubt be passed on a year or two later to the chips that really need leakage reduction: the desktop/server chips.

Believe me, Intel will be very happy if they can reduce leakage by a factor of 10 for those processors, because that will not mean a .45 wattage decrease, but more than 45 watts. That’s a big, real-world difference.

The Moral of the Story

What these future notebook processors will or will not leak will probably be of little to no interest to the typical person reading this.

What’s important to note here is that Intel has been turning up the hype/BS level on its announcement, which means that they need to read with an increasingly skeptical eye, like a lawyer looking for loopholes.

Ed