Intel Announces 65nm Breakthrough

[seer]

Intel claims to have achieved a significant milestone in developing next-generation chip manufacturing technology.

In what is proving to be a self-fullfilling prophecy, Intel has managed to shrink transistors enough so that chips can have more of them. It was Intel's founder, Gordon Moore, who, in 1965, claimed that the number of transistors on a chip would roughly double every two years. This prediction has come to be known as Moore's law and this latest achievement, Intel claims, confirms its founders prediction.

More here

I suppose IBM is looking towards doing the same.

[mikeymike]

Considering they haven't properly 'broken through' with 90nm technology*, I'm skeptical.

* - leakage problems, much more wasted energy than with previous processors.

[Glaucus]

Ummm....  what are they at now???

  - Mike

[chsedge]

leakage is a problem concerning the mosfet transistors. You can reduce feature size even more because there's a physical limit of materials involved. After that the mosfets aren't under control anymore...

[KennyR]

They've just started to investigate diamond and silicon carbide transistors. Maybe Intel and IBM should invest in that direction. A CPU core built from these would suffer much less leakage.

[minator]

They'll all be working on this and probably have been for many years, if you read the right stuff you'll see the odd comment about plans for 22nm which is 7+ years away.

I can remember reading about IBM experimenting with 65nm while Commodore were still around, but only at the lab level.

Freescale (and partners) announcing a while back they'll be going into pilot production in under a year.

Considering they haven't properly 'broken through' with 90nm technology*, I'm skeptical.

Intel will be using SOI for 65nm so their leakage problems should reduce considerably.

[Waccoon]

Funny, after decades of research some things are still theories, like the Theory of Evolution, the Big Bang Theory, and the Theory of Relativity...

...but in the fast paced world of computers, we have Moore's Law.

Sorry, that just always ticks me off.

[bloodline]

Well "Moor's Rule of Thumb", would be more accurate... but less of a soundbite, in an industry built on soundbites.

[mikeymike]

Ummm.... what are they at now???

They're at 90nm at the moment, but the Prescott core is a total bar heater (~130W wasted energy when the CPU is being run hard?).  IBM have also been trying to get a 90nm Power5 core out but have been suffering from similar leakage problems.  I haven't heard much about AMD's attempts.

I think the announcement is stockholder candy.

[itix]

Interesting Amiga news I must say ;-)

But seriously it is unfortunate Moore's law is not true in the PPC world.

[Hammer]

AMD uses Low-K Black Diamond in its 90nm process. All of the current 90nm AMD64s (Q3 2004) are being shipped to the mobile market (one of the premium processor markets).

[Waccoon]

But seriously it is unfortunate Moore's law is not true in the PPC world.

Or in the GPU world.  For quite a while GPUs were trippling in performance every 18 months.  Of course, competition is much tougher in that industry.

[billt]

>Funny, after decades of research some things are
>still theories, like the Theory of Evolution, the
>Big Bang Theory, and the Theory of Relativity...

They've come a long way since they were mere hypotheses...

They've been working on this for a few years at least. Same for smaller technologies, already being worked on, and thre was a recent story on slashdot about 35nm... Lab experiments have created transistors using only 5 atoms also, I read about that some months ago. That's when Moore's Law will truely hit a wall, when there's no more atoms to leave out and have the thing still work. We'll need a profound paradigm shift of some sort at that point to improve performance again.

I'm looking forward to what diamond substrate can make possible. Saw a documentary a while ago about artificial pure diamonds that can be grown to pretty much any size, which is needed for chip wafer production. You can't sensibly make chips using those little natural jewelry sized things... Of course there's always waiting for that diamond production to pick up, they're nearly an underground thing due to fears of retaliation from Debeers and friends, who are of course extremely not happy about the idea of truely pure and easy to make "artificial" diamonds. Diamond substrate will allow far far far better heat tolerance and dissipation both, to allow these extremely small transistors to make really complex chips which will get very very hot.

It's pretty cool to hear about though, I'm still designing at 180nm for the cips I work on. The company is moving to 130nm, but it'll be a while before we're pumping the things out of the fab.

[The_Power_of_the_Ginger]

Reminds me of the story I heard back in 2000 that by 2090 computer chips will have to generate small nuclear explosions to get the necessary procssing power required for Moore's Law (by then, I'm sure, history will have changes so that it was Sir Roger Moore who made the original statement  :-D ).

Question is, why bother? Soon you will have all the power you need, like a car doesn't need unlimited horsepower, and you only need a few hundred nuclear bombs to destroy a planet, so why make more?

You'd have to have some INCREDIBLY inefficient programs to require processing power by then.

Actually, Windows 2090 (Oh God, I hope not) would be just the application. It would still run slowly and crash...

[Valan]

Who will buy the next round of chips?

Even the 3d video industry will be satisfied soon with the progress of software efficiency bringing down rendering times.

Hardware is at the stage where most of the market is already satisfied.