Moore's Law Has No End in Sight..

[Burpo]

Interesting article..

GaAs is VLSI's inevitable successor to silicon.
Gallium arsenide (GaAs) will soon outpace silicon (Si) as a higher-performance choice for implementing integrated circuits, according to POET Technologies Inc. of Storrs-Mansfield, Conn. The handwriting has been on the wall since the 1980s..
"GaAs, as opposed to silicon, will boost electrical transistor performance while integrating optical circuitry capabilities. These qualities enable both higher performance and novel IC architectures, thereby extending Moore's Law indefinitely."

http://www.eetimes.com/document.asp?doc_id=1323892&

"Thus 40 nanometer GaAs compares to 14 nanometer in speed and 10 nanometer in power."

 

[ElFenix]

they've been talking about gallium arsenide for as long as i can remember. the stuff sunk seymour cray.

though, iirc some phone ICs use gallium arsenide because they need the frequencies.


diamond would be even better.

 

[jdubs03]

it is inevitable that we will finally get to see commercialization with these new channel materials. by 7nm if we see these results we're talking about at least 3-4x faster than today, hence moore's law will continue. 14nm performance/10nm energy properties at 40nm planar? that's awesome, at 14nm these materials would at least double our current performance at 14nm, and that doesn't include the transistor structure, so 2x+, but either way its 3+ years away. good article though.

 

[jhu]

No end in sight? Not looking far enough.

 

[jhu]

Silicon digital logic hits the wall at 4 GHz, but we can produce small gallium arsenide [GaAs] analog circuits switching at 100 GHz today and 400 GHz in the not too distant future

Doesn't appear they're comparing the same things.

 

[Greenman]

I believe you mean "Moore's speculation". It's not a law at all.
Gallium arsenide has been the next big thing for years, I think I first read about it around the time I got my blazing fast 25mhz 386.

 

[Idontcare]

Good stuff, we have an entire periodic table of elements to use in a zillion different combinations for improving electrical properties and performance.

 

[Xpage]

personally I would be happy with with the "lower operating voltage -- as low as 0.3 volts" assuming he meant threshold value. I believe Si is around 0.7v. So if we assume we will be above that by 0.2v like we currently are, then we'd halve the voltage.

Since the energy draw is based on the square of the voltage, that's some nice savings. I'll take a current gen 65 watt processor to have 16 watts of power consumption. Consumers will be happy, chip companies will make tons of money and then never sell another processor since it will be beyond good enough until MS releases Windows 10 and sucks all the cycles as we all know only the odd releases are good

 

[OS]

There is active research in using individual atoms embedded in silicon for either bits or quantum states, if any of the research yields anything useful, moores law might go all the way down to atomic scale.

 

[MarkizSchnitzel]

Outside of super computers, will there be a market for these future super advanced chips? I mean, most people today can get by with a Baytrail. Sure, power savings are nice, but performance not so much.

Also, when all of the major players do reach that point, what happens to competition? How do they differentiate?

 

[moonbogg]

I hope I don't have a heart attack or get hit by a truck before I get to game on one of these things.

 

[witeken]

I believe you mean "Moore's speculation". It's not a law at all.

It's isn't a speculation. History has been behaving exactly as Gordon Moore predicted almost 50 years ago. It is the roadmap for the industry.

Gallium arsenide has been the next big thing for years, I think I first read about it around the time I got my blazing fast 25mhz 386.

Why do people always have to use such nonarguments? Here's something to consider: Tri-Gate has been "the next big thing" (which actually isn't true at all, but it are your words) for 1-2 decades. If you buy an Intel processor, it will have those 3D transistors as promised.

=> It takes time to implement technology. Companies follow a logical roadmap; strained silicon, HKMG and Tri-Gate are the predecessors of III-V. Just like we don't have graphene transistors yet.

 

[witeken]

personally I would be happy with with the "lower operating voltage -- as low as 0.3 volts" assuming he meant threshold value. I believe Si is around 0.7v. So if we assume we will be above that by 0.2v like we currently are, then we'd halve the voltage.

The threshold voltage of silicon is more like 0.3V.

http://www.realworldtech.com/near-threshold-voltage/

 

[witeken]

Outside of super computers, will there be a market for these future super advanced chips? I mean, most people today can get by with a Baytrail. Sure, power savings are nice, but performance not so much.

Those future super advanced chips will be in all your devices.

Also, when all of the major players do reach that point, what happens to competition? How do they differentiate?

Reach what point? We don't know when scaling, performance and power increases will end. Don't forget that pretty much the only company that really differentiates itself is Intel with its 2-4 year lead.

 

[BrightCandle]

It might extend the inevitable but it wont stop the problem that at some point you are counting 1's with single electrons and transistors are only an atom wide. Going below that with conventional computing just isn't possible as far as our current science knowledge goes. In reality I doubt we can get down to that limit either.

Physics is already kicking us every which way with todays processes, the rate of new processes and the doubling of density of transistors it ought to bring is no longer on an 18 month turn around, it has been 24 months and now its looking more like it will be 30 months. Moving to GaAs givers a one time improvement and then puts it onto a similar trajectory of difficulty.

 

[witeken]

It might extend the inevitable but it wont stop the problem that at some point you are counting 1's with single electrons and transistors are only an atom wide. Going below that with conventional computing just isn't possible as far as our current science knowledge goes. In reality I doubt we can get down to that limit either.

Physics is already kicking us every which way with todays processes, the rate of new processes and the doubling of density of transistors it ought to bring is no longer on an 18 month turn around, it has been 24 months and now its looking more like it will be 30 months. Moving to GaAs givers a one time improvement and then puts it onto a similar trajectory of difficulty.

Has it ever been 18 months? Intel's Tick-Tock schedule is 1 new step every year.

In any case, power and performance will probably go up for a while even after density scaling ended. If you need more transistors, you can also use the third dimension.

 

[know of fence]

[disregard]