19-10-2016 | | By Paul Whytock
Could it be possible that the architectural and computational performance of future computers lies within the anatomy of a 1mm long roundworm? Quite possibly is the answer to that if the continuing research by the Si elegans project proves successful.
But just what is this worm? The Caenorhabditis elegans is a none parasitic transparent roundworm that lives in temperate soil environments. The name is a blend of the Greek caeno (recent), rhabditis (rod-like) and Latin elegans (elegant).
More commonly referred to as C elegans, it is the first multi-cellular organism to have its whole genome mapped. The reason for this is that biological nervous systems work in different ways than conventional computers and any simulation running on them. In nature, information is processed in a highly parallel way and successfully emulating this process potentially offers orders of magnitude increases over current technologies in computational speed.
The Computational Intelligence Laboratory of Nottingham Trent University is part of the Si elegans project, established in 2013 to create such an emulation platform for the C elegans nervous system. Their plan is to create a reconfigurable, scalable and modular open-access computational platform that will provide a technological blueprint for a new era of brain-like computational architectures. In conjunction with Nottingham Trent University.
Verotec, a specialist thermal management and electronic enclosure manufacturer has provided custom 6U 450mm deep 19inch KM6-II subracks (pictured), configured to house 376 dynamically reconfigurable field programmable gate arrays (drFPGAs) arranged in a 3D interconnection configuration. Each subrack houses eight FPGA boards and every one of these has an integral power supply unit. Interconnection is via Gigabit Ethernet.
These will replicate the neuron and muscle connectome of C elegans; connectomes being a map of connections in the brain or in other words a neural circuit diagram.
There is no doubt that Moores Law could do with a computational boost right now. Forty years ago it said that the number of components per integrated circuit would double every year. By 1975 that forecast was revised to every two years and this estimate turned out to be on the money and proved accurate for twenty years.
However, Moores Law has only ever been a projection and these are not a perfect science and some industry pundits see it declining or potential disappearing as we move into 10 nanometre node technology. Hopefully this research will prove them wrong.