Posted on 11/26/2001 11:14:23 AM PST by milestogo
By Duncan Martell
SAN FRANCISCO (Reuters) - Intel Corp. (Nasdaq:INTC - news) has devised a new structure for transistors -- the tiny switches that make up semiconductors -- in a development it said could lead to microprocessors that run at blazing speeds and consume far less power than conventional ones.
The technology, Intel said, solves two of the more intractable problems facing the development and manufacture of microprocessors today as more and more transistors are packed onto each chip: power consumption and heat.
In addition, as the geometries on chips become ever smaller, it becomes increasingly difficult to ensure that electrons go where they were engineered to go, a problem that can lead to current leakage within the device.
The advances, which Intel calls the TeraHertz transistor because it cycles on and off 1 trillion times per second, could ultimately lead to new applications, such as real-time voice and face recognition, computing without keyboards and ever-smaller electronic gizmos with higher performance and improved battery life. To compare, it would take a person more than 15,000 years to turn a light switch on and off a trillion times.
``The real significance of this is that they've basically invented a new transistor technology that's fundamentally different and it's manufacturable,'' said analyst Dan Hutchinson of market research firm VLSI Research. ``They've completely re-engineering the transistor as we know it.''
This leakage of electrical current within a transistor, which is simply a solid-state switch that toggles between an ''on'' and an ``off'' position when current is applied, can quickly become problematic. Leakage can lead to a microprocessor generating data errors, producing so much heat that the silicon actually melts, and consuming too much power.
``The general issue of heat is one that everybody is confronting,'' said analyst Nathan Brookwood of market research consultant Insight 64.
KEEPING MOORE'S LAW ALIVE
The technology that Intel is using in the TeraHertz transistor is also important, because it is another step in extending Moore's Law into the next decade. Moore's Law, an observation rather than a law of physics, is named after Intel co-founder Gordon Moore, who noted in 1965 that the number of transistors on a chip doubles every 18 months to 24 months with an attendant 50 percent reduction in cost.
``What's going to limit transistor performance is the power consumption, not the speed and not the size,'' said Gerald Marcyk, director of components research in Intel's technology and manufacturing group. He added that Intel is aiming to have 25 times more transistors in processors than in current ones, running at 10 times the speed, yet with no increase in power.
The Santa Clara, California-based chipmaker said it might start incorporating portions of this new transistor structure into its microprocessors as early as 2005. Intel plans to present two major elements of the new transistor structure at the International Electron Device Meeting, a leading gathering of scientists researching device physics, in Washington on Dec. 3.
A BILLION TRANSISTORS
The type of transistor that most of the semiconductor industry uses is called a CMOS (complementary metal-oxide semiconductor) transistor, a switch that, when ``flipped,'' allows current to pass through, completing a circuit.
The combination of millions of ``on'' transistors, represented in the digital language by the numeral one and ''off'' transistors, represented by zero, flipping back and forth between the states is what allows a microprocessor to crunch and manipulate data.
Intel's flagship Pentium 4 chip has about 42 million transistors. Microprocessors with a billion transistors are expected in the second half of this decade.
Think of a CMOS transistor as a channel of water. The left bank of the transistor is called the source, and has continuous power supplied to it. The right bank is called the drain. For electrons to cross the channel, passing from the source to the drain and to complete the circuit, a bridge is needed.
That bridge is known as the gate. When the gate is charged with electrons, a kind of drawbridge is lowered, and power can flow from the source to the drain, completing the circuit, turning it ``on'' and registering as a 1.
One element of the new transistor structure is what Intel calls a ``depleted substrate transistor,'' a form of a technology called silicon on insulator. Using this technology, a transistor is built in an ultra-thin layer of silicon on top of an embedded layer of insulator.
Because this insulator is thinner than other types now used, all of the electrons get used up, resulting in what is called maximum drive current, which helps the transistor switch on and off faster.
Other chipmakers, such as International Business Machines Corp. (NYSE:IBM - news) are already using what are known as partially depleted substrate transistors, but they were not designed to address what is known as off-state leakage, or the passage of electrons from the source to the drain when no power is supplied to the gate, Hutchinson, the VLSI analyst, said. Intel said this type of transistor results in 100 times lower leakage than traditional silicon-on-insulator implementations.
The second key element, Intel said, is the development of a new material, which it is not disclosing, that replaces silicon dioxide on the wafer, a layer at the bottom of the gate and between the source and the drain.
Super-speedy transistors showcased in the last year have gate dialectrics only about three atomic layers thick, meaning that increasing numbers of electrons are leaking through, and this leakage has become one of the largest sources of chips' power consumption.
Intel said that this new material reduces gate leakage by more than 10,000 times compared to silicon dioxide. The combination of these two developments will result in chips that have higher performance, produce less heat and allow for significantly longer battery life.
``These developments are definitely in the category where it makes sure we stay on Moore's Law,'' analyst Brookwood said. It represents some pretty serious changes in the way people are doing things.''
Okay now, stop playing with the lights!!!
With the right software - this could do some impressive VR!
At these speeds, the software could use brute force and still be pretty cool.
Besides, EE's are a dime a dozen thanks to India and China. MBA's are a lot more valuable, and a lot more rare. Last I saw China and India were not turning out tens of thousands of MBA's every year, and Silicon Valley wasn't expecting to hire MBA's for $30k a year. See what they are paying that MSEE from India and you'll see what I mean. Becoming an engineer nowadays is a sucker deal, you'll work twice as hard as anyone else for half as much money.
I went into patent law. It was the best transition for me. Going the MBA/middle management route has alot of risks.
You're smart by sticking close to marketing -- those people are often clowns and a bit of competence can go a long way.
Does that include bathroom breaks?
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.