Google
 

View Full Version : New Technology for Computer Chips


Manu
06-22-2001, 12:08 PM
Kicking Computer Chips Up a Notch
Chip Research Developments Push the Extent of ‘Moore’s Law’

By Paul Eng

June 22 — You think computer chips are fast? You ain’t seen nothing yet.





For the past four decades, microprocessors have followed “Moore’s Law” faithfully. That observation, made by Intel co-founder Gordon Moore in the 1960s, states that the power of microchips doubles at least every 18 months.
Still, many in the microprocessor industry — including Moore himself — doubted that this rate of growth could continue forever. That's because it was getting extremely difficult to make transistors, the heart of a microprocessor’s power, smaller and more powerful than before.

But nay-sayers shouldn’t write off the chip-making chief just yet.

At a chip developers' conference in Kyoto, Japan, last month, both Intel and IBM announced independent research developments that could extend Moore’s Law well into the next decade, if not beyond.

'Strained Silicon'

Researchers from Armonk, N.Y.-based IBM believe that one of the easiest ways to make faster chips is to make transistors more efficient by fiddling with their basic element: silicon.

The IBM development is based on a decade-old concept called “strained silicon.” The concept shows electricity can move through a transistor much faster if it is made up of stretched, or “strained,” molecules of silicon.

Molecules of any given substance have a shape, size and pattern. To stretch the shape of silicon molecules, researchers at IBM use an underlying layer of silicon and germanium, another element. That combination effectively forces the molecules of the layer on top to try to conform to the shape of the layer below.

That allows the electricity to move through the strained silicon at speeds of up to 70 percent faster than through non-strained silicon.

Jeff Welser, manager of high performance semiconductor technology at IBM, says chips using such strained silicon transistors will be 35 percent faster than chips using similar-sized, non-strained transistors.

'Devil in the Details'

Welser says more work is needed on producing pure underlying, or "substrate," layers of silicon-germanium. And as such, IBM has managed to create only a few hundred of these strained silicon transistors rather than whole processors. “It’s the ‘devil-is-in-the-details’ kind of thing,” he says.

However, “Once you have the substrate,” says Welser, “It can run through a normal production line.” And since IBM’s strained-silicon technology doesn’t require new chip-making machines or processes, Welser says faster chips could be available by the end of 2003.

But by far, the most ambitious of the Moore’s Law-extending developments came from Intel.


Much Smaller Than a Human Hair

The Santa Clara, Calif., chip-making company announced it has created a transistor that is a mere 20 nanometers in size. That’s about 500 times narrower than a strand of human hair, or about 30 percent smaller than the current fastest transistors being researched.

The new Intel design could one day fit up to a billion transistors onto a chip the size of the current Intel Pentium 4 speed demon, which holds 42 million transistors. Such chips will be able to complete almost a billion calculations in 1/50 of a second — literally the time it takes to blink an eye, more than 10 times faster than current PC chips.

“That’s going to lead to very, very, very, powerful computers,” says Rob Willoner, marketing analyst with Intel’s technology and manufacturing group.

Getting to that point, however, still requires overcoming many challenges. The largest: Changing how tiny transistors are crafted.

Industry Shift

To produce such miniscule transistors, Intel and other chip makers must perfect a process called Extreme Ultraviolet, or EUV, lithography. This method uses a complex set of mirrors and photon beams to trace the microscopic patterns of transistors onto silicon wafers.

But EUV lithography is still a very tricky technique and quite a few years away from being adopted industry-wide. Willoner notes, for example, that the company has managed to create only “tens” of 20-nanometer transistors despite having done “tremendous amounts of work” on EUV technology.

But Willoner is confident that the first chips built with 20-nanometer transistors could be available by 2007. “The industry has 40 years of experience,” says Willoner. “There are plenty of hurdles, but we think we can overcome them.”
www.abcnews.com (http://www.abcnews.com)

------------------
Manu Narayan

74Mav
06-22-2001, 03:22 PM
The strained silicon part makes sense. I guess on a molecular level, using less material to make the same transistor would increase its efficency. How about reliability? I'd like to see Intels description of how they made such a small transistor.

I'm begining to feel old. I remember going to Radio Shack and drooling over a TRS-80 Model 1 & Model 3 computer with a z-80 CPU. Had two Tandy Color Computers, helped the local computer shop build my mom & dad's 1st PC, an 8086 at a scorching 10MHz clock spped and a 20MB Hdd. Had an Apple IIe self destruct on me on the last day of jr high. Man how things have changed...

------------------
Uh, what country is this??

74Mav
06-22-2001, 03:28 PM
Another take on Moore's law with a touch of interesting tidbits...

http://www.anandtech.com/news/shownews.html?i=13512&t=an

------------------
Uh, what country is this??

Manu
06-22-2001, 05:39 PM
That was a nice article.

Yeah, id think with the strained silicon you'd run into more heating/shorting problems. But hey thats why they pay these guys the big bucks.

The article mentioned something about mixing it with another element, that problem strengthens it physically and acts as an 'immediate' heat sink.

------------------
Manu Narayan

74Mav
06-25-2001, 08:08 AM
Another blurb on htis subject. Seem IBM may beat the others to the punch...
http://www.electic.com/cgi-bin/news/News/Stories/2001/06/24/993434022.shtmlother blurb on this subject.

------------------
Uh, what country is this??

Manu
06-25-2001, 11:32 AM
hmm, what check the link? I got a 404 error.

------------------
Manu Narayan

74Mav
06-26-2001, 07:53 AM
try this...

http://www.electic.com/cgi-bin/news/News/Stories/2001/06/24/993434022.shtml

------------------
Uh, what country is this??

Manu
06-26-2001, 11:24 AM
Wow.

That is amazing. 210GHz.

I cannot wait for what the future has to hold. It will be amazing.

------------------
Manu Narayan

Google