Slashdot Mirror
IBM Announcements on Chip Design/Nanocommunications
mr was one of the folks who wrote about some IBM scientists who have discovered a way to transport information on the atomic scale that uses the wave nature of electrons instead of conventional wiring. The new phenomenon, called the "quantum mirage" effect, may enable data transfer within future nanoscale electronic circuits too small to use wires. Big Blue also unveiled some new chip technology, called "Interlocked Pipelined CMOS" that starts at 1 GHz, and will be able to deliver between 3.3 - 4.5 Ghz.
ISO-9K is a great way to turn any company into a shithole.
- A.P.
--
"One World, one Web, one Program" - Microsoft promotional ad
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
This stuff is going to make our current PCs look like vacuume tube computers. There is no way such a fundamentan shift in `direction' is going to be only an incremental step. This will be like going from plates and grids and electron streams to silicon layers and fields. This has a lot of potential.
Bill - aka taniwha
--
Leave others their otherness. -- Aratak
It's got to be more then that. The TI SuperSPARC did that in '95 or so. Most of the CPU ran at (I think) 50Mhz, but the register file at 100Mhz. In that case (and all similar cases I know of) all the clocks are multiples of each other (no relitavly prime clock rates).
More over if the decode unit runs faster then the dispatch unit there is no useful gain. The dispatch unit has to be fed decoded instructions. The same is true for many (but not all) other parts of the CPU.
I think (with no proof, and no help from that watered-down article) that most of the parts in this chip run at the same speed (say 1Ghz). They each have their own 1Ghz clock distribution net, and they are not sync'ed with each other (so the ALU could see start of clock while the decoder is half way through a clock, and the renamer has just seen the end of a clock pulse). The boundry between each clocked unit must have a small async buffer.
That would trade the pain of clock distribution for the pain of having a bunch of async buffering all over the place adding to latency. Given how painful clock distrubution is in really big chips this is probbably a positave tradeoff. At least for latency intolerent workloads.
OTOH, IBM has been noted for their research programs. About 20 yrs ago, there was the big three in terms of corporate research. They were AT&T, IBM, and Exxon. Well Exxon has greatly reduced their research efforts (as had the other major oil companies), AT&T has been split up and then again split up again (Bell Labs is part of Lucent), while IBM has redirected their researchers to perform more applied research. But IBM research is still very impressive. Low temperature super-conductivity was an IBM product that came out their Zurich research facility.
Off topic, but Thomas Watson many years ago had the now-famous Think posters put up. I used to have a cartoon in my office that showed the Think poster with a guy saying, "I'll like to, but I have too much work to do."
Ross Perot founded EDS (after being an IBM salesman) to provide software for IBM mainframes. Back then, IBM philosophy was to sell hardware, software was just an afterthought. Hmmm, I wonder if anybody else got rich for selling software that ran on IBM hardware.
Yup, I'm just rambling. IBM is a "friend" of linux at this moment. They have been very good for the time being. However, as a person who has witnessed the might of IBM in the past, I'm scared of what IBM could potentially do to screw things up. Remember, the enemy (linux) of my enemy (MS) is my friend. Of course, we all live by the ancient Chinese saying/curse, "May you live interesting times."
to run quake4.
;) Now all of my machines are out of date, and I won't be able to play any new games at all.
Great.
This sig is false.
My question is are the clocks phase locked to a master timing source or are they free running?
Mea navis aericumbens anguillis abundat
I've read some papers on research operating systems that use huge global virtual address spaces that are shared across multiple computers. Each object gets a globally unique virtual address that is never reused. That can use up a lot of address bits.
Mea navis aericumbens anguillis abundat
Is it just me, or has IBM made a real turnaround in the last 5+ years? It seems they understand the whole open source movement, they've pretty much ditched they're sorry aptivas, and they seem to be a leader in new technologies. On top of that they've changed the way people percieve them. I remember hearing stories about how they had to wear knee-high black socks to match their black suits long ago, and now I go to an interview with them, and the guy is wearing jeans and a Polo shirt!
Honestly, this is one large corporation I have respect for. And there aren't too many of those left now and days.
My grandma could do that in her sleep, and she's dead! Company A did nothing special. Besides, there was a paper written at an obscure university ten years ago which made reference to something like what is described
Hmmmm...
1: It's the "obscure university" that Alan Turing was a professor at.
2: It's the "obscure university" that built the world's first stored-program computer (the Manchester Mark I/"Baby")
3: It's either very similar or identical technology.
4: They've built the chips. They have prototypes.
5: Funny how everyone jumps up and down in defence of IBM, when they quite happily quote unrelated tech after unrelated tech to prove that Microsoft doesn't innovate...
Simon
Coming soon - pyrogyra
This was probably quite difficult to implement, but isn't exactly conceptually brilliant. Modern computers already run at different clock rates internally. Your disk I/O bus runs at one speed, your video processor runs at another speed and the CPU still spends a lot of time waiting for stuff to come down the system bus from memory.
;-)
It's even less conceptually brilliant, when you see what people elsewhere have been working on - namely wavepipelined architectures.
Funny... people just keep on reinventing the wheel... fire... and then they patent it to hell.
IIRC, the guys at Manchester University were working on this back in 1989/1990 (or at least they were when I went on a tour of the place...). Back then, it was just called the "wave pipelined RISC chip" - these days, it's the "Amulet". Check it out. It's based on ye olde ARM processor architecture - but the implementation is completely asynchronous -- that is, each individual logic element is clocked separately.
Sure, it's still experimental... sure, it's slower than other chips - but it also predates IBM's announcement by about 11 years. Just goes to show - academia ain't entirely useless
Links
Architectural Overview at Berkeley
The Amulet Asynchronous Logic Group at Manchester University
Who needs clocks? Bah!
Simon
Coming soon - pyrogyra
What's really nice is that IIRC alot of Drexler et al's work on nanotech is concerned with avoiding quantum effects that would disrupt their atomic-scale gears, etc. Here the scientists are turning the problem on its head and using the quantum nature of matter at the nano scale for their nanocomputing device. However, obviously heat is still a problem (cooled to 4K - don't think Kryotech will cut it anytime soon ;-p)
#include "disclaim.h"
"All the best people in life seem to like LINUX." - Steve Wozniak
#include "disclaim.h"
"All the best people in life seem to like LINUX." - Steve Wozniak
To increase speed, IBM researchers decentralized the clock, using locally generated clocks to run smaller sections of circuits. The design thus allows faster sections of circuits the freedom to run at higher cycles. It also significantly reduces power requirements.
/. a few days ago.
This was probably quite difficult to implement, but isn't exactly conceptually brilliant. Modern computers already run at different clock rates internally. Your disk I/O bus runs at one speed, your video processor runs at another speed and the CPU still spends a lot of time waiting for stuff to come down the system bus from memory.
As far as I can see, IBM have scaled this down to a single chip, which will increase overall throughput considerably. Difficult to do, very worthwhile, but conceptually all they have done is to get the latency issues into a smaller space.
OTOH, this could lead to an architecture with considerably lower power consumtion, which is definitely worth doing.
The bit about 'quantum mirages' has already been discussed on
They say they use multiple clocks to increase speed. Sometimes the best inventions are those that simply make sense. I mean, most commodity chips heretofore are locked down to one clock. That means the tiniest circuit still has to wait for clock to compute another value. That doesn't make any sense. Have independent smaller clocks, and make the computing asynchronous and have each component just fill up a queue. Match up the results with their ids/stamps and there you go. Without independent clocks, the slowest component will dictate the overall cpu speed.
Jazilla.org - the Java Mozilla
It's 10 PM. Do you know if you're un-American?
It will be interesting to see if this results in a practical quantum waveguide to replace wiring. Just insert (or remove) an electron at the one end of the pipe and it will produce (or delete) a marage at the other end.
I wonder if the "mirage" could be interpreted as the electrons of the cobalt atom tunneling to the image location and spending a fraction of their time there? That less-than-half strength might be because the nucleus is still at the other location and makes the electron density "prefer" that region because it is lower energy, due to the attraction of the positive charge.
I also wonder what is the speed of propagation of the effect? Switch a gate's output by dropping an electon into an electron trap at one end of the waveguide, and it appears (at, say, 50% density) at the other end, and affects the logic there. How long does it take to happen? Does it exceed the speed of signals in a wire? (That's a very small fraction of lightspeed on a chip, where the wire resistance and stray capacatance form a delay line.) Does it approach that of light in vacuum? Does it EXCEED that of light in vacuum? (Even if the total system can't send signals faster than light in emptyness, which is a very slight improvement on light in quantum vacuum.)
Whatever it is, my bet is that it will happen at tunneling speed.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Well, crusoe isn't the fastest processor at all. It's just the lowest power using (wrt to speed) x86 compatable chip.
I'd much rather have a cluster of P3 or Athlons.
No you don't. It'd be faster, but you could do it with just 4 32bit numbers.
Ok the reference for this work is:
H. C. Manoharan, C. P. Lutz & D. M. Eigler, Nature403, 512-515(2000).
In this experiment a few Cobalt atoms were deposited on a Copper surface. Using a scanning tunneling microscope the Co atoms were gently dragged into an elliptic(coral) structure, and one Co atom was placed at the focus of the ellipse. (The images of this stuff are gorgeous and more cool STM images of atoms and atomic maniputation can be found at the STM Image Gallery).
Due to the magnetic nature of the Co atom electrons near the atom tend to align their spins with the Co magnetic field screening the magnetic moment. This local phenomina can be imaged by the STM, the surprising result is that another mirage image appears at the second focus of the ellipse. This suggests some sort of long range electon ordering.
These experiments are being done with a low temperature ultra-high vacuum stm (this stuff is damn hard) and to reproduce these same results in a next generation processor as a means to transport data is unlikely in the near future. Nevertheless, these results will have a great effect in our understanding of macroscopic quantum systems and ordering.
Using the 'quantum mirage' process, previously posted Slashdot stories magically reappear at another time and place.
Moore's Law sucks it took less then a year for my 400mhz and 500mhz computers to be outdated.
This stuff is kinda of cool though, maybe we'll get to see some nice VR stuff and better speech recognition. Eventually the hardware will be fast enough to run windows.
I would have liked to have seen intel go to a 128 bit architecture instead of 64 it would have lasted longer.
Environmentalists are their own worst enemy. ~tricklenews.com
I can see the error messages for Win2010 now:
"Error - user32.exe performed an illegal operation in this universe - please continue in another universe or restart..."
this is very true, the technology is being upgrade so fast that none can keep up and it's regoddamdiculous to even try to buy the fastest CPU on the market, not only will it be out of date tomorrow, you won't be able to use it for lack of a supporting mobo (motherboard for those of you who don't know). this is making many more clock speeds available, but making a stable and longlasting system less likey. think about it, they're making quantity, not quality now. sure, i want a gighertz processor as much as anyone, but i think i'll wait a year before i get one so i know it works and there's a motherboard for it, but than, it will no longer be available nor will supporting motherboards. Well, I guess I'd just like to share my opinion that things are moving TOO fast now, and the race needs to slow down or... well, there has to be a way that the end user doesn't get screwed. - AZ
- AZ
So does this mean that the IBM 1-piece 386 dos based terminals in my highschool's computer lab aren't state of the art anymore?
-- Just the FAQs Ma'am.