4 Billion Synapses, 16.7 Million Neurons, Visual System

DigiCortex NVIDIA V100 ("Volta") Update: 12.7M Neurons Simulated in Real-time on 8x V100 GPU

NVIDIA Volta has just been made available to wide public through AWS. We have tested DigiCortex v1.22 on a 8x NVIDIA V100 system and achieved real-time simulation performance of 12.6 million neurons and 431 million synapses, which is almost 3x speedup compared to K80!

DigiCortex NVIDIA P100 ("Pascal") Update: 5M Neurons Simulated in Real-time on Dual GP100 GPU

We have tested DigiCortex v1.22 on a dual NVIDIA P100 accelerator with significant results: real-time simulations of 5 million neurons and 200 million synapses are possible, thanks to the HBM2 memory bandwidth and other Tesla P100 features. Click on the article to read more.

DigiCortex 1.21 - Support for NVIDIA Pascal - 1.8 Million Neurons Simulated in Real-Time!

DigiCortex v1.21 brings improvements in the GPU area: it now includes support for multi-GPU CUDA-accelerated simulations and a full support for the latest generation NVIDIA Pascal GPU architecture

SpikeFun 1.16 - Bugfix for Big Haswell Processor Group Bug (>64 logical cores)

DigiCortex 1.16 fixes a nasty bug which resulted in crashes when running on systems with more than 64 logical cores. This happened on our test system based on dual-CPU setup with Intel Xeon E5 2699 V3, which is the first time a 2S system has >64 logical cores. Fix is available with DigiCortex 1.16, and the article shows SpikeFun 1.16 managing to take 100% of all 72 logical Haswell EP cores.

SpikeFun 1.14 - Running on Big Maxwell (NVIDIA GeForce GTX TITAN X)

NVIDIA GeForce GTX TITAN X comes with 12 GB of video memory, making it ideal GPU for neural network simulations, including spiking neural networks.Upcoming version of DigiCortex engine (v1.14) can simulate more than million multi-compartment neurons on a single TITAN X GPU!

SpikeFun 1.10 - Support for Intel Haswell EP/EX

SpikeFun 1.10 release comes with performance monitoring code update adding support for Intel Haswell EP/EX (Xeon E5/E7 v3) CPUs as well as Intel Crystalwell CPUs.

SpikeFun 1.05 - First release of DigiCortex SDK

SpikeFun 1.05 release comes with the first version of DigiCortex SDK. With DigiCortex SDK it is now possible to connect to the running simulation and obtain information about neuron states, inject current into any compartment and get spiking information. Click on the title to read more...

SpikeFun 1.01 – Spontaneous emergence of orientation sensitivity in population of V1 cells with voltage-based Spike Timing Dependent Plasticity (STDP)

DigiCortex Engine can be used to model some of the emergent phenomena of the developing cortex. In this article we show how spontaneous orientation sensitivity of the primary visual cortex (V1) cells develops in presence of natural video signal and voltage-dependent spike-timing dependent plasticity (vSTDP). Click on the title to read more...

DigiCortex 1.01 - Test results running on Intel Xeon 2697 v2 (Ivy Bridge EP)

Since Intel Xeon E5 2600 v2 has been out publically, it is time to show some results of DigiCortex engine on the new dual 12-core setup (24 cores / 48 threads total) based on Intel Xeon 2697 v2 CPU. Result: we clock nearly perfect 1.45x speed-up! Click on the article to read more!

Pages

Subscribe to www.digicortex.net RSS