iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Neuromorphic computing takes ideas from biological nervous systems to build computing systems, algorithms and sensors. It does not reproduce a complete brain. In an interview published by AIhub on 1 October 2026, Oliver Rhodes, Senior Lecturer in Bio-Inspired Computing at the University of Manchester, explains how event-driven processing and platforms such as SpiNNaker work—and why applications such as patient-specific disease simulations remain research goals, not clinical services.
What does “neuromorphic computing” mean?
Neuromorphic computing is a broad research and engineering field that uses the brain as inspiration at several levels: algorithms, computing architectures and sensing hardware. It is not another name for all artificial intelligence, and spiking neural networks are one approach within the field rather than a definition of it.
“Neuromorphic computing is quite a broad subject, which essentially looks to biology as inspiration to develop next-generation computing systems,” Rhodes told AIhub. The inspiration is useful, but it is not a claim that engineers have recreated the brain: Rhodes notes that much about how the brain represents information is still not understood.
How can event-driven processing save work?
Respond to activity rather than process continuously
In a spiking neural network, information is represented through discrete events, or spikes. In an event-driven design, a processor can remain inactive when no incoming spikes arrive and do work when activity does arrive. This contrasts with approaches that repeatedly process data whether or not the scene or signal has changed.
#1 Best Overall
The potential efficiency comes from doing less work when there is little activity. It is not a guarantee that every neuromorphic device will use less energy than every conventional computer: results depend on the task, the architecture and how well the algorithm fits the hardware.
Keep data near the computation
A second design idea is to reduce the movement of information between a processor and a separate memory area. Rhodes contrasts the brain’s distributed storage with a conventional computer’s processor-and-RAM workflow. Keeping information near where it is used can reduce data movement, but specialist hardware imposes constraints on which algorithms it can run.
That makes hardware and software design interdependent. Researchers must adapt algorithms to a device and map the work onto it effectively; an algorithm that performs well in one setting may not translate directly to another.
What is SpiNNaker, and what does it demonstrate?
SpiNNaker is a large-scale research platform developed at the University of Manchester to model spiking neural networks. Rhodes describes it as a one-million-core system built through a 20-year effort. The university’s own description says it incorporates over one million ARM mobile-phone processors and can model spiking networks at mouse-brain scale in biological real time. Those are institutional descriptions of the platform’s design and modelling capability, not a general benchmark for neuromorphic computers.
Many low-power processing elements work together, while a routing architecture sends small packets representing neural spikes among processors. The goal is to support neural simulations at scale, rather than to serve as a general-purpose replacement for everyday computers.
Rhodes also identifies SpiNNaker2 as a second-generation system and Intel’s Loihi as another platform built around related principles. The University of Manchester’s International Centre for Neuromorphic Systems (ICNS) describes Loihi 2 hardware hosted there for the Edgy Organism project. The available information does not provide comparable price, energy, throughput or workload data for these platforms, so it does not support ranking them against one another or against GPUs.
Rank #3
Where is neuromorphic technology being used?
Event-based vision sensors
A conventional camera typically produces image frames at intervals. An event-based vision sensor instead reports pixel changes as events, rather than repeatedly outputting unchanged parts of the image. That can suit sparse visual data or scenes with fast movement and high contrast.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Rhodes illustrates the difference with a rocket launch: he says a conventional image can be saturated by the ignition, while event-based footage can retain detail in the plume and sky. This is an illustrative example from the interview, not a quantified comparison of camera performance.
Rhodes describes neuromorphic vision sensors as commercially available and among the field’s more mature products. However, users often process sensor output with conventional AI because conventional algorithms are more accessible than neuromorphic processors and algorithms. ICNS also describes event-driven sensors as useful for sparse data and lists research combining vision with processing. The interview does not identify a particular manufacturer, camera model or price.
Neuroscience simulation
SpiNNaker was designed in part to help accelerate neural simulations. Rhodes recounts a past milestone in which a cortical model ran in real time, while noting that newer conventional computers have since surpassed that record. The interview does not give the benchmark paper or test conditions, so this account should not be treated as a current, independently comparable performance result.
Rhodes also discusses possible patient-specific models for conditions such as Alzheimer’s disease, or simulations that could help researchers understand responses to deep brain stimulation for Parkinson’s disease. He describes this as an early research area and says a local doctor cannot currently run such a model. Neuromorphic computing is therefore not an established way to diagnose, predict or treat either condition.
Recommended Free Tools
Edge devices, robotics and smart glasses
Processing data close to a sensor could be useful where power, connectivity or computing capacity is limited, including remote settings and small devices. Rhodes points to smart glasses as a possible application, not a finished consumer product.
Best Value
One project example is NimbleAI, which combines event-based vision, foveated sensing and a small hardware accelerator. Foveated sensing directs higher-resolution attention to regions of interest, rather than treating every part of a scene equally. ICNS says NimbleAI was an EU Horizon Europe project that ended in March 2026; Manchester’s contribution included foveated-sensing algorithms and real-time near-sensor hardware. These project activities do not establish that a completed smart-glasses product is available.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What are the main limitations?
- Less mature software tooling: Rhodes says neuromorphic computing lacks the established software stack available for GPU-based machine learning. Compiling work for specialist hardware and distributing it across a device remain research challenges.
- Performance depends on mapping: How a model is mapped onto a chip can affect its results. Hardware capability alone does not determine whether a particular application will work well.
- Biological inspiration has limits: The brain is not fully understood, so its mechanisms cannot simply be copied into a complete computing blueprint.
- Learning is not human-like: Online learning and reinforcement learning are active research areas, but the interview says human-like learning has not been achieved.
- Comparisons need matched tasks: Rhodes cautions against directly comparing academic neuromorphic work with systems such as ChatGPT when their training resources differ. The interview does not establish a general energy, latency or accuracy advantage over conventional computing.
As Rhodes puts it, “We build systems to try to harness this energy saving feature, but it places constraints on the types of algorithms that you can implement.” The trade-off is central: an architecture designed to exploit sparse, event-driven activity may be a poor fit for a workload that does not have that structure.
How should you compare neuromorphic systems with conventional computing?
There is no single fair score that captures every system. A useful comparison starts with the intended workload and asks how each design handles it:
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problems- Activity model: Does the task benefit from processing discrete events, or does it need regular frames or continuous data processing?
- Data movement: Can information stay near the computation, or must it repeatedly travel between separate memory and processing units?
- Software readiness: Are the algorithms, compilers and tools available for the target hardware?
- Measured task performance: Is there a comparable test for the same workload, with stated conditions and metrics?
Without matched tests, claims that one platform is simply “more efficient” or “faster” than another overreach. SpiNNaker, SpiNNaker2, Loihi, event cameras and GPUs serve different roles, and the cited material does not establish a head-to-head ranking.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
