NEC and Toshiba hop on IBM's Semiconductor Alliance train for the ride to 28nm
By Tim Stevens 
posted June 18th 2009 7:39AM
posted June 18th 2009 7:39AM
IBM seems seriously intent to beat Intel to the tiny, 28nm processor punch, and has enlisted even more help to get there first. After securing deals with Samsung, Globalfoundries, and a few other merry chipmakers in April, NEC and Toshiba are now joining in on the Semiconductor Alliance fun to create next-generation processors before the biggest name in current-generation processors. Goals are smaller footprints, lower power consumption, and of course greater performance. Mind you, that greater performance is still likely two years away from anything we can hope to buy.
It's a start
With 32nm coming this year, one would think the Intel should surely be able to produce 28nm (or small) processes within 2 years.
Yeah, I expect Intel will ship 22nm procs within months(+ or -) of these 28nm chips.
One upside, hopefully the next Xbox and PSX will launch with 28nm CPUs.
Change to optical already.
Probably have to spend billions on foundries to do that.
optical what?
Current electronics use voltages to denote 0s and 1s, 0volt for 0, 5 volts for 1 (the voltage is lower in the cpus we use, but that's the basic theory that's applied). The circuits calculate everything using tiny pulses of electricity to perform logic operations and give you an output.
The problem with this is electricity generates heat. The faster you calculate (the higher the frequency), the hotter it gets, so there's an upper limit to the frequency you can achieve. This is why recently they've been concentrating on making processors multicore and perform better instead of simply increasing the frequency.
The thing is, we don't care what is doing the calculations, we just want the output. If we were to use pulses of LIGHT (on = 1, off = 0) instead of electricity to perform calculations, processors could get much faster and much cooler. The final output would still be electric though, but that's just fine.
@superhobo
Ok genius, let's see your design plans for making a processor core using "optical" logic gates...
@loosely
"All-optical XOR gate using semiconductor optical amplifiers without additional input beam
Jae Hun Kim; Young Min Jhon; Young Tae Byun; Seok Lee; Deok Ha Woo; Sun Ho Kim
Photonics Technology Letters, IEEE
Volume 14, Issue 10, Oct 2002 Page(s): 1436 - 1438"
People have been working on optical circuits for years.
You being ignorant doesn't make other people stupid.
Yeah, what jon said.
and keep in mind, that while process is a big determining factor, it isn't everything. One has to wonder if Intel might still be able to produce a better platform at 32nm than the semiconductor alliance could at 28nm, especially if they get bogged down in unforeseen complications. Hopefully it puts the pressure on to really get some powerful mobile devices out into the wild.
I thought intel's next step after 32nm was to 22nm?
and yes, i realize that this topic is about IBM. however, it's a little silly to say you beat somebody somewhere when the other party isn't even going to the same place. that's like saying "hey, bill, I beat you to chicago" when bill wasn't going to chicago.
yea, intel doesn't do half steps in this area.