GPUs ‘well-suited’ for powerful brain simulations

GPUs could play a key role in creating simulations of the human brain.
19 December 2018 | 366 Shares

GPUs could provide a cheaper and much more powerful architecture for brain simulations. Source: Shutterstock

Researchers at a UK university have been able to create the fastest and most energy-efficient simulation of a rat’s brain using ‘off-the-shelf’ computer hardware.

Using GeNN neuronal network simulation software and readily-available GPUs (graphics processing units), Dr James Knight and Professor Thomas Nowotny of the UK’s University of Sussex (UoS) claim to have achieved brain simulations that exceeded the capabilities of a “top-50 supercomputer”.

The development of faster and more efficient simulators could help in our understanding of brain function and neurological structure. Findings could be applied in the treatment of neurological damage and disorders in patients, such as pinpointing areas of the brain that cause epileptic seizures.

Improved simulators could also accelerate progress within the development of artificial intelligence (AI) technology. UoS is already using GeNN to build autonomous robots, including flying drones which can be controlled through simulated insect brains.

While computers have become “drastically more powerful” over the last three decades thanks to our ability to create computer chips with smaller and smaller components, Nowotny believes progress has now “hit a wall”— it has become much harder to build faster computers without employing “radically different” architectures.

“GPUs are one such architecture and our work shows that, in the near term, they are a competitive design for high-performance computing and have the potential to make advances far beyond where CPUs have brought us to so far,” he said.

Using their own GeNN software, the team tested two established computational models, including a cortical microcircuit— “consisting of eight populations of neurons and a balanced random network with spike-timing-dependent plasticity”— a process which has been fundamental to biological learning.

Using just one GPU, the researchers were able to achieve processing speeds up to 10 percent faster with 10 times the energy savings of what’s currently possible using either a supercomputer or the SpiNNaker neuromorphic system, a custom-built machine developed as a part of the £1 billion (US$1.27 billion) European Human Brain Project (HBP).

As a result of their flexibility and power, the UoS team believes GPUs could play a key role in creating simulations capable of running models that begin to approach the complexity of the human brain.

“Although we’re a long way from having the understanding necessary to build models of the entire human brain, we’re approaching the point where the latest exascale supercomputers have the raw computing power that would be required to simulate them,” said Knight.

“Many of these systems rely on GPUs so we’re delighted with these latest results which show how well-suited GPUs are to brain simulations.”

Over the next year, the researchers hope to extend the work to a model 50 times larger, with the aim of simulating a monkey’s visual system, using multiple, interconnected GPUs.