Digitimes says that the next generation Intel processor which is supposed to put out the flames will come out somewhere between the middle of 2005 and the middle of 2006.
Since Intel is going to have to retool a new fab to make these things at 65nm as they wish, a full scale release will probably not occur until towards the end of that time window.
What does this mean?
To be sure, Intel will tinker with PressHot to tone down the heat requirements. They’ll probably futz around more fundamentally with Tejas.
But jumping to Jonah is far more fundamental a change than any of that.
For the next 18-24 months, Intel strategy is essentially damage control. They’ll tinker with PressHot even to get 4-4.2GHz out of it. Maybe they’ll play with the electrochemistry on Tejas for another 20%, maybe they won’t, maybe they can’t.
Intel may well be able to not get too badly hurt being walking wounded for two years. However, in the overclocking world, wounded might as well be dead.
Unless Intel can pull out some real magic tricks, they’re going to be pretty much out of action in the overclocking arena for a year, maybe two. Sure, a few will overclock them, hell, a few are doing it now.
But for most overclocking people, and especially those who rely on air to tame the silicon dragon, AMD becomes the choice by default.
Haven’t we been through this before?
Haven’t We Been Through This Before
Within the last four years, Intel has done the following:
- Pushed the PIII until it broke at 1.13GHz
- Sat on the sidelines until they rushed the PIV into action a process generation too soon and gave us Willamette.
- Put out Northwood, which on the surface looked good, provided you ignored this:
(Both graphs from The Economist, 3/13/03, subscription may be necessary)
- Eventually, push Prescott close to the point where they have to bring on advanced life-support systems (i.e., socket 775, BTX) just to keep it from breaking.
This is not a terribly admirable record.
Mind you, they’ve been doing this since about 2001 while at the same time talking along the lines illustrated below:
In its past and current CPU decisions Intel has basically said, “Heat of nuclear reactor OK, heat of rocket nozzle maybe a little too much.”
Not that AMD’s track record is any better. It took years to delay to get Hammer and SOI down (if, indeed, they really do even at this point), but all that’s probably done is move AMD from a generation ahead of Intel in heat to a generation or maybe two behind.
But are we being unfair here? When everyone is flopping, maybe there’s more to it than mistakes and errors?
Who To Blame?
It’s easy to say about Intel, “They should have known better,” but speed is a powerful drug indeed, and one we’re all hooked on.
You can say that all the CPU and GPU makers looked at the power curves, said, “To hell with it,” and that’s just where their products are ending up. Some may get there sooner rather than later, but all are on the same road.
The average person reading this may say, “They’ll figure it out,” but the point to this article is to point out that they’re not figuring it out, at least not easily or in the timeframes we’ve been accustomed to.
And in an age where more and more people say, “They’re fast enough,” we have a scissors effect of more effort, less demand.
At the least, we’re looking at a world where progress as measured by speed will become much more difficult and time-consuming.
Speed will slow down.
Jumping to Jonah is an admission by Intel that it can’t do business as usual any more; that they can’t pursue speed at any cost anymore.
And that’s bad news for speed junkies.
90nm Hammers will give a long-delayed thrill. Then we’ll have dual-core CPUs, but after that, the well doesn’t look too wet.