From Eugene Izhikevich’s website (which by the way is full of cool stuff that we’ll probably post about someday):
“On October 27, 2005 I finished simulation of a model that has the size of the human brain. The model has 100,000,000,000 neurons (hundred billion or 10^11) and almost 1,000,000,000,000,000 (one quadrillion or 10^15) synapses. It represents 300×300 mm^2 of mammalian thalamo-cortical surface, specific, non-specific, and reticular thalamic nuclei, and spiking neurons with firing properties corresponding to those recorded in the mammalian brain. The model exhibited alpha and gamma rhythms, moving clusters of neurons in up- and down-states, and other interesting phenomena (watch a 25M .avi or .mov movie).
One second of simulation took 50 days on a beowulf cluster of 27
processors (3GHz each). Why did I do that?”
neurodudes 
Thanks to Erik Flister for telling me about this
LikeLike
He says that “Instead of saving synaptic connections, I re-generated the anatomy every time step.” I wonder how much additional overhead this technique caused, and how much longer the simulation ran as a result.
Also, is anyone else satisfied with waiting 40 years to run these quickly? This makes it especially clear that we’ll need alternative hardware. Good luck to Blue Brain nonetheless…
LikeLike
Hm. That synaptic connection bit also means that a) his simulated brain is static and can’t learn more than what sustained activity can hold in memory and b) it has some highly redundant structure allowing him to compute the weights. Both are not true for natural brains, making his estimate somewhat optimistic, I think.
LikeLike