Simulating 1 second of real brain activity takes 40 minutes and 83K processors

A team of Japanese and German researchers have carried out the largest-ever simulation of neural activity in the human brain, and the numbers are both amazing and humbling.
The hardware necessary to simulate the activity of 1.73 billion nerve cells connected by 10.4 trillion synapses (just 1 percent of a brain’s total neural network) for 1 biological second: 82,944 processors on the K supercomputer and 1 petabyte of memory (24  bytes per synapse). That 1 second of biological time took 40 minutes, on one of the world’s most-powerful systems, to compute.
If computing time scales linearly with the size of the network (a big if; I have no idea if this would be the case), it would take nearly two and half days to simulate 1 second of activity for an entire brain.

The K computer: it's big.

The K computer: it’s big. Source: RIKEN

Still, the researchers are excited by what they’ve accomplished. According to a quote from project leader Markus Diesmann in the press release announcing the simulation: “If peta-scale computers like the K computer are capable of representing 1% of the network of a human brain today, then we know that simulating the whole brain at the level of the individual nerve cell and its synapses will be possible with exa-scale computers hopefully available within the next decade.”
Although they’re measured in FLOPS — floating point operations per second — rather than bytes, the prefixes measuring supercomputer performance are the same as those measuring data storage. A system operating at 1 exaflop would be 1,000 times more powerful than a system operating at 1 petaflop. K — now the world’s fourth fastest supercomputer — is capable of 10.51 petaflops.
Importantly, though, the recent simulation was merely a test of the open source NEST simulation software the researchers have been developing. Research into specific diseases, and projects such as Europe’s Human Brain Project and the United States’ BRAIN initiative, will require more job-specific tuning.
Figuring out the mysteries of the human brain isn’t just a matter of sheer scale and advanced software, though. As GigaOM’s Stacey Higginbotham has reported, numerous companies and institutions, including IBM, are working on projects to better our understanding of the brain. Some are even developing chips designed to mimic it — albeit on a much smaller scale.
This article was updated at 11:34 p.m. on Aug. 3 to correct the amount of memory allocated per synapse to 24 bytes.
Feature image courtesy of Shutterstock user StudioSmart.