Researchers synchronize 105,000 tiny magnetic oscillators in 45 nanoseconds, pointing to faster, low-power computing

What just happened?

Researchers have synchronized 105,000 tiny magnetic oscillators in just 45 nanoseconds, creating the largest network of its kind ever built, and a real sign that this odd corner of computing might actually scale.

To put that number into perspective, the previous record was 64 oscillators. This new experiment increased that number to 105,000, more than 1,600 times as many, and the sync time barely moved.

• With 100 oscillators, synchronization took 10 nanoseconds.
• With 105,000, it took 45 nanoseconds.

That's the part researchers care about most. Usually, adding more components slows a system down. Here, the system barely slowed down.

How it works

Each oscillator is tiny, just 10 to 20 nanometers wide. There's no external clock running the show. Once the grid gets nudged, the oscillators fall into sync on their own, using nothing but their natural magnetic spin. Think of ripples spreading across a pond until the whole surface moves the same way. And they do it using very little power.

Why this matters

Because a grid of synchronized oscillators can solve certain problems just by settling into place. Anything involving waves, statistics, approximation, or pattern recognition tends to fit. Two applications already mentioned in the research are:

• Ising machines, which are used for optimization problems
• Reservoir computing, which processes time-based data using the natural behavior of a physical system rather than traditional logic circuits

Next steps

The next step is making these grids programmable - adjusting each oscillator's frequency, phase, and coupling strength so the whole array settles into a specific answer. You'd read the result straight from how the grid synchronizes, no separate decoding required.

If that works at scale, people are eyeing applications such as:

• High-speed networking
• Financial modeling
• Real-time analytics
• AI acceleration

Speed and performance

Speed is the other draw. The paper says grids like these could run at tens of gigahertz while using comparatively little energy. And that 45-nanosecond settling time for all 105,000 oscillators is roughly the time a regular CPU takes to perform one calculation across a full matrix.

There's also a clear contrast with quantum computing, which needs heavy error correction just to keep working. This system doesn't have that problem. Once it settles, the signal it produces is sharp and stable. The quality factor measured in the experiment was over a million, meaning the resulting frequency was about as clean and precise as the tone from a tuning fork.

The study

The study, published in Nature Nanotechnology, was conducted by researchers from the University of Gothenburg, IIT Bhubaneswar, and Tohoku University.