The SpiNNaker Neuromorphic Processor: Architecture, Principles, and Significance

By /
Spinnaker, neuromorphic, processor

1. Introduction 

Neuromorphic processor is a paradigm inspired by the human brain, has got traction as a promising avenue for developing energy-efficient and massively parallel computing architectures. The escalating demand for enhanced computational power coupled with the necessity for sustainable energy consumption in artificial intelligence and large-scale brain simulations has fueled this interest. Traditional von Neumann architectures, with their inherent separation of processing and memory units, face limitations in efficiently handling the parallelism and complex connectivity characteristic of neural networks.

In this context, the Spiking Neural Network Architecture (SpiNNaker) represents a pioneering effort in the field of neuromorphic processors. Developed at the University of Manchester, SpiNNaker is a massively parallel neuromorphic computer architecture distinguished by its scale, with a million-core system, and its primary focus on the real-time simulation of spiking neural networks (SNNs). This innovative architecture breaks away from the deterministic and repeatable communication principles of conventional supercomputers, instead embracing a model where processing elements communicate using simple, inherently unreliable messages known as spikes.

2. Core Operational Principles of the SpiNNaker Neuromorphic Processor

The fundamental principle of operation for the SpiNNaker neuromorphic processor spins around the simulation of neural activity through a vast network of interconnected processing elements. Unlike traditional computers that execute instructions sequentially, SpiNNaker employs a massively parallel approach, where each of its numerous cores independently models the behavior of one or more neurons. These artificial neurons communicate with each other using discrete, asynchronous events known as spikes, mirroring the communication method employed by biological neurons. In this computational model, information is not encoded in the amplitude or shape of these spikes, but rather in their precise timing.

A key capability of the SpiNNaker architecture is its ability to perform simulations in real time. This means that the artificial neural networks running on SpiNNaker can operate at speeds comparable to those observed in biological nervous systems, where time progresses continuously. This capability is crucial for applications such as neurorobotics, where the processing of sensory input and the generation of motor output must occur in a timely manner to interact effectively with the physical world.

Furthermore, the design of SpiNNaker emphasizes energy efficiency, a hallmark of biological brains. By utilizing an event-driven communication paradigm, where processing and communication are triggered only by the occurrence of spikes, SpiNNaker minimizes power consumption. Only the active neurons and synapses involved in processing a spike consume energy, while the rest of the system remains idle. This approach contrasts sharply with traditional, synchronous computing systems where power is consumed regardless of the level of activity. The core design of SpiNNaker, therefore, prioritizes mimicking the brain’s neural signaling through spikes, leading to both biological plausibility and computational efficiency

3. Internal Architecture

  • 3.1. The SpiNNaker Chip: ARM9 Cores and their Functionality

The fundamental building block of the SpiNNaker system is the SpiNNaker chip, a complex System-on-Chip (SoC) designed with a Globally Asynchronous Locally Synchronous (GALS) architecture. Each chip originally housed 18 ARM9 processor cores, each operating at a clock speed of 200 MHz. These cores serve as the computational units responsible for simulating the dynamics of individual neurons. One of these cores on each chip is typically designated as a monitor core, tasked with performing system management functions, while the remaining cores are available for running the user’s neural network applications. The GALS nature of the chip means that the individual processing cores function as synchronous islands, operating with their own local clocks, while the communication between these islands and across the chip is handled asynchronously. This design choice, utilizing energy-efficient ARM cores, reflects a deliberate trade-off where some performance is sacrificed to achieve significantly enhanced power efficiency, aligning with the project’s overarching goals. The subsequent generation, SpiNNaker2, features a significant increase in the number of cores per chip, incorporating 153 ARM cores, further enhancing the platform’s parallel processing capabilities.

  • 3.2. Memory Hierarchy: Local Tightly-Coupled Memory and Shared SDRAM

The SpiNNaker chip employs a two-level memory hierarchy to manage the storage and retrieval of data required for neural simulations. Each of the ARM9 cores is equipped with a small but fast local memory, known as Tightly-Coupled Memory (TCM). This TCM is further divided into 32KB for storing instructions and 64KB for data. The neuron states, which represent the dynamic properties of the simulated neurons (such as membrane potential), are typically stored in this fast local memory to enable rapid access during the simulation process.

In addition to the local TCM, each SpiNNaker chip is paired with a larger, shared off-chip SDRAM (Synchronous Dynamic Random Access Memory) with a capacity of 128MB. The synaptic parameters, which define the strength and other characteristics of the connections between neurons, are primarily stored in this shared SDRAM. Given the vast number of synapses in even moderately sized neural networks, the larger capacity of the SDRAM is essential for accommodating this data. To facilitate efficient data transfer between the off-chip SDRAM and the local TCM of the processing cores, the SpiNNaker architecture incorporates Direct Memory Access (DMA) controllers. These DMA units allow the cores to request and receive blocks of synaptic data from the SDRAM directly into their local memory without requiring constant intervention from the processor itself, thereby optimizing the simulation performance. This memory hierarchy, with its balance between fast local memory for active computations and larger shared memory for storing network parameters, is carefully designed to meet the specific demands of simulating large-scale spiking neural networks within the constraints of the hardware.

  • 3.3. The Asynchronous Packet Routing System and Inter-Chip Communication

A defining feature of the SpiNNaker architecture is its custom-designed packet routing system, an intricate communication fabric inspired by the highly interconnected nature of the brain. This system facilitates the exchange of small packets of data, representing neural spikes, between the numerous processing cores both within a single chip and across multiple interconnected chips. The communication relies on the principle of Address Event Representation (AER), where the information conveyed is primarily the identity (address) of the neuron that has fired a spike, with the timing being implicit in the arrival of the packet. A key aspect of SpiNNaker’s communication infrastructure is its efficient multicast routing mechanism. This allows a single spike originating from one core to be simultaneously delivered to multiple destination cores, which is crucial for emulating the high fan-out connectivity observed in biological neural networks. Each SpiNNaker chip is equipped with six bidirectional links that enable it to connect to its neighboring chips, typically arranged in a 2D or 3D torus topology, forming a large-scale, interconnected system. This sophisticated communication infrastructure allows the SpiNNaker machine to achieve a very high bisection bandwidth, enabling a large volume of spike traffic to be routed efficiently across the system.

The routing of these spike packets is managed by routing tables stored within the router of each SpiNNaker chip. These tables contain entries that map the address of a spiking neuron to a set of destination cores or neighboring chips. When a spike packet arrives at a router, the router consults its routing table to determine where to forward the packet. If no specific route is found for a given spike, a default routing mechanism is employed, typically sending the packet out of the link opposite to the one it was received on. The SpiNNaker routing system also incorporates features for fault tolerance, allowing the system to continue operating even in the presence of some chip or link failures. The asynchronous nature of spike transmission in SpiNNaker contributes significantly to its efficiency, as communication only occurs when neurons fire, minimizing unnecessary data transfer and power consumption.

4. Implementation of Spiking Neural Networks on SpiNNaker

  • 4.1. Supported Neuron Models for Biological Realism

The SpiNNaker neuromorphic processor distinguishes itself through the programmability of its ARM cores, which allows researchers to implement a wide variety of neuron models in software. This flexibility is crucial for achieving biological realism in simulations, as different types of neurons in the brain exhibit diverse firing patterns and dynamics. Among the commonly implemented models is the Leaky Integrate-and-Fire (LIF) model, a computationally efficient model that captures the basic behavior of a neuron integrating incoming signals and firing a spike when its membrane potential reaches a threshold.

For more biologically accurate simulations, researchers can also implement more complex neuron models on SpiNNaker. These include conductance-based models, which simulate the flow of ions through specific ion channels in the neuron’s membrane, the Izhikevich model, known for its ability to reproduce a wide range of neuronal firing patterns with relatively low computational cost, and the Adaptive Exponential Integrate-and-Fire (AdEx) model, which incorporates mechanisms for spike frequency adaptation. Implementing these more detailed models often involves a trade-off with computational cost and the number of neurons that can be simulated in real time. However, the ongoing advancements in the SpiNNaker architecture, such as the increased processing power in SpiNNaker2, are enabling researchers to explore more complex neuron models with greater efficiency.

  • 4.2. Mechanisms for Synaptic Plasticity and Learning

Synaptic plasticity, the capacity of synapses to modify their strength over time in response to neural activity, is believed to be a fundamental mechanism for learning and memory in the brain. SpiNNaker provides mechanisms to implement various forms of synaptic plasticity, allowing researchers to investigate these crucial processes in their neural network simulations. In the original SpiNNaker architecture, a dedicated plasticity core approach was introduced, where a subset of the ARM cores is exclusively dedicated to processing synaptic plasticity updates, operating in the background while other cores handle the neural and synaptic simulations. This separation allows for efficient management of the computational resources required for both neural dynamics and weight modifications.

SpiNNaker supports a range of spike-timing dependent plasticity (STDP) rules, where the strength of a synapse is adjusted based on the relative timing of the pre- and post-synaptic spikes. These rules, which can lead to either long-term potentiation (strengthening) or long-term depression (weakening) of synaptic connections, are crucial for learning temporal sequences and forming associations in neural networks. Due to the programmable nature of the ARM cores, SpiNNaker offers the flexibility to implement custom learning rules beyond the standard STDP variations, enabling the exploration of novel learning paradigms. Furthermore, the development of SpiNNaker2 has brought advancements in on-chip learning capabilities, suggesting a continued focus on enabling more sophisticated and biologically plausible learning mechanisms within the architecture. Implementing these plasticity mechanisms on the large scale offered by SpiNNaker presents challenges in terms of managing computational resources and ensuring efficient access to synaptic weights stored in memory. However, the ongoing research and development efforts continue to push the boundaries of what is possible in simulating learning and adaptive behavior on neuromorphic hardware.

  • 4.3. Leveraging Event-Driven Communication for Efficiency

A cornerstone of SpiNNaker’s architecture and its ability to efficiently implement spiking neural networks is its reliance on asynchronous, event-driven communication between processing elements. As previously discussed, neurons on SpiNNaker communicate through the transmission of small packets representing spikes. This sparse communication paradigm is highly energy-efficient, as power is primarily consumed only when a neuron fires and transmits a spike. This event-driven approach also minimizes the bandwidth required for communication, as information is only exchanged when there is a change in neural activity.

The use of event-driven processing on SpiNNaker directly contributes to its real-time performance capabilities. The SpiNNaker router is specifically optimized for handling a high volume of these small packets (spikes) with minimal latency. This approach, inspired by the brain’s own communication methods, is a key factor in SpiNNaker’s ability to simulate complex neural networks efficiently.

5. The Genesis of SpiNNaker: History of Development

  • 5.1. Key Individuals and Institutions Involved

The SpiNNaker project is primarily the brainchild of the Advanced Processor Technologies (APT) Research Group within the Department of Computer Science at the University of Manchester. The leading figure behind this ambitious endeavor is Professor Steve Furber, a renowned computer engineer who conceived the initial architectural innovations required for real-time brain modeling between 2000 and 2005. The project has also involved collaborations with other academic institutions and industrial partners, including the University of Southampton and ARM Ltd., which contributed to various aspects of the hardware and software development. Dr. Andrew Rowley has played a crucial role in leading the software development efforts for SpiNNaker since 2014, focusing on creating tools that enable the compilation and execution of large neural networks on the million-core machine. Furthermore, Dr. Gengting Liu has been instrumental in the hardware design and low-level software system of the second-generation SpiNNaker machine (SpiNNaker2). The sustained involvement of these key individuals underscores the long-term commitment and deep expertise driving the SpiNNaker project. The collaborative nature of the project highlights the interdisciplinary approach required to tackle the complex challenges of neuromorphic computing. The University of Manchester’s rich history in computer science, marked by pioneering work from figures like Alan Turing, provides a strong foundation and context for the ambitious goals of the SpiNNaker project.

  • 5.2. Timeline of Key Milestones and Achievements

The SpiNNaker project officially began in 2005, although its conceptualization dates back to the late 1990s. The development of the first prototype SpiNNaker chips was a significant milestone, with the initial test chip being delivered in December 2009. The first full SpiNNaker chip with 18 processors was delivered in 2011, marking a significant step towards the project’s goals. After years of intensive development and construction, the landmark achievement of creating a one-million-core SpiNNaker machine was reached in 2018. This marked a significant milestone in the project’s 20-year journey from conception to realization. The development of the second-generation SpiNNaker machine, SpiNNaker2, began in 2013 in collaboration with the Technische Universität Dresden, with the aim of achieving a tenfold increase in computing performance while maintaining similar power consumption. The first SpiNNaker2 silicon became available in late 2022, with larger boards becoming available subsequently. The construction of a large-scale SpiNNaker2 system in Dresden, known as SpiNNcloud, is underway and expected to be available for cloud access. These milestones highlight the continuous progress and evolution of the SpiNNaker platform, driven by the dedication of researchers at the University of Manchester and their collaborators.

6. Unveiling the Motivations and Design Philosophy Behind SpiNNaker

  • 6.1. Addressing the Challenges of Brain Simulation

The primary motivation behind the SpiNNaker project was to create a computing platform specifically designed to tackle the immense challenges of simulating large-scale neural networks with a level of detail and speed that could contribute to our understanding of how the brain functions. The sheer scale and complexity of the mammalian nervous system, with its billions of neurons and trillions of synapses, present a formidable computational challenge. Researchers aimed to create a platform capable of modeling a significant portion of the human brain, such as 1%, or an entire mouse brain, in biological real time. This ambition required a departure from conventional supercomputing architectures, which often struggle with the unique communication demands of neural networks, where individual neurons connect to thousands of others.

  • 6.2. Emphasis on Parallelism, Real-time Operation, and Energy Efficiency

The design philosophy of SpiNNaker is heavily influenced by the fundamental characteristics of biological brains, particularly their massive parallelism, real-time operation, and remarkable energy efficiency. The architecture is designed to support a massive number of processing cores, enabling the parallel execution of neural simulations. The ability to simulate neural networks in real time is another crucial design goal, enabling the integration of SpiNNaker with robotic systems and other real-world applications. Furthermore, energy efficiency was a critical consideration from the outset, aiming to replicate the brain’s remarkable ability to perform complex computations with very low power consumption. The design choices in SpiNNaker, such as the use of low-power ARM cores and an event-driven communication system, reflect this emphasis on minimizing energy usage. The project’s focus on these three key aspects – parallelism, real-time operation, and energy efficiency – distinguishes it from traditional high-performance computing systems and underscores its ambition to provide a truly brain-inspired computing platform.

Sources

  1. SpiNNaker – Human Brain Project, fecha de acceso: abril 8, 2025, https://www.humanbrainproject.eu/en/collaborate-hbp/innovation-industry/technology-catalogue/spinnaker/
  2. Neuromorphic Engineering: Developing Brain-Inspired Machines …, fecha de acceso: abril 8, 2025, https://viso.ai/deep-learning/neuromorphic-engineering/
  3. Neuromorphic Computing. An overview of the history of… | by QuAIL Technologies – Medium, fecha de acceso: abril 8, 2025, https://medium.com/quail-technologies/day-18-neuromorphic-computing-752e6020332d
  4. SpiNNaker Home Page – Research Groups: APT – Advanced Processor Technologies (School of Computer Science – The University of Manchester), fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/
  5. (PDF) SpiNNaker: A Spiking Neural Network Architecture – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/publication/340279616_SpiNNaker_A_Spiking_Neural_Network_Architecture
  6. Overview of the SpiNNaker system architecture – ePrints Soton – University of Southampton, fecha de acceso: abril 8, 2025, https://eprints.soton.ac.uk/350495/1/TCv2.pdf
  7. SpiNNTools: The Execution Engine for the SpiNNaker Platform – Frontiers, fecha de acceso: abril 8, 2025, https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2019.00231/full
  8. SpiNNaker – Wikipedia, fecha de acceso: abril 8, 2025, https://en.wikipedia.org/wiki/SpiNNaker
  9. (PDF) The SpiNNaker project – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/publication/262025847_The_SpiNNaker_project
  10. Implementation of an unsupervised fully-connected spiking neural network on SpiNNaker for pattern classification – Webthesis – Politecnico di Torino, fecha de acceso: abril 8, 2025, https://webthesis.biblio.polito.it/25523/
  11. SPINNAKER – OAPEN Library, fecha de acceso: abril 8, 2025, https://library.oapen.org/bitstream/handle/20.500.12657/47874/9781680836530.pdf?sequence=1&isAllowed=y
  12. ‘Human brain’ supercomputer with 1 million processors switched on for first time – The University of Manchester, fecha de acceso: abril 8, 2025, https://www.manchester.ac.uk/about/news/human-brain-supercomputer-with-1million-processors-switched-on-for-first-time/
  13. SpiNNaker – The University of Manchester, fecha de acceso: abril 8, 2025, https://www.scieng.manchester.ac.uk/tomorrowlabs/spinnaker/
  14. Neuromorphic Sentiment Analysis Using Spiking Neural Networks – PubMed, fecha de acceso: abril 8, 2025, https://pubmed.ncbi.nlm.nih.gov/37765758/
  15. Performance Comparison of the Digital Neuromorphic Hardware SpiNNaker and the Neural Network Simulation Software NEST for a Full-Scale Cortical Microcircuit Model – PMC, fecha de acceso: abril 8, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC5974216/
  16. The SpiNNaker Supercomputer, Modeled After the Human Brain, Is Up and Running, fecha de acceso: abril 8, 2025, https://singularityhub.com/2018/11/19/the-million-core-spinnaker-supercomputer-is-up-and-running/
  17. A Look at SpiNNaker 2 – University of Dresden – Neuromorphic Chip, fecha de acceso: abril 8, 2025, https://open-neuromorphic.org/neuromorphic-computing/hardware/spinnaker-2-university-of-dresden/
  18. Spinnaker Real Estate Partners: Home, fecha de acceso: abril 8, 2025, https://spinrep.com/
  19. SpiNNaker Project – Research Groups: APT – Advanced Processor Technologies (School of Computer Science – The University of Manchester), fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/project/
  20. Neuromorphic Computing – Human Brain Project, fecha de acceso: abril 8, 2025, https://www.humanbrainproject.eu/en/science-development/focus-areas/neuromorphic-computing/
  21. SpiNNaker Project – Boards and Machines. – Advanced Processor Technologies (APT) – The University of Manchester, fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/hardware/
  22. [2103.08392] The SpiNNaker 2 Processing Element Architecture for Hybrid Digital Neuromorphic Computing – arXiv, fecha de acceso: abril 8, 2025, https://arxiv.org/abs/2103.08392
  23. SpiNNaker2 Chip Topology, fecha de acceso: abril 8, 2025, https://spinnaker2.gitlab.io/getting-started/2-s2-chip-topology/
  24. Evaluation of a Large-Scale SpiNNaker System – UPV/EHU, fecha de acceso: abril 8, 2025, http://www.sc.ehu.es/ccwbayes/members/jnavaridas/pubs/jor08.pdf
  25. SpiNNaker2 – SpiNNcloud Systems, fecha de acceso: abril 8, 2025, https://spinncloud.com/portfolio/spinnaker2/
  26. Low-Power Deep Learning Inference using the SpiNNaker Neuromorphic Platform – OSTI.GOV, fecha de acceso: abril 8, 2025, https://www.osti.gov/servlets/purl/1761866
  27. Neuro-Evolution of Spiking Neural Networks on SpiNNaker Neuromorphic Hardware, fecha de acceso: abril 8, 2025, http://vigir.missouri.edu/~gdesouza/Research/Conference_CDs/IEEE_SSCI_2016/pdf/SSCI16_paper_607.pdf
  28. SpiNNaker: A Spiking Neural Network Architecture – Now Publishers, fecha de acceso: abril 8, 2025, https://www.nowpublishers.com/article/BookDetails/9781680836523
  29. Structural Plasticity on the SpiNNaker Many-Core Neuromorphic System – Frontiers, fecha de acceso: abril 8, 2025, https://www.frontiersin.org/articles/10.3389/fnins.2018.00434/full
  30. SpiNNaker Event-driven Neural Application Model. – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/SpiNNaker-Event-driven-Neural-Application-Model_fig10_220277134
  31. Overview of the SpiNNaker system architecture – YouTube, fecha de acceso: abril 8, 2025, https://www.youtube.com/watch?v=EhPpxsK2Ia0
  32. SpiNNaker – Science and Industry Museum blog -, fecha de acceso: abril 8, 2025, https://blog.scienceandindustrymuseum.org.uk/spinnaker/
  33. 1. Origins – Now Publishers, fecha de acceso: abril 8, 2025, https://nowpublishers.com/article/Chapter/9781680836523?cId=978-1-68083-653-0.ch1
  34. Spinnaker Residence – Dubai – Propsearch.ae, fecha de acceso: abril 8, 2025, https://propsearch.ae/dubai/spinnaker-residence
  35. SpiNNaker brain simulation project hits one million cores on a single machine – DCD, fecha de acceso: abril 8, 2025, https://www.datacenterdynamics.com/en/news/spinnaker-brain-simulation-project-hits-one-million-cores-single-machine/
  36. Carver Mead: Microelectronics, neuromorphic computing, and life at the frontiers of science and technology – SPIE, fecha de acceso: abril 8, 2025, https://spie.org/news/photonics-focus/septoct-2024/inventing-the-integrated-circuit
  37. Neuromorphic computing – Wikipedia, fecha de acceso: abril 8, 2025, https://en.wikipedia.org/wiki/Neuromorphic_computing
  38. August 1, 2020 | Neuromorphic Chips Take Shape – California NanoSystems Institute, fecha de acceso: abril 8, 2025, https://cnsi.ucla.edu/august-1-2020-neuromorphic-chips-take-shape/
  39. IBM’s Incredible TrueNorth Chip || Neuromorphic Computing – YouTube, fecha de acceso: abril 8, 2025, https://www.youtube.com/watch?v=X2TYAcr36r0
  40. TrueNorth: A Deep Dive into IBM’s Neuromorphic Chip Design, fecha de acceso: abril 8, 2025, https://open-neuromorphic.org/blog/truenorth-deep-dive-ibm-neuromorphic-chip-design/
  41. TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip, fecha de acceso: abril 8, 2025, https://redwood.berkeley.edu/wp-content/uploads/2021/08/Akopyan2015.pdf
  42. TrueNorth Architecture IBM’s Neuromorphic Chip – Janathjsk – Medium, fecha de acceso: abril 8, 2025, https://janathjsk.medium.com/truenorth-architecture-ibms-neuromorphic-chip-63cbfec42b98
  43. A Look at Loihi – Intel – Neuromorphic Chip, fecha de acceso: abril 8, 2025, https://open-neuromorphic.org/neuromorphic-computing/hardware/loihi-intel/
  44. Intel Loihi: A Neuromorphic Breakthrough | by Prayag Mohanty – Medium, fecha de acceso: abril 8, 2025, https://medium.com/@topper1728/intel-loihi-a-neuromorphic-experiment-4e08bb901dae
  45. Neuromorphic Computing and Engineering with AI | Intel®, fecha de acceso: abril 8, 2025, https://www.intel.com/content/www/us/en/research/neuromorphic-computing.html
  46. SpiNNaker Jupyter service (7.2.1) – EBRAINS, fecha de acceso: abril 8, 2025, https://search.kg.ebrains.eu/instances/0da1752b-1c42-4d96-847e-10c4daf14076?noSilentSSO=true
  47. Spinnaker chip organization. (NoC: network on a chip.) – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/Spinnaker-chip-organization-NoC-network-on-a-chip_fig6_220651350
  48. SpiNNaker: Mapping Neural Networks onto a Massively-Parallel Chip Multiprocessor – Advanced Processor Technologies (APT), fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/ftp/pub/apt/papers/MMK_IJCNN08.pdf
  49. What Is Neuromorphic Computing and How Could It Impact Enterprise IT? – Turn-key Technologies, fecha de acceso: abril 8, 2025, https://www.turn-keytechnologies.com/blog/network-solutions/what-is-neuromorphic-computing-how-it-impacts-enterprise-it
  50. Exploring Neuromorphic Computing for AI: Why Spikes? (Part One), fecha de acceso: abril 8, 2025, https://www.edge-ai-vision.com/2019/07/exploring-neuromorphic-computing-for-ai-why-spikes-part-one/
  51. Intel Ramps Up Neuromorphic Computing Effort with New Research Partners – TOP500, fecha de acceso: abril 8, 2025, https://www.top500.org/news/intel-ramps-up-neuromorphic-computing-effort-with-new-research-partners/
  52. Intel Debuts Pohoiki Beach, Its 8M Neuron Neuromorphic Development System – HPCwire, fecha de acceso: abril 8, 2025, https://www.hpcwire.com/2019/07/17/intel-debuts-pohoiki-beach-its-8m-neuron-neuromorphic-development-system/
  53. Intel Scales Neuromorphic Research System to 100 Million Neurons, fecha de acceso: abril 8, 2025, https://www.intc.com/news-events/press-releases/detail/12/intel-scales-neuromorphic-research-system-to-100-million
  54. Intel Rolls Out New Loihi 2 Neuromorphic Chip: Built on Early Intel 4 Process – AnandTech, fecha de acceso: abril 8, 2025, https://www.anandtech.com/show/16960/intel-loihi-2-intel-4nm-4
  55. Intel’s Loihi ‘brain chips’ now available to researchers – TechCentral.ie, fecha de acceso: abril 8, 2025, https://www.techcentral.ie/intels-loihi-brain-chips-now-available-to-researchers/
  56. Intel Editorial: Intel’s New Self-Learning Chip Promises to Accelerate Artificial Intelligence, fecha de acceso: abril 8, 2025, https://www.intc.com/news-events/press-releases/detail/202/intel-editorial-intels-new-self-learning-chip-promises
  57. Intel Announces Neuromorphic Research Progress, fecha de acceso: abril 8, 2025, https://download.intel.com/newsroom/2021/archive/2018-12-06-news-intel-announces-neuromorphic-computing-research-collaborators.pdf
  58. Update on Intel’s Neuromorphic Ecosystem Growth and Progress, fecha de acceso: abril 8, 2025, https://www.intel.com/content/www/us/en/newsroom/news/neuromorphic-ecosystem-growth-progress.html
  59. Intel Announces Neuromorphic Computing Research Collaborators – UTK-EECS, fecha de acceso: abril 8, 2025, https://www.eecs.utk.edu/intel-announces-neuromorphic-computing-research-collaborators/
  60. SpiNNaker, the Million-Core Supercomputer, Finally Switched On – News – All About Circuits, fecha de acceso: abril 8, 2025, https://www.allaboutcircuits.com/news/simulate-human-brain-spinnaker-million-core-computer-switched-on/
  61. SpiNNaker: A Supercomputer Based on the Human Brain – YouTube, fecha de acceso: abril 8, 2025, https://www.youtube.com/watch?v=8LSWp5CQgjY
  62. Supercomputer ‘SpiNNaker’ Designed To Work Like Human Brain Switched On For The First Time – Swarajya, fecha de acceso: abril 8, 2025, https://swarajyamag.com/insta/supercomputer-spinnaker-designed-to-work-like-human-brain-switched-on-for-the-first-time
  63. Neuromorphic Computing and the Human Brain Project: An Interview with Prof. Steve Furber on SpiNNaker and Cross-Disciplinary Brain Research, fecha de acceso: abril 8, 2025, https://www.humanbrainproject.eu/en/follow-hbp/news/2023/09/01/neuromorphic-computing-and-human-brain-project-interview-prof-steve-furber-spinnaker-and-cross-disciplinary-brain-research/
  64. SpiNNaker: Next-level thinking – ARM Community Forums, fecha de acceso: abril 8, 2025, https://community.arm.com/arm-research/b/articles/posts/spinnaker-next-level-thinking
  65. What Is Neuromorphic Computing? – IBM, fecha de acceso: abril 8, 2025, https://www.ibm.com/think/topics/neuromorphic-computing
  66. MODELING SPIKING NEURAL NETWORKS ON SPINNAKER – Advanced Processor Technologies (APT), fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/ftp/pub/apt/papers/XJIN_CISE_10.pdf
  67. Home – Modelling the Human Brain with specialised CPUs – Gazettabyte, fecha de acceso: abril 8, 2025, https://www.gazettabyte.com/home/2023/7/30/modelling-the-human-brain-with-specialised-cpus.html
  68. SpiNNaker chip – Advanced Processor Technologies (APT) – The University of Manchester, fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/SpiNNchip/
  69. SpiNNaker – APT – The University of Manchester – WikiChip, fecha de acceso: abril 8, 2025, https://en.wikichip.org/wiki/university_of_manchester/spinnaker
  70. SpiNNaker datasheet version 2.02 6 January 2011, fecha de acceso: abril 8, 2025, https://spinnakermanchester.github.io/docs/SpiNN2DataShtV202.pdf
  71. (PDF) SpiNNaker: A 1-W 18-Core System-on-Chip for Massively-Parallel Neural Network Simulation – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/publication/260648207_SpiNNaker_A_1-W_18-Core_System-on-Chip_for_Massively-Parallel_Neural_Network_Simulation
  72. Understanding the Interconnection Network of SpiNNaker – Advanced Processor Technologies (APT), fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/ftp/pub/apt/papers/JavN_ICS09.pdf
  73. The SpiNNaker 2 Processing Element Architecture for Hybrid Digital Neuromorphic Computing – arXiv, fecha de acceso: abril 8, 2025, https://arxiv.org/pdf/2103.08392
  74. SpiNNaker2: A Large-Scale Neuromorphic System for Event-Based and Asynchronous Machine Learning – arXiv, fecha de acceso: abril 8, 2025, https://arxiv.org/html/2401.04491v1
  75. SpiNNaker 2: A 10 Million Core Processor System for Brain Simulation and Machine Learning – arXiv, fecha de acceso: abril 8, 2025, https://arxiv.org/pdf/1911.02385
  76. (PDF) Overview of the SpiNNaker System Architecture – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/publication/260585643_Overview_of_the_SpiNNaker_System_Architecture
  77. Synapse-Centric Mapping of Cortical Models to the SpiNNaker Neuromorphic Architecture, fecha de acceso: abril 8, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC5022244/
  78. A framework for plasticity implementation on the SpiNNaker neural architecture – PMC, fecha de acceso: abril 8, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC4299433/
  79. Breaking the millisecond barrier on SpiNNaker: implementing asynchronous event-based plastic models with microsecond resolution – Frontiers, fecha de acceso: abril 8, 2025, https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2015.00206/full
  80. SpiNNaker Project – System Software. – Advanced Processor Technologies (APT) – The University of Manchester, fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/software/
  81. SpiNNaker: Event-based simulation – quantitative behaviour – ePrints Soton – University of Southampton, fecha de acceso: abril 8, 2025, https://eprints.soton.ac.uk/413671/1/TMSCS_2017_05_0030_main.pdf
  82. Scalable Event-Driven Native Parallel Processing: The SpiNNaker Neuromimetic System – Advanced Processor Technologies (APT), fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/ftp/pub/apt/papers/Rast_ACMCF10.pdf
  83. SpiNNTools: The Execution Engine for the SpiNNaker Platform – PMC, fecha de acceso: abril 8, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC6444189/
  84. SpiNNaker: Mapping Neural Networks onto a Massively-Parallel Chip Multiprocessor, fecha de acceso: abril 8, 2025, https://www.researchgate.net/publication/221531727_SpiNNaker_Mapping_Neural_Networks_onto_a_Massively-Parallel_Chip_Multiprocessor
  85. Breaking the millisecond barrier on SpiNNaker: implementing asynchronous event-based plastic models with microsecond resolution – PubMed Central, fecha de acceso: abril 8, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC4458614/
  86. Building a Spiking Neural Network Model of the Basal Ganglia on SpiNNaker – White Rose Research Online, fecha de acceso: abril 8, 2025, https://eprints.whiterose.ac.uk/129097/1/Basabdatta_etal_AuthorProof.pdf
  87. Real-time cortical simulation on neuromorphic hardware – PMC, fecha de acceso: abril 8, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC6939236/
  88. E-prop on SpiNNaker 2: Exploring online learning in spiking RNNs on neuromorphic hardware – Frontiers, fecha de acceso: abril 8, 2025, https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2022.1018006/full
  89. Privacy Policy – Spinnaker Projects, fecha de acceso: abril 8, 2025, https://spinnakerprojects.com/privacy-policy/
  90. SpiNNaker – Documentation for HPC – GWDG, fecha de acceso: abril 8, 2025, https://docs.hpc.gwdg.de/services/neuromorphic-computing/spinnaker/index.html
  91. Memory-Efficient Deep Learning on a SpiNNaker 2 Prototype – Frontiers, fecha de acceso: abril 8, 2025, https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2018.00840/full
  92. Event-based backpropagation on the neuromorphic platform SpiNNaker2 – arXiv, fecha de acceso: abril 8, 2025, https://arxiv.org/html/2412.15021v1
  93. ARM Inventor Steve Furber on SpiNNaker 1, 2 and Beyond – EE Times Podcast, fecha de acceso: abril 8, 2025, https://www.eetimes.com/podcasts/arm-inventor-steve-furber-on-spinnaker-1-2-and-beyond/
  94. History – Spinnaker Global, fecha de acceso: abril 8, 2025, https://spinnaker-global.com/about/history/
  95. governance/rfc/eol_kubernetes_v1.md at master · spinnaker/governance – GitHub, fecha de acceso: abril 8, 2025, https://github.com/spinnaker/governance/blob/master/rfc/eol_kubernetes_v1.md
  96. Carver Mead – Lemelson-MIT Program, fecha de acceso: abril 8, 2025, https://lemelson.mit.edu/resources/carver-mead
  97. Neuromorphic electronic systems – Proceedings of the IEEE, fecha de acceso: abril 8, 2025, https://hasler.ece.gatech.edu/Published_papers/Technology_overview/MeadNeuro1990.pdf
  98. TrueNorth is a brain-inspired chip architecture built from an… | Download Scientific Diagram – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/TrueNorth-is-a-brain-inspired-chip-architecture-built-from-an-interconnected-network-of_fig4_260869530
  99. Loihi – Intel – WikiChip, fecha de acceso: abril 8, 2025, https://en.wikichip.org/wiki/intel/loihi
  100. Simplified Diagram of Loihi 2 Neurocore Architecture. – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/Simplified-Diagram-of-Loihi-2-Neurocore-Architecture_fig1_372584421
  101. SpiNNaker Publications SpiNNaker Publications – Advanced Processor Technologies (APT) – The University of Manchester, fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/Publications/
  102. TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip – IBM Research, fecha de acceso: abril 8, 2025, https://research.ibm.com/publications/truenorth-design-and-tool-flow-of-a-65-mw-1-million-neuron-programmable-neurosynaptic-chip?utm_source=thegpu.ai&utm_medium=referral&utm_campaign=issue-8-the-ultimate-cheat-sheet-for-ai-processors
  103. Large-Scale Simulations of Plastic Neural Networks on Neuromorphic Hardware – PMC, fecha de acceso: abril 8, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC4823276/
  104. Neuromorphic Embedded Systems – TUM Neuroscientific System Theory (NST), fecha de acceso: abril 8, 2025, https://tum.neurocomputing.systems/en/research/neuromorphic-systems/index.html
  105. Neuromorphic Computing for Accelerating AI Models: SpiNNaker and Loihi Platforms, fecha de acceso: abril 8, 2025, https://www.cmc.ca/wp-content/uploads/2020/03/Idir-Mellal-University-of-Toronto.pdf
  106. Steve Furber – CHM – Computer History Museum, fecha de acceso: abril 8, 2025, https://computerhistory.org/profile/steve-furber/
  107. Research IT and SpiNNaker – The University of Manchester, fecha de acceso: abril 8, 2025, https://research-it.manchester.ac.uk/news/2020/04/28/research-it-and-spinnaker/
  108. Neuromorphic Hardware Guide, fecha de acceso: abril 8, 2025, https://open-neuromorphic.org/neuromorphic-computing/hardware/
  109. SpiNNaker-Based Supercomputer Launches in Dresden – EE Times, fecha de acceso: abril 8, 2025, https://www.eetimes.com/spinnaker-based-neuromorphic-supercomputer-launches-in-dresden/
  110. Research Groups: APT – Advanced Processor Technologies (School …, fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/project/Access/
  111. SpiNNaker – Tools – EBRAINS, fecha de acceso: abril 8, 2025, https://www.ebrains.eu/tools/spinnaker
  112. Carver Mead Earns Lifetime Contribution Award for Neuromorphic Engineering – Caltech, fecha de acceso: abril 8, 2025, https://www.caltech.edu/about/news/carver-mead-earns-lifetime-contribution-award-for-neuromorphic-engineering
  113. Loihi 2 – Intel – WikiChip, fecha de acceso: abril 8, 2025, https://en.wikichip.org/wiki/intel/loihi_2
  114. Intel Builds World’s Largest Neuromorphic System to Enable More Sustainable AI, fecha de acceso: abril 8, 2025, https://newsroom.intel.com/artificial-intelligence/intel-builds-worlds-largest-neuromorphic-system-to-enable-more-sustainable-ai
  115. SpiNNaker Neural Network Architecture | Ontosight – AI Research Assistant, fecha de acceso: abril 8, 2025, https://ontosight.ai/glossary/term/spinnaker-neural-network-architecture–67a142766c3593987a5519f4
  116. Getting started with spiNNaker – Google Groups, fecha de acceso: abril 8, 2025, https://groups.google.com/g/spinnakerusers/c/11AaZitvDQU
  117. Low-Power Deep Learning Inference using the SpiNNaker Neuromorphic Platform – OSTI.GOV, fecha de acceso: abril 8, 2025, https://www.osti.gov/servlets/purl/1641484
  118. Documentation – Spinnaker, fecha de acceso: abril 8, 2025, https://spinnaker.io/docs/
  119. Andrew Rowley: Using the SpiNNaker neuromorphic compute system interactively, fecha de acceso: abril 8, 2025, https://www.youtube.com/watch?v=xtsaH40MrLQ
  120. try.spinnaker.io, fecha de acceso: abril 8, 2025, https://spinnaker.io/docs/community/gsoc/projects/2021/try-spinnaker-io/
  121. Spinnaker Workshop Part Four – AWS Quick Start – Thomas McGonagle, Armory – YouTube, fecha de acceso: abril 8, 2025, https://www.youtube.com/watch?v=QFZPCHHKFTU
  122. I just downloaded Spinnaker, how do I start using it? – Apliter Termografia, fecha de acceso: abril 8, 2025, https://www.apliter.com/en/blog/ufaq/acabo-de-descargar-spinnaker-como-empiezo-a-utilizarlo/
  123. Getting Started – Spinnaker, fecha de acceso: abril 8, 2025, https://spinnaker.io/docs/guides/user/managed-delivery/getting-started/
  124. Get Started Using Spinnaker, fecha de acceso: abril 8, 2025, https://spinnaker.io/docs/guides/user/get-started/
  125. SpiNNaker Tutorial, fecha de acceso: abril 8, 2025, https://spinnaker-tutorial.readthedocs.io/
  126. Hands-on SpiNNaker tutorial at the European Institute for Neuromorphic Computing, fecha de acceso: abril 8, 2025, https://www.ebrains.eu/news-and-events/hands-on-spinnaker-tutorial-at-the-european-institute-for-neuromorphic-computing
  127. Getting started with Armory Spinnaker | by DataCouch – Medium, fecha de acceso: abril 8, 2025, https://datacouch.medium.com/getting-started-with-armory-spinnaker-1421c1749a38
  128. Spinnaker Version 1.0 Theory Guide | ORNL, fecha de acceso: abril 8, 2025, https://www.ornl.gov/publication/spinnaker-version-10-theory-guide
  129. SpiNNaker Application Notes – Advanced Processor Technologies (APT) – The University of Manchester, fecha de acceso: abril 8, 2025, https://apt.cs.manchester.ac.uk/projects/SpiNNaker/support/
  130. Articles and Interviews – Carver Mead – Caltech, fecha de acceso: abril 8, 2025, http://www.carvermead.caltech.edu/articles.html
  131. Carver Mead Writes the First Book on Neuromorphic Computing – History of Information, fecha de acceso: abril 8, 2025, https://www.historyofinformation.com/detail.php?id=3900
  132. Carver Mead on Quantum Computing and Neuromorphic Design – HPCwire, fecha de acceso: abril 8, 2025, https://www.hpcwire.com/2013/11/25/carver-mead-quantum-computing-neuromorphic-design/
  133. The creation of the electronic brain – DCD – Data Center Dynamics, fecha de acceso: abril 8, 2025, https://www.datacenterdynamics.com/en/analysis/creation-of-the-electronic-brain/
  134. Spinnaker, fecha de acceso: abril 8, 2025, https://spinnaker.io/
  135. Spinnaker Projects | Project Management Services in East Anglia & London : Spinnaker Projects, fecha de acceso: abril 8, 2025, https://spinnakerprojects.com/
  136. Spinnaker Watches, fecha de acceso: abril 8, 2025, https://spinnaker-watches.com/
  137. SpiNNaker—Programming Model – ePrints Soton – University of Southampton, fecha de acceso: abril 8, 2025, https://eprints.soton.ac.uk/369378/1/__userfiles.soton.ac.uk_Users_slb1_mydocuments_TC2329686.pdf
  138. The basic architecture of a SpiNNaker chip. – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/The-basic-architecture-of-a-SpiNNaker-chip_fig1_308293734
  139. Network bandwidth (BL) required to support regular operation of… | Download Scientific Diagram – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/Network-bandwidth-BL-required-to-support-regular-operation-of-SpiNNaker-18000-neurons_fig3_261161704
  140. External memory read bandwidth used by SpiNNaker chip containing… – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/External-memory-read-bandwidth-used-by-SpiNNaker-chip-containing-increasing-numbers-of_fig8_308293734
  141. SpiNNaker spike packet format (used throughout the machine). – ResearchGate, fecha de acceso: abril 8, 2025, https://www.researchgate.net/figure/SpiNNaker-spike-packet-format-used-throughout-the-machine_fig2_261465042
  142. sPyNNaker: A Software Package for Running PyNN Simulations on SpiNNaker – Frontiers, fecha de acceso: abril 8, 2025, https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2018.00816/full
  143. About us – Spinnaker Cementing Solutions, fecha de acceso: abril 8, 2025, https://spinnakeroil.com/about-us/
  144. Spinnaker Plugin Framework: A Major Milestone | by Ryan Pei, fecha de acceso: abril 8, 2025, https://blog.spinnaker.io/spinnaker-plugin-framework-a-major-milestone-872c379f9737
  145. A Step-by-Step Guide to ERP Implementation – Spinnaker Support, fecha de acceso: abril 8, 2025, https://www.spinnakersupport.com/blog/2023/09/20/a-step-by-step-guide-to-erp-implementation/
  146. SPINNAKER dev.indd, fecha de acceso: abril 8, 2025, http://www.spinndev.com/wp-content/uploads/2016/06/SPINNAKER_FALL-14_HIGH-RES.pdf
  147. Spinnaker Development, fecha de acceso: abril 8, 2025, https://www.spinndev.com/
  148. Tech Podcast: Carver Mead Says Neuromorphic Efficiency Can Help AI | EE Times Current, fecha de acceso: abril 8, 2025, https://www.youtube.com/watch?v=LJSpyyhG_6g
  149. Carver Mead Says Neuromorphic Efficiency Can Help AI – EE Times Podcast, fecha de acceso: abril 8, 2025, https://www.eetimes.com/podcasts/carver-mead-says-neuromorphic-efficiency-can-help-ai/
  150. A Look at TrueNorth – IBM – Neuromorphic Chip, fecha de acceso: abril 8, 2025, https://open-neuromorphic.org/neuromorphic-computing/hardware/truenorth-ibm/
  151. Cognitive computer – Wikipedia, fecha de acceso: abril 8, 2025, https://en.wikipedia.org/wiki/Cognitive_computer
  152. Deep learning for medical image segmentation – using the IBM TrueNorth Neurosynaptic System – eScholarship.org, fecha de acceso: abril 8, 2025, https://escholarship.org/content/qt3n66b3rv/qt3n66b3rv_noSplash_b548e4970525bc6891de6e4b5a0b2883.pdf
  153. TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip – IBM Research, fecha de acceso: abril 8, 2025, https://research.ibm.com/publications/truenorth-design-and-tool-flow-of-a-65-mw-1-million-neuron-programmable-neurosynaptic-chip
  154. How neuromorphic computing takes inspiration from our brains – IBM Research, fecha de acceso: abril 8, 2025, https://research.ibm.com/blog/what-is-neuromorphic-or-brain-inspired-computing
  155. A Look at Loihi 2 – Intel – Open Neuromorphic, fecha de acceso: abril 8, 2025, https://open-neuromorphic.org/neuromorphic-computing/hardware/loihi-2-intel/
  156. Loihi: A Neuromorphic Manycore Processor with On-Chip Learning, fecha de acceso: abril 8, 2025, https://redwood.berkeley.edu/wp-content/uploads/2021/08/Davies2018.pdf
  157. Intel’s Silicon Brain System a Blueprint for Future AI Computing Architectures – HPCwire, fecha de acceso: abril 8, 2025, https://www.hpcwire.com/2024/04/24/intels-silicon-brain-system-a-blueprint-for-future-ai-computing-architectures/
  158. A Brief Introduction to Neuromorphic Processors – Spectra – Mathpix, fecha de acceso: abril 8, 2025, https://spectra.mathpix.com/article/2022.09.00090/a-brief-introduction-to-neuromorphic-processors

Related Posts

Discord
Email
Phone
Telegram
WhatsApp
WhatsApp
33618047944
Phone
tel:+33618047944
Email
mailto:info@ideant.xyz
Telegram
https://telegram.me/IDEANT_electronics
Discord
https://discord.com/channels/1338488205216845874/1338488206454427712
Scroll to Top