Modern digital chips operate are two-state machines, working only with 0s and 1s. These states are normally electric currents/charge, where some value below a (pre-determined) threshold (say 3 Volts) would be a 0, and above would be a 1. Chips spend a lot of energy ensuring that a 0-state is indeed a 0-state (and not a 1-state with a large negative error in the voltage), and vice-versa. For handheld devices (PDAs, MP3 players) that run on batteries/solar-power, this is a problem, because a) efficient power management allows such devices to run for longer time, and b) smaller devices have problems dissipating the excess heat.
PCMOS based on the PBITS framework (Courtesy: Georgia-Tech
This PBITs model is now backed by measurements of an actual probabilistic CMOS device which the researchers call PCMOS (Probabilistic Complementary Metal-Oxide Semiconductor). Supported by Defense Advanced Research Projects Agency (DARPA), the research team estimates that 100-fold improvements are possible to the energy consumed and performance of complex applications such as neural networks, which are used for pattern recognition and other applications such as spoken alphabet recognition programs used on cell phones.
According to the vice-president of Research Operations (Dr. Ralph K. Cavin III) at Semiconductor Research Corporation, The most striking thing about the work is the idea that we can utilize a phenomena normally viewed as unwanted (noise on the chip) as a vehicle to address an important and limiting problem (reducing heat dissipation). There is something powerful and appealing about turning a problem into a feature!
PCMOS based on the PBITS framework (Courtesy: Georgia-Tech
This PBITs model is now backed by measurements of an actual probabilistic CMOS device which the researchers call PCMOS (Probabilistic Complementary Metal-Oxide Semiconductor). Supported by Defense Advanced Research Projects Agency (DARPA), the research team estimates that 100-fold improvements are possible to the energy consumed and performance of complex applications such as neural networks, which are used for pattern recognition and other applications such as spoken alphabet recognition programs used on cell phones.
According to the vice-president of Research Operations (Dr. Ralph K. Cavin III) at Semiconductor Research Corporation, The most striking thing about the work is the idea that we can utilize a phenomena normally viewed as unwanted (noise on the chip) as a vehicle to address an important and limiting problem (reducing heat dissipation). There is something powerful and appealing about turning a problem into a feature!
Fuzzy Computing
(Hide)
(Show)
8 Comments:
Lucretia: Agreed. Also, a probabilistic computer is like a poor man's quantum computer.. instead of two state, you could in theory have a mixture, and therefore, solve problems parallely.
Wayne: Quantum computing has a long way to go! In the meantime, things are only going to get smaller (and I dont expect to see quantum computing in handhelds anytime soon!). So we need something that is more power efficient.
By harnessing living cells, the idea of binary calculations becomes a little quaint when there are so many other possible states available.
And, given that Japanese researchers have figured out how to extract chemical energy directly from the bloodstream, how long before motor dysfunctional disorders and such like become a thing of past?
This kind of technology really is the future...
If you want a notebook, it might make sense to go directly to centrino. For desktops, if you wait a year or so, yo ushould be able to get the new 64 bit processors!
Post a Comment