Slashdot Mirror
Experimental Transistor Breaks 600 Gigahertz
neutron_p writes "The goal of a terahertz transistor for high-speed computing and communications applications could now be within reach. A new type of transistor structure, invented by scientists at the University of Illinois, has broken the 600 gigahertz speed barrier.
A new type of transistor - built from indium phosphide and indium gallium arsenide - is designed with a compositionally graded collector, base and emitter to reduce transit time and improve current density. With their pseudomorphic heterojunction bipolar transistor, the researchers have demonstrated a speed of 604 gigahertz - the fastest transistor operation to date."
If we assumed that all transistors on a chip (say a P4) were this type of transistor, and could run at 600 GHz, I know there is time required for a signal to cross all of these transistors, etc., and that some chips have a billion transistors on them, how fast could the current chips run with these transistors?
Rhymes that keep their secrets will unfold behind the clouds.There upon the rainbow is the answer to a neverending story
Silicon is widely available for current transistors. Are indium phosphide and indium gallium arsenide just as available, or are they the doping materials.
Will material prices be the main determining cost of chips made from these products?
I didn't RTFA -- it was slashdotted.
OK, I cannot RTFM right now as it is /.ed, but:
This sounds an awful lot like they are giving the zero-gain bandwidth of the transistor - the frequency at which the transistor does NOT amplfy a signal anymore.
So, at 599GHz the transistor will amplify a little. At 600 GHz the transistor takes as much power to drive the input as it is able to switch at the output. At 601 GHz the transistor takes more power to control than it can switch.
Given a 600 GHz zero-gain bandwidth transistor you ARE NOT going to make a 600 GHz clockspeed processor.
www.eFax.com are spammers
But seriously, a previous poster had a point, what's the relationship between the speed of a transitor and the speed of a proccessor? Because 600GHz is a HUGE jump over 3.4GHz. If there's a 1:1 ratio, then a proccsoor of with 600GHz transistors would have 176 tiems the proccessing power over the current breed. A Beowolf cluster in a single chip!
Free MacMini
That was the last time I remember Bipolar transistors being hyped as the next revolution in CPU technology. Back then Exponential Technology http://bwrc.eecs.berkeley.edu/CIC/announce/1996/x7 04-533.htmlwas developing a PowerPC processor that was claimed would be able to run at the unheard of speed of 533MHz. The Mac fans of the time were drooling over the prospects for Pentium crushing performance.(About 200-300Mhz at the time)
BYTE magazine article from the time http://www.byte.com/art/9611/sec6/art14.htm
We always seem to forget that all of our cool new toys carry an enormous environmental cost. Anyone have any idea if indium phosphide and indium gallium arsenide are better or worse for the planet than current technologies?
Replace every room heater in the world with a (networked) computer of the same wattage, get a supercomputer running with no additional energy consumption on top of normal heating needs. The nodes of the supercomputer that are active will mostly follow the nightside of the Earth and winter hemisphere, but who cares, as long as there are enough nodes at any given time, it doesn't matter whether they are located in Alaska or Chile. *This* is the future of computing.
the researchers have demonstrated a speed of 604 gigahertz - the fastest transistor operation to date.
How does one measure 604 gigahertz? Just asking.
Spare me. Please.
I call this the race to Google. It is a test for how long it takes for a desktop machine to actually become useable. This is usually a better measure than the ambiguous "boots in x seconds", that we often see. Here is how to perform the test.
Take any modern linux distro you like and install it as a dual boot with Windows XP. Now time how long it takes from pressing <enter> in grub (or Lilo if you are so inclined) and when you can see the main Google page. Try this with both Linux and Windows XP using Firefox as the browser on both.
The results? So far I have tested SuSE, Fedora, Mandrake, Slackware, Ubuntu and yes even Gentoo and in ALL cases, Windows XP was able to access Google faster than Linux on the same machine.
why would they switch technologies after investing $50 billion a year in their CMOS foundries etc.
Hopefully, competition.
$5 / month hosted VPS on linux = awesome!
I was under the impression that information had 0 thermodynamic value. Where'd you hear otherwise? (I'm curious to know, not flaming or doubting).
-Jesse
Nothing says "unprofessional job" like wrinkles in your duct tape.
Of course. A lot of Microsoft engineers spent a long time making XP boot as fast as possible. They'll probably do the same thing with Longhorn.
What remains is that when I minise a window on my parents' 700MHz XP box, it takes noticeable time to render the desktop, and these kind of delays will be much worse in Longhorn (probably).
CPUs have stalled out at about 4ghz overall clocking, cutting edge transistors seem to be hitting a wall at about 500-600ghz.
Now granted faster gate transitions make for faster CPUs, but multiple gate operations are necessary for each state change, add signaling and propagation delay and who knows what you can really clock the CPU at (I am not an Electrical Engineer).
Here is a page link claiming a record 562ghz transistor switching in Oct. 2002 article
here is another claimed record of 509ghz, Nov, 2003 article
Obviously at odds with the 2002 anoucment. Undoubtedly it should narrow its claim for a specific transistor type.
Here is a U of I annoucment calming a record 382 ghz Jan. 30, 2003 article
But expects 700ghz by early 2004 (I'm guessing they didn't make it).
Lets assume 562ghz in 2002, so we - drum roll please --- 7.5% increase in speed in 2 ½ years!
This is not going to keep Moore's Law humming along.
Even stranger, here are claims of TerraHertz transistors at Intel in 2002 article
Ironically, while googling for transistor or gate speed will show hundreds of hits, you can't actually find the switching speed for individual gates in a P4 or AMD chip. This stuff seems to be super secret stuff, and only the overall CPU clock it published. I wouldn't be surprised if the individual gates and transistors are transitioning at several dozens of ghz if not a couple of hundred or more. While Moore's Law death claims may have been premature 10 and 20 years ago, they may not be now.
I hope I'm wrong, I want my Holodeck Playstation 5 in 2015.
Letter To Iran
Sort of like with gold jewelry. It's too expensive for many people, so they mix in other metals to reduce the price.
I suppose this is a joke, but pure gold is rather soft. mixing it with other materials makes it hard enough to survive normal use.
"We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
But there's no "barrier" at 600 GHz or any other nice round number. It's just a number, and I wish tech writers and marketeers would quit using the "barrier" word in cases like this.
How did they measure this?
The wavelength of 600GHz is only half a millimetre (if I got that right). That means you could make an inverting oscillator by looping the output back to the input with a quarter millimetre piece of wire.
The wavelength issue is going to make design with these in digital circuits a real challenge. It will take a few orders of magnitude in process improvement to exploit these to the max.
Engineering is the art of compromise.
I'm surprised no one has noted that indium is rare enough that this transistor wouldn't be practical. Most of the indium is used in the transparent electrode (indium tin oxide) for LCD screens, etc. and it's in very short supply. Certainly not like silicon.