Slashdot Mirror
Double-Gate Transistors
democritus writes "Scientific
American is reporting that IBM has made a
new double-gate transistor" As transistors become
shorter, current bleeds through them when they are supposed
to be off. The new design should enable anything down to 25 nm
long transistors. Currently transistors are 250 nm long, but
to attain higher speeds, circuit designers often use fatter
transistors which can drive a wire to a high or low state (binary 1 and 0)
faster (hence very hot CPUs). The new design acts as if it
doubled the length of the transistor, enabling higher speed
circuits in the same area. Finally, because each transistor
has two gates, a chip using the new transitor could use either
both gates (for fast operation) or a single gate (to save power
for portable applications).
And then there's SOI (silicon-on-insulator)...
I wonder, though, since they're beginning to show that they can do chip fab'ing pretty well, why (as I recall) they Pentithangs (Blue Lightning?) bit.
Or perhaps I misremember.
Like, wow.. I think..
wired? march 99? its not march yet?
They ship them to subscribers a month in advance. It's got Jeff Bezos (the amazon.com guy) on the cover.
Because PPC, which is made by Motorola and IBM and, at one point, Exponential, already has areas in which it is faster than x86, and already is significantly more energy-efficient than x86. :)
Therefore, although Pentiums need these speed and efficiency improvements far more, they can't have them because IBM makes PPCs (and lots of other things besides), and therefore we want to 'replace' x86 by having really solid competition come up and basically obliterate x86 in performance.
The fact that this will give people three main options- MacOS, LinuxPPC and Be- is icing on the cake. Support for x86 == support for Intel, and Microsoft. Support for PPC is far more likely to drive Linux, simply because not everybody wants to use MacOS
Screw ATX mobos, it's time for a PPC based Netwinder >;) that sets out from the start to tap into the cost savings of aiming for Linux from the beginning. _That_ would be the future.
Apple's kinda cool- they oscillate between sucking horribly and being totally cool. Did you hear that now Sys7.5 is available on the web for download at no cost, same with sys6 and sys7.1? That you can download the rather unixish MPW development environment, with the same compilers as used to compile MacOS itself, also at no cost? That Apple is supporting OpenGL and aiming to make _that_ available to developers and users at no cost also? (note avoidance of 'free'! ;) ) ;) :) yay!) have explicitly said they will not blindly copy the bits of VB that suck, even if they are begged to by suits and VB weenies ;) ;) it dualboots with a terrific little control panel app that's super convenient (or open firmware if you insist on not dualbooting MacOS), and the hardware is more predictable so it's easier to set up and maintain. I can't say enough good things about LinuxPPC- I just love it, and my first experiences with Linux have been linuxish, sometimes frustrating (unixgeekery novice here), but in the end rewarding- it did not defeat me, in fact I got to where I understood and loved it. ;) and hey, Apple's not so bad either! Dualboot to it for games, or when you want to just passively webbrowse. MacOS Navigator 4.06 is dead stable...
Any half competent Linux hacker can make a version of MacOS 8, 8.1, 8.5 sit up and do tricks. You only have to do a bit of research, and make sure MS doesn't have a place on your box for anything. There's an active software ecology with lots of developers (hell, even game developers!) getting by writing Mac software. You don't hear about them any more than you hear about Linux developers in the Doze mags, but they are out there, lots of them.
There are even GPLed Mac software programs- I know this for a fact 'cos _I_ write 'em
There's a rapid IDE Basic out there just in the last two years or so called REALbasic, written by an Australian who's moved to Texas. REALbasic is absolutely phenomenal, and though it's vaguely syntax compatible with VB, the vendors (six people in Texas! Not counting the coder's newborn baby daughter
And, of course, there's LinuxPPC. I think this may be the most approachable path to Linux you can get short of buying a prebuilt x86 box with Linux on
Isn't that what all this is about, in the end?
SUPPORT PPC AND LINUXPPC
Posted by JosephKer:
Up to is entirely correct. 25nm is the limit; therefore, we can move up to the implied limit.
Posted by Bill, the Galactic Hero:
Although I'm not sure -- Won't it be a little harder to create charge-inversion layers at the oxide-semiconductor interfaces, since there's no back-bias on channel? I mean, normally the body contact is used to pin the voltage on the back of the channel, which reduces the body effect.
Now your body effect comes back, it seems. So I wonder if the transconductance per channel isn't quite what you'd get with a normal MOSFET. That would mean that a two-channel MOSFET wouldn't match a normal normal MOSFET with a double-width gate.
"Free" , in the case of cell phones, is meaningless. Dealers pay quite a bit of money for them. They make it back in spades by making you sign a contract for service. They get a part of that service fee for as long as the contract lasts, and possibly as long as you use the phone.
Cell phones are, in no sense, free.
Why? Can you refute any of his points?
It's far easier to forgive your enemy after you get even with him.
I wouldn't put too much faith in the Motorola mobos. They are way too expensive. $2400 for a 300MHz 604e motherboard and $3800 for a dual 333MHz is pretty bad. We need someone to develop a cheap motherboard based on the powerPC.
Ways to make it cheaper:
Use IDE instead of SCSI.
Standard DIMMS (SDRAM)
Stick to PCI (no ISA slots or bridge chips)
Use standard AGP slot
ship with a standard BGA ZIF socket so you can buy a G3 or G4
Use USB, parrallel, firewire
It's far easier to forgive your enemy after you get even with him.
IBM publishes an amazing amount of research results, not all of which even get patented, so they effectively make a lot of things available free.
If it allows anything up to 25nm, does that mean it allows 20nm, 2nm, 0.0002nm, etc.?
--
Matthew Sachs
matthewg@interport.net
Considering Intel invented the damn chip, yeah. It's an example of PHB's at Intel more concerned with short-term profits. IBM is a collossus -- they still dwarf Microsoft, and at IBM, lifetime employment for their researchers at least is still a reality.
Hell, this is the same company that's been doing research in *teleportation*. Fast propogation time? How's zero grab ya?
I've finally had it: until slashdot gets article moderation, I am not coming back.
Alot of the stuff that gets anounced these days contains what the SEC likes to call "forward looking statements". Hehehe. Also known as: "Me and da boys got it to work in the back lab, when the planets aligned, while standing on one foot."
In other words, the stuff is still in BIG TIME development. In case you didn't read the story, they note that while they might have a possible manufacturing process, there are innumerable checks to make, things to do, etc.
Businesses don't just put out stuff because it's cool, or neat. They put it out when the economics of production make sense for the market. Since this technology will undoubtably require major reworking at the Fab level (and we all know they run $10-20 billion to construct these days), this isn't something that's going to happen overnight. Like the story says, maybe by 2010.
So stop whining about conspiracies. /. sometimes sounds like a bunch of X-Files groupies stuck in Roswell, NM.
There are always four sides to every story: your side, their side, the truth, and what really happened.
Wow it is nice to see someone here who know a bit about electronics (I am an EE myself) and understands technology...
.. isn't that when the e- going from s to d has no barrier and it becomes like a wire with a small resistance...
I imagine that next year the P II's will be dropping in price and I'll finally go and get a DUAL P II (or maybe a quad?) at about the price that I bought my P 233MMX's this year. I also imagine that with smaller transistors well have PII 5600+ Mhz CPU's....
punchthru
cool tech
Only 'flamers' flame!
Some things IBM(among others) have researched that you are currently using...
The ultra dense disc drives currently sold and used. OF course it isn't the absolute densest, but no one would pay for that until the data becomes overwhelming! With their magneto-resistive technologies, IBM changed the hard disc industry.
IBM also pioneered the phase variation and modulation of multiple optical data layers. They had a seven layer CD stack operational; today you will see this in your average DVD drive and disc. IBM also found a way to use copper substrate in their CPUs, and you will see these in newer Apples, as well as in their Power3 and Netfinity servers.
Things invented we use that perhaps IBM didn't do(THey also might have, but I don't know). Sony's(?) ultra high power diode lasers, found in newer DVD drives, rather than conventional IR lasers. Heck, diode lasers period, which are found in CD players. Whoever originally thought lasers were geeky toys probably are regretting it now. IBM has now discovered double gate transistors, for ultra small gates, x-ray lithography, for extra small layout, gallium arsenide transistors, for the ultra fast transistor, and copper substrate, for the ultra efficient wire. Put all 4 together, and expect to see, in a few years, some of the fastest CPUs in the world. Moore's Law is alive, dammit!
If you don't see change fast enough, its because you, and others, including me, don't pay enough in early adoption fees to get this new Technology. That means to see better LCDs, you should go out now and buy a large SGI flat panel. If you want to see better faster processors, go and dump money for a Quad Dec Alpha system; show them there is demand. If you want better faster video, go and buy a VooDoo3 3000 when they get out, or the Riva TNT2, and let them know affordable consumer level devices aren't enough; you want bleeding edge!
AS
-AS
*Pikachu*
I agree totaly with your assesment of the situation. I was ranting mainly because this guy was griping about the slow state of technology...
Miracles were happening under his nose that he didn't realize. I forgot to mention that IBM had a hand in the UPC, the bar code affixed to everything today as well...
Copper is not ultra efficient, you're right. An exaggeration on my part, but it is more efficient and robust than the current aluminum wires =)
I'm wondering if the clever electron lithography technique involves constructive and destructive interference of 'electron waves', or is something else entirely.
I liked your analysis =)
AS
-AS
*Pikachu*
The increased speed of the double gate mosfet is
due to it's behaving as if it's twice as WIDE (doubling the current driving capability), not twice as LONG (which would decrease transconductance for a given gate width). Longer devices are not as prone to drain-source punchthrough, which is a mechanism by which transistors can be stuck "on", so I suppose it is in this sense that the device exhibits the behavior of a longer transistor.
In browsing the comments to this posting I was surprised at the general lack of knowledge concerning the incorporation of research results into products. Where does everyone think the price cuts across almost all consumer electronic product categories come from? Have you opened up a CD player from 1988 and compared it to one from 1998? Sure, they sound the same, but that's the point: a very large amount of research is required to pack the same features and performance into products that sell for less and less. The complexity of a cell phone far exceeds that of many products that do something which seems more "cutting edge" (like the latest video accelerator card), yet many service providers give them away free.
I used the term "punchthrough" to refer to the merging of the drain and source depletion regions, so you're correct...the transistor essentially looks like a resistor (plus all the parasitic capacitances and source/drain diodes)
What will happen to the rest of the Industry. Is IBM going to share its technology with other companies like Intel, Sun, and SGI. I hope they do becuse it would be very good for the computer industry overall.