|
|
|
TRANSISTOR DIMMER switches that are not simply turned to fixed off and on positions could be the technology that enables 1,000 times performance improvements over today's computers, say scientists.
Computers have been based upon digital principles with bits as ones and zeros and boolean logic gates. Transistors perform operations with these bits, in a serial on, off fashion. But a new technology varies the energy at a logic gate, creating values between one or zero, values that represent probability.
Using this principle, researchers who invented the technology at Massachusetts based Lyric Semiconductor say this probability logic can be used to perform many computations in parallel. This form of computation is what scientists think will realise 1,000 fold performance improvements in computing power.
"After a decade of development...we are starting with Lyric Error Correction [for flash memories] but ultimately plan to develop a more general purpose probability processor that will truly change the landscape for many applications," says Lyric's CEO and co-founder Ben Vigoda.
Flash memory error correction is increasingly required because read error rates have become problematic, with one in every thousand bits stored in a flash memory being read wrongly.
At first glance this technology reminds The INQUIRER of Douglas Adam's infinite improbability drive. Based on the probability of a subatomic particle being in a particular place, such as near the nucleus of an atom, but also being at an entirely different place, potentially, very far from its point of origin, in all probability the spaceship propelled by the drive could be anywhere at any point in time, even inside a Sperm Whale that had been called into existence, several miles above the surface of an alien planet, and since this is not a naturally tenable position for a whale, this innocent creature had very little time to come to terms with its identity before it would plunge to its death.
A spin-off from the Massachusetts Institute of Technology, unsurprisingly Lyric was funded by the US military's wacky technology development organisation, the Defense Advanced Research Projects Agency. µ
You have an analog computer in your head! -- Okay, that's just *my* conclusion. If you can otherwise explain how slow chemical "wet-ware" is able to do complex real-time processing that digital computers still don't even approach, then have at it.
As to definition: if this system has signals "that represent probability", then it's closer to analog, at least going the right direction.
Analog computing is something completely different from this. This is still, ultimately, using 1's and 0's. Only difference is that they are transposed over one another. So a logic gate that now handles only 1 operation at a time can handle 10+ (though early chips will probably be closer to 2-3). That's still a heck of a space savings. And probably a power/heat saving as well, as the chip will almost certainly not require more (or much more) power to handle the requests.
If you want to compare it to something, compare it to multilayer optical discs. Multiple bits of information is stored and retrieved in the same space, depending on what wavelength the ODD laser writes/reads at.
If they can get this working at a high enough density, it'll be revolutionary, both for performance and price. Imagine a single CUDA running 512 cores. The newest GTX could be replaced by a tiny chip while still incorporating a couple of gigs of ram right on the dye.
Or 10GB of random access memory in your netbook; a 12 core Atom processor; or your cell phone having the power of the highest end workstation.
The reason being that it SUCKS.
The advantages of digital systems are still what they always were, and multiplexing 1000 CPUs onto a single physical die using different signalling bands like phase quadrature or whatever doesnt suddenly make it an analog computer.
In my ever humble opinion, of course.
I suspect that digital designers haven't really grasped analog computation, even while converging toward it in a "back to the future" way. I think some sort of massively parallel analog computer is the only possible way to even approach Artificial Intelligence, but we're in the grip of Digital Monopolists who are now holding back progress.
Catch phrase:
"I was analog when analog wasn't cool."
We need a law that says anyone claiming to be inventing a technology that was widely known decades ago, should be shot in the damn head.
How dare they fund such a crazy idea. Next thing you know such agencies will be funding things like health care advances, or even moon flights! What are we thinking? Advancements? Those are for the idiots that think we can use our laps to hold a computer. Bah!
How dare they fund such a crazy idea. Next thing you know such agencies will be funding things like health care advances, or even moon flights! What are we thinking? Advancements? Those are for the idiots that think we can user out laps to hold a computer. Bah!
- Researchers use CPUs to feed GPUs
- IBM shows off sub-10nm carbon nanotube transistors
- Microsoft shows off touchy feely surfaces of the future
- Boffins invent an operating system for cities
- Nike is readying Back to the Future shoes
- IBM and 3M are developing semiconductor adhesives
- Amazon chief fails to deliver rocket to space
- Lightbulbs can be used to broadcast wireless internet
