Hardware manufacturer, IBM has a processor built around such functions as the brain. The neural processor is fundamentally different than current processors and is specialized tasks much more fuel-efficient than current processors.
The neural processor TrueNorth has 4096 cores, made on Samsung’s 28nm process. In total, the processor has 5.4 billion transistors. Each processor core has over 100,000 bits of memory to store whatever his state is, and what connections he may impose.
The difference with conventional processors is that the information does not come from the binary view of the to or stand of the transistors, but from spikes, where the information is in the timing and frequency of these spikes. In addition, the ‘neurons’ in the processor that spikes send to and receive from 256 other ” neurons; with conventional processors is that not possible.
Another architecture is the processor is much more efficient at performing certain tasks, such as object recognition in video images. The researchers from IBM and Cornell University say in the scientific journal Science that in the analysis of video images of 240×400 pixels at 30fps the chip 65mW consumed, current processors have much more energy.
IBM has spent years working with neural processors. The began to own say six years ago and presented last year in a software ecosystem for these processors, because current software is not working.
Comments
(100)