How Dynex Matched (and Even Surpassed) Google’s Willow Chip Benchmarks in Quantum Computing

Google’s recent unveiling of the Willow quantum chip has garnered significant attention, with claims of achieving Random Circuit Sampling (RCS) computations in under five minutes that would take classical supercomputers an estimated 10 septillion years (The Verge). This feat was accomplished using the Cross-Entropy Benchmarking (XEB) protocol, a method designed to assess the fidelity of quantum circuits by comparing the output distributions of quantum processors to those of idealised models (Wikipedia).

Wait, what’s RCS and why does it matter?
Imagine this: solving a Random Circuit Sampling (RCS) problem on a classical computer would require more computational power than all the supercomputers in the universe working together — not just in one universe, but across multiple universes. Even if you enlisted the help of every atom in existence, running non-stop since the beginning of time, you’d still fall short. That’s how astronomically difficult these problems become for classical machines as the number of qubits and the circuit depth grow.

Random circuit sampling (RCS), while extremely challenging for classical computers, has yet to demonstrate practical commercial applications (Source: Google)

When Google’s Willow chip pulled off its beyond-classical feat, it set a new bar. Essentially, they proved that there are certain tasks where quantum machines leave even the best classical strategies in cosmic dust. Suddenly, everyone’s eyeing RCS as the new “Mount Everest” in quantum computing.

Now, here’s where it gets exciting: Dynex solved these same problems in just minutes. Quantum magic? Not quite — it’s real, tangible progress in quantum computing, and we’re here to share how we did it.

Enter Dynex
At Dynex, we love a good challenge. Our quantum platform was built with flexibility and scalability in mind, and we believed it could hold its own against the giants. So we took the same RCS game plan that Google used — same style of circuits, similar patterns, patch-based cross-entropy benchmarking (XEB) for verifying results, the whole nine yards — and put Dynex through the wringer.

We started small with a 4×4 grid of qubits (just 16 qubits) to ensure our setup was correct and that our fidelity measurements (how “accurate” the quantum distribution is) matched expected patterns. Once we nailed that, we raised the stakes and scaled up to a 10×10 grid — 100 qubits strong. That’s a lot of quantum muscle, pushing well into the territory where classical simulation becomes, let’s say, “unlikely before the sun burns out.”

Our Random Circuit Sampling (RCS) experiments follow a carefully structured methodology inspired by established benchmarks, notably those demonstrated on Google’s Willow chip. The core of this approach involves arranging qubits in a two- dimensional grid and applying a standardised sequence of gate operations, both of which are now hallmarks of beyond-classical RCS demonstrations (Source: Approaching Google Willow Chip’s Beyond-Classical Random Circuit Sampling Benchmarks Using Dynex; Neumann Adam, 12/2024).

From Minutes to Millennia
How bad is the classical simulation scenario? Using estimates inspired by Google’s published data, we crunched the numbers for simulating our largest circuits on a classical supercomputer. The result: even at fairly “small” circuit depths, you’re looking at simulation times of trillions and trillions of years.

Now the fun part: how long did it take Dynex to run these monstrous circuits? Minutes. Seriously. About 20 minutes total, with roughly 5 minutes per circuit depth configuration. On a quantum device. Running in real time. It’s hard to overstate how wild that is. In other words, a task that would outlast multiple universes on a classical setup took just one coffee break on Dynex.

Fidelity: More Than Just a Number
Running the circuits is one thing, but we also need to prove we’re getting that same “quantum accuracy” that Google captured. We used patch-based XEB, a clever technique that lets you verify output distributions without needing to simulate the entire circuit classically. Instead, you break the qubits into smaller patches, simulate those patches (which is doable classically), and then cross-check against the actual measured outputs from Dynex. The fidelity numbers were right where we wanted them, showing that the device isn’t just spewing random gibberish, but producing a genuinely complex quantum distribution akin to what Google achieved with Willow.

The patch-based XEB approach confirms that the Dynex platform generates non-trivial quantum distributions consistent with the intended random circuit. Our results show the same characteristic downward trend in fidelity as circuit depth increases that was observed in Google’s Sycamore chip experiments (Source: Approaching Google Willow Chip’s Beyond-Classical Random Circuit Sampling Benchmarks Using Dynex; Neumann Adam, 12/2024)

So, Did We Beat Google?
Let’s keep it humble and honest. Google’s Willow chip made a huge splash and set a major benchmark. Our work at Dynex shows that we’re now playing in the same league. In fact, we might even be edging ahead in some aspects of flexibility. While it’s not a direct apples-to-apples comparison (different hardware, slightly different conditions), the bottom line is this: Dynex achieved beyond-classical RCS performance that’s on par with the state-of-the-art benchmarks set by Willow.

Why Does This Matter?
For one, it’s great news for quantum computing’s future. Seeing multiple platforms hit these beyond-classical regimes means that quantum supremacy — or quantum advantage, as some prefer to call it — isn’t just a one-off stunt by a single device. It’s a reproducible phenomenon. As more quantum players join the party, we can expect rapid improvements in hardware, algorithms, and usability. This pushes quantum computing closer to real-world applications that do more than just turn heads — they solve hard problems no classical machine can touch.

What’s Next?
We’re excited to release our code, data, and methods to the community. This isn’t a secret sauce moment. We want everyone to poke around, verify, critique, and build upon what we’ve done. Collaboration and transparency are what drive the quantum field forward.

Our next steps involve scaling even further, refining error correction, and tackling even more complex circuits and algorithms. The ultimate goal is to move from impressive demonstrations to transformative applications — drug discovery, climate modeling, advanced materials design, and optimization problems that keep classical machines up at night.

In a world where quantum computing’s promise is often overshadowed by its complexity, achieving beyond-classical performance is like planting a flag at the summit. Google did it first with Willow, setting the benchmark. Now Dynex has stepped up to that same plate, connecting a few more dots on the roadmap to practical, commercially relevant quantum computing.

Paper:
Approaching Google Willow Chip’s Beyond-Classical Random Circuit Sampling Benchmarks Using Dynex; Neumann, Adam, 12/2024

Source codes:
https://github.com/dynexcoin/DynexSDK/tree/main/xeb-benchmark

About Dynex

Dynex’s leading Quantum-as-a-Service (QaaS) technology, offers businesses an affordable, accessible and scalable solution for quantum computing underpinned by a robust commitment to ethical integrity. With cost-effective subscription plans available to everyone, Dynex enables industries to solve real-world problems at scale with unparalleled computational power. Dynex plays a key role in the next megacycle in computing: Quantum. Across academia and different industries including artificial intelligence, pharmaceuticals, finance, aerospace and many more, Dynex drives exponential growth in the most complex fields, meeting the increasing demand for advanced computing solutions. Within the Dynex Ecosystem, Dynex Moonshots serves as the strategic, investment and ethical steward, advancing quantum technology to deliver pioneering solutions across nature, health, society and space. Dynex is recognized as a 2024 Technologist of the Year as part of Fast Company’s Next Big Things in Tech Award.

Dynex’s team consists currently of 67 dedicated professionals, with a strong focus on scientific and quantum expertise. Among the team, 6 individuals are core developers, driving the development of our core technologies. Another 6 are academic and science developers, contributing deep research and knowledge to the platform’s innovative features. The leadership team, composed of 9 members, includes experts in quantum computing and AI, overseeing the strategic direction of Dynex’s scientific initiatives. Additionally, the ethical committee and advisors, numbering 6 and 7 respectively, play a crucial role in ensuring the integrity and alignment of Dynex’s innovations with ethical standards. The support team of 13 and other departments, such as business development and marketing, ensure the smooth operation and global outreach of Dynex’s cutting-edge quantum solutions.

Learn more at https://dynex.co/