Video below shows SpikeFun in action simulating 8 million neurons with 1.45 billion (yes, that's right - billion) synapses. This simulation takes approx ~130 GB of RAM + 20 more for additional data (visualization, DSI tracts, etc...).
Check the video out (it is also available in Full HD and 2560x1440):
SpikeFun v0.68 introduces simulation video capture directly to Google WebM format - eliminating the need to manually transcode captured video frames. In addition, SpikeFun 0.68 also brings support for processor groups, feature introduced with Windows 7 / Windows Server 2008 R2. With processor group support, SpikeFun is now able to fully utilize all processors on systems with more than 64 processors installed (up to 256 logical processors with Windows Server 2008 R2). Click on the title to find out more!
SpikeFun v0.67 brings support for CPU and memory bandwidth benchmarking (SpikeBench) as well as support for hardware monitoring using Intel(r) Performance Monitoring Unit (PMU) present in Intel's Core i7(TM) processors. Results on the reference system are analyzed... Click on the title to find out more!
SpikeFun v0.65 contains extensive optimizations for NUMA architectures (such as Intel(R) Xeon(TM) E5-2600) with multiple CPU sockets and memory divided per CPU groups. Speed up of over 25% is measured on our reference system. Click on the title to find out more!
SpikeFun v0.63 brings support for user-configurable neuron parameters, variable cortical thickness, optimized OpenGL rendering using GPU shaders as well as more hand-optimizations for CPUs supporting Advanced Vector Extensions (AVX) instruction set and much more! Click on the title to find out what's new!
SpikeFun v0.57 has just been released and it is loaded with new features ranging from the new network generation algorithm all the way down to the new command line interface... click on the title to find out what's new!
SpikeFun v0.54 has just been released and it adds the ability to follow anatomical data related to axonal pathways (connectomes). During the simulation build-up, SpikeFun will use connectivity matrices to route parts of axons belonging to pyramidal and thalamic relay neurons. This feature is in a very early stage, so expect lots of improvements and better precision in the future.
As described in the previous post, SpikeFun v0.5 is using 3D brain model for neural network build-up (if desired). With quite a big jump in complexity of the neural network, there are some very interesting emerging phenomena that could be observed if the network is fairly large. Click on the title to find out more!
SpikeFun v0.5 has just been released and it contains some quite big improvements. How about real 3D model of the brain? If you are interested - click on the title to read more!
