Strained Silicon 2: Back on Track?

Intel says that they’ll be using a new generation of strained silicon to make its new-generation 65nm chips.

Is this the fix to heat problems we’ve been waiting for? Does this put Intel back on track?

Not really.

Intel says the following about this:

“The second generation of Intel strained silicon increases transistor performance by 10 to 15 percent without increasing leakage. Conversely, these transistors can cut leakage by four times at constant performance compared to 90nm transistors. As a result, the transistors on Intel’s 65nm process have improved performance without significant increase in leakage.”

What does that mean in real-life terms?

Northwoods leak about 40% of their power. Prescotts leak more than that, maybe 50% or even a bit more than that.

If we take Intel at its word, this development will drop power leakage at current speeds down to 10-15% of total power expended, or about the level of pre-PIV generation chips.

This would mean that instead of having a 110W Prescott, you’d have a 70W one. That’s certainly good news, provided you’re satisfied with Prescott-like speed.

But, assuming you’re an overclocker, that’s not likely to be the case.

Go back to the statement, and see the part that says, “increases transistor performance by 10 to 15 percent without increasing leakage. (compared to 90nm transistors)” What that would look to mean is that if you boost CPU speed 10-15 percent, you’re back in current Prescott leakage territory again.

And mind you, these less-than-welcome numbers include any benefits from a process shrinkage, which in the old days would cut power by about half running at the same speed and allow for far more than 10-15% performance improvement.

This is a patch to get Intel through one more generation of CPUs (and/or have tolerable dual-core systems). It’s not a fix.

Is there a more permanent fix in the making?

For the moment, Intel thinks that tri-gate transistors are the way to go (put simply, a tri-gate is a 3D rather than a 2D transistor), and currently plans to incorporate that technology into chips made in 2007. AMD is also looking into the technology.

The Future?

The issue is not “Can this problem be eventually fixed?” Of course it can be fixed, there’s plenty of proposed fixes in the research labs of the CPU companies being worked on as we speak.

In the short term, though, there are no quick fixes which can be easily implemented. Any potential fixes are pretty much going back to the drawing board and designing a new CPU from scratch type of activity.

In the medium term, the question is likely to turn into “Is this problem worth fixing?”

If you think that answer is blindingly obvious, you haven’t thought enough about it. Let me put it to you this way, if two years from now, we find out a “fixed” generation of processors will cost a minimum $1,000, and that price isn’t going to get any lower, are you going to buy one? More importantly, do you think the armies of Joe Sixpacks will?

The PC industry has never been about technology. It has been about affordable technology. If the latest and greatest suddenly takes a big leap upward in price simply because it is hellaciously difficult and expensive to make, the vast majority of computer users will say, “No thanks.”

Of course some people will buy such machines, just as people bought $6,000 PCs twenty years ago. But if that ends up being the price tag, don’t expect many buyers.

If the choice is between a $6,000 machine two or three times faster than today’s, or a $200 equivalent of today’s machine, guess what almost every computer user will buy.

Not saying this will happen, that will be as foolish as saying that it can’t. I’m just saying it could; nobody really knows yet.


Be the first to comment

Leave a Reply