IBM says new chip 250x faster than todays....

[Dr Mordrid]

Still developing on the wire....

Mon Jun 19 2006 22:57:25 ET

Researchers at IBM and the Georgia Institute of Technology are set to announce Tuesday that they have broken the speed record for silicon-based chips with a semiconductor that operates 250 times faster than chips commonly used now.

NEW YORK TIMES:
>
The achievement is a major step in the evolution of computer semiconductor technology that could eventually lead to faster networks and more powerful electronics at lower prices, said Bernard Meyerson, vice president and chief technologist in IBM 's systems and technology group. He said developments like this one typically find their way into commercial products in 12 to 24 months.
>

Dr. Mordrid

[Technoid]

IBM seems to do that annually

[High_Jumbllama]

Parallel??

Link??

[agallag]

http://www.nytimes.com/2006/06/20/technology/20chip.html?ei=5090&en=215511bacfc970b5&ex=1308456000&partner=rssuserland&emc=rss&pagewanted=print

[High_Jumbllama]

http://www.eetimes.com/news/semi/sho...leID=189500692

[Wombat]

Notice they brought this chip down to something ridiculous, like 4 Kelvin. Lots of things will run damn fast at that temperature.

[Dr Mordrid]

U of M announced a quantum computing chip architecture that uses on-chip laser cryo-cooling (<1 deg. Kelvin) and manipulates single atoms in magneto-optical ion traps that can be built using normal fab techniques. They're scaling it up as we speak.

Dr. Mordrid

[Wombat]

Great. Scaling it up to what? We've been able to bounce a single electron around a near-0K superconductor for years. Not a big deal. Not applicable to any reasonable computing environment.

Electrons can only move so fast, folks. They can't get across a CPU-sized chip if you don't give them enough time. Remember that at 5GHz, a photon only moves 6cm (in an optimal vacuum).

[Dr Mordrid]

You do mean an electron, right?

Check the Dec. 11, 2005 issue of Nature Physics for the initial work. This is more recent;

University of Michigan develops quantum processor

Humphrey Cheung

January 17, 2006 20:04

It's much too early for AMD or Intel to start looking over their shoulder, but the University of Michigan has developed a quantum chip that contains one cadmium ion. The ion, which is suspended in electrical fields, can exist in many possible states which collapse into one when viewed by an outsider. Quantum computing has been touted as great leap in computing, but still faces many challenges.

Composed of gallium arsenide, the quantum chip was made with the same microlithography process that many modern processors are made of. Miniature lasers blast the trapped ion giving it various "spin" states. Ions must be protected from the environment to prevent "decoherence" a process where the ion's data is corrupted by the surrounding environment.

"We levitate the atom in the chip by applying certain electrical signals to the tiny nearby electrodes," explained Professor Christopher Monroe, University of Michigan Physics professor and co-inventor of the chip. While other researchers use neutral atoms, Monroe's chip traps ions - atoms with missing or extra electrons - on his chip.


Ions being unstable, require extra radio-frequency waves to hold in place. This instability is actually a blessing because the ion trap can scaled up to hold many ions. In contrast, stable neutral atoms can be held solely by magnetic fields, but researchers are having a tough time getting separate atoms to interact.

Unlike regular processors which only recognize 1s and 0s, quantum computers use qubits. These qubits can be 1, 0 or anything in between. The correct value is shown only when the chip checks the ion.

Scalability is key to producing a viable quantum computer because one qubit doesn't do any good - but a series of them can allow for much faster computers, especially with equations involving factoring. A traditional computer would do everything in series, meaning multiply one by one, but a quantum computer has already calculated all the answers. The difficulty is getting identifying the correct one.

Future quantum computers won't necessarily supplant traditional processors from the likes of AMD or Intel. Quantum computers can excel in computations involving waveform analysis or cryptography or anything where you must reduce a large set of data to find an answer, but don't do as well with Microsoft Word or checking email. In addition, new formulas must be made to deal with the self-collapsing nature of these computers.

The ordering and construction of equations can collapse all the qubits, which effectively nullifies original purpose of a quantum computer, namely to store many states at once. In addition, error correction is tricky business because any measurement causes changes in the system - Checking for errors can cause errors.

In addition to solving specialized equations, quantum computing could offer tremendous storage. Assuming researchers get past the decoherence problem; each additional qubit doubles your storage capacity. Two qubits can store 4 regular bits of data, while three can store 8 bits. While those numbers may not look exciting, 1024 regular bits can be stored with just ten qubits and the sky is the limit after that.

A few firms and government agencies are using quantum technology already to transfer data over encrypted fiber optic links. Research is being funded by several agencies such as the National Science Foundation and the National Security Agency.

The one ion chip is only the first step for the University of Michigan researchers and Monroe says that the chip could be scaled up to include "hundreds of thousands of electrodes."

[Wombat]

No, I meant a photon. I was intentionally comparing to the fastest known-possible velocity for information. Information transmission via electrons in a conventional conductor is significantly slower (supercooled superconductors close that gap some).


You're posting incoherently. What does a quantum-comp research project have to do with conventional (non-quantum) chip architectures?

You always try to drown people who oppose you with information. Even when that information is unrelated to the point in debate.

[Dr Mordrid]

Not drowning at all. Just got carried away due to U/M connection plus that there is more than one way to skin a cat, or do a calculation. Call it an 'ain't this cool too?' post.

That said; if someone else had put it in the thread as 'ain't this cool too?' would your reaction have been the same? Hmmmmm.....

Dr. Mordrid

[Wombat]

Yeah, actually. I still keep a lot of ties to the industry, and this is not exciting compared to other accomplishments.

[High_Jumbllama]

Notice they brought this chip down to something ridiculous, like 4 Kelvin. Lots of things will run damn fast at that temperature.

From the links, it says that it operates at 350GHZ at room temperature. That's not too shabby either.

[Wombat]

But for what? If it were just a clock (and you can't get an electron in 2ps anyway)....

[High_Jumbllama]

But for what? If it were just a clock (and you can't get an electron in 2ps anyway)....

For the clock as that was the chain of conversation.