Rice University today used the International Solid-State Circuits Conference to reveal that it has tested a real-world example of a processor founded on probability math. Called a PCMOS (probability-based complementary metal-oxide semiconductor), the chip abandons the either/or Boolean logic of all current processors in favor of calculations that rely on the most likely answer in most cases. By avoiding a reliance on getting an exact answer when unnecessary, the CPU uses just a fraction of the power to accomplish the same work as today's chips; the example chip's voltage is dropped to where it consumes 30 times less power than an equivalent, ordinary CMOS processor.
As it has to wait less often for answers, the result is also about seven times faster and is especially suited to math where imprecision is unnoticeable or even encouraged, such as small image rendering or data encryption. The Houston-based university believes PCMOS would be particularly useful for typical home users in cellphones, video cards and other devices with embedded processors.
Co-developed with Singapore's Nanyang Technological University, the technology is still in the early stages but, according to inventing professor Krishna Palem, should enter real-world production within the next four years.