Ruminations on the electronics industry from David Manners, Senior Components Editor on Electronics Weekly.
Manchester Cracks Parallel Processing Bottleneck
A Manchester University group, called SpiNNaker, under the leadership of Professor Steve Furber, co-designer of the ARM microprocessor, may have cracked the key problem affecting the widespread adoption of the technology.
Parallel processing has become the accepted way forward for computing, now that IC scaling is becoming too expensive and too leaky. But the problem with parallel processing is that adding more processors results in diminishing performance returns. The main reason why you don’t get a linear scaling effect on performance from the addition of more processors, is the problem of the communications links between the processors and the rest of the system. These communications links get clogged up, leading to a slowing down of the system. It is this communications bottleneck which Furber’s group may have solved. The SpiNNaker solution is currently the subject of a patent application. SpiNNaker is backed with £1m in funding from the Engineering and Physical Sciences Research Council (EPSRC) and aims to build a computer which mimics how nerve cells in the brain interact. Manchester is the spiritual and technological home of worldwide computing, being home to the first computer to be made with all the features of a modern computer, the Small Scale Experimental Model in 1948. The SSED was followed a year later by the Manchester Mark 1 computer, and the world’s first commercially available computer the Ferranti Mark1, the first of which rolled off its production line in 1951.Tags: Ferranti, mark 1 computer, parallel processing, physical sciences research council, ssed