Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Simulating the Cortical Microcircuit Significantly Faster Than Real Time on the IBM INC-3000 Neural Supercomputer.

Arne Heittmann1, Georgia Psychou1, Guido Trensch2

  • 1JARA-Institute Green IT (PGI-10), Jülich Research Centre, Jülich, Germany.

Frontiers in Neuroscience
|February 7, 2022
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Development and cross-validation of a novel multi-omic assay to assess locoregional recurrence risk and adjuvant therapy benefit in early-stage hormone receptor positive invasive breast cancer patients.

Breast cancer research : BCR·2026
Same author

Accelerating hybrid XOR-CNF Boolean satisfiability problems natively with in-memory computing.

Nature communications·2026
Same author

Building on models-a perspective for computational neuroscience.

Cerebral cortex (New York, N.Y. : 1991)·2025
Same author

Modeling neuron-astrocyte interactions in neural networks using distributed simulation.

PLoS computational biology·2025
Same author

The coming decade of digital brain research: A vision for neuroscience at the intersection of technology and computing.

Imaging neuroscience (Cambridge, Mass.)·2025
Same author

Reconciliation of weak pairwise spike-train correlations and highly coherent local field potentials across space.

Cerebral cortex (New York, N.Y. : 1991)·2024

This study demonstrates the potential of Field-Programmable Gate Array (FPGA) systems for neural modeling by simulating a cortical microcircuit on the IBM INC-3000 supercomputer. The FPGA-based approach achieved significant speed-up compared to biological real time.

Area of Science:

  • Computational neuroscience
  • Hardware acceleration for AI
  • Neuromorphic engineering

Background:

  • Cortical microcircuit models are crucial benchmarks for neural simulation performance.
  • Existing simulation architectures face challenges in achieving biological real-time performance.
  • Field-Programmable Gate Arrays (FPGAs) offer potential for high-performance computing in neuroscience.

Purpose of the Study:

  • To implement and evaluate a widely used cortical microcircuit model on the IBM INC-3000 FPGA-based neural supercomputer.
  • To assess the performance and speed-up factor of FPGA systems for large-scale neural simulations.
  • To demonstrate the flexibility of the FPGA approach with various neuron and synapse models.

Main Methods:

  • Simulation of an 80,000-neuron, 300-million-synapse cortical microcircuit model on the IBM INC-3000 FPGA prototype.
Keywords:
FPGA clusterneuromorphic computingparallel computingperformance benchmarkingprocedural connectivityreconfigurable computingspiking neural networks

Related Experiment Videos

  • Utilizing the programmable logic part of FPGA nodes with single-floating point precision arithmetic.
  • Employing exact exponential integration for linear LIF neurons and current-based synapses, and Runge-Kutta/Parker-Sochacki for non-linear models.

    • Main Results:

    • Achieved a speed-up factor 2.4 times greater than previously reported literature values.
    • Exceeded biological real-time simulation by a factor of four.
    • Demonstrated minimal performance decrease (< few percent) with non-linear neuron models and conductance-based synapses.
    • Identified communication system latency as the primary limitation for further speed-up.

    Conclusions:

    • FPGA-based systems, exemplified by the IBM INC-3000, show significant potential for accelerating neural modeling.
    • The implemented approach offers flexibility for simulating diverse neural network configurations.
    • Further improvements in FPGA communication systems are key to unlocking even greater simulation speeds.