IBM Quantum Summit 2021: QPU Eagle Breaks the 100-qubit Barrier
Building on year-over-year engineering advancements IBM unveils Eagle, a 127-qubit quantum processor, and the next generation of cloud-based quantum systems.
The IBM Quantum Summit 2021 saw the rise of the Eagle, a 127-qubit quantum chip built as a milestone in the roadmap to achieving "quantum advantage." IBM's Quantum Advantage is said to be the point where quantum operations overtake classical computation for algorithm execution.
A rendered view of the IBM Eagle quantum processor. Image used courtesy of IBM
As a 127-qubit processor, Eagle is a third-generation (in this line) quantum processor and the first to exceed 100 qubits. It is intended to be a massive milestone in the roadmap of IBM's quantum development.
Let's dive into some of the details surrounding this years' achievements in quantum development, the performance metrics behind the hardware, and what we can expect of quantum advantage in the future.
Introducing IBM’s Eagle Quantum Processor
The entire generation of IBM quantum chips is named after birds. Each generation, like Falcon (2019) and Hummingbird (2020), has required key advancements such as optimized lattices or scalable readout (multiplexing).
The Eagle's key advantage is novel packaging and control.
Exploded view of the Eagle QPU showing qubit layers and interposers. Image used courtesy of IBM
The Eagle quantum processor's exploded view (as seen in IBM's Quantum State of the Summit video) reveals a two-chip system.
The first chip is the Josephson Junction base superconducting transmon attached to a separate interposer chip through bump bonds. The interposer layer allows for "tremendous flexibility" in routing signals and laying out the device through buried wiring layers and substrate vias, similar to the CMOS world.
Key technology breakthroughs and a rising qubit count are not the only measures of success for the IBM quantum team.
As with any engineering project, engineers must devise quantitative methods to measure overall system performance. To that end, IBM has three significant performance metrics.
Performance Metrics for IBM’s Flock of Quantum Computers
Three main drivers are said to indicate increased performance for Quantum Processor Units (QPUs): scale, quality, and speed. Each of these three metrics has an associated unit to measure relative generational performance.
The three main metrics of quantum processor performance are scale, quality, and speed. Image used courtesy of IBM
The first metric is scale, which is the number of qubits present in the QPU, and this measure has nearly doubled yearly for the past three years, from 27 qubits in 2019 to 127 qubits in 2021. The IBM roadmap for scaling quantum technology is projecting a 1000+ qubit system in 2023.
The second metric, quality, is said to be a measure of "quantum volume." Quantum volume is a measure of performance, similar to transistor counts for semiconductor electronics.
This measure is based on a statistical method called the "heavy output generation problem," which determines a confidence level in the quantum circuit outputting the correct bits strings (as opposed to errors).
IBM states that higher quantum volume directly equates to higher processor performance. In 2020, the gate error rate was just below 10-2. This year, with the Falcon-R10 system (yellow dot), gate error has been reduced below 10-3.
Reducing gate error rates over four years. Screenshot [video] used courtesy of IBM Research
The final metric, speed, is measured in-circuit layer operations per second (CLOPS). At the high level, CLOPS measures how fast operations are performed on the QPU.
It's important to understand that QPUs (at least for the time being) are intended to be tightly coupled to classical computation systems. Using the Qisket Runtime environment, users request access to the QPU for algorithm acceleration. This new environment contributed to a 120x speed improvement in simulating molecules.
To better understand IBM's future in quantum development, it's necessary to take a look at its System One computer system.
IBM’s System One & The Future of Quantum Development
System One claims to be the world's first quantum computer system. It exists at the intersection of numerous scientific fields, including cryogenics, systems engineering, and industrial design.
IBM's System One. Image used courtesy of IBM
IBM System One was initially constructed (mechanically) in 2018 at IBM partner Goppion's headquarters in Milan. Now, IBM System One is a global system with operations running in North America, Germany, and Japan.
The cloud-based system allows for classical computing interaction with the quantum circuitry while maintaining the demanding tolerances required to keep qubits stable and calibrated.
System Two, tentatively unveiled at IBM Quantum Summit 2021, is projected to replace System One in the coming years.
Future Applications Primed For Quantum Acceleration
To wrap up today's coverage of IBM's quantum progress, let's look forward to 2025.
Targeting applications for financial services, drug discovery, and advancements in material design are expected to be the first to benefit from the advancements of IBM's future Osprey (2022) and Condor (2023) chips.
Early projections of viable applications using QPUs. Screenshot [video] used courtesy of IBM Research
Economically, these forward-looking applications are expected to generate more than $1B USD annually as early as 2023, according to the Boston Consulting Group (BCG).
IBM's development path is aggressive. Stretching from Eagle to Condor represents a nearly 10x increase in qubit count, along with the advancements in System Two, in just two short years.
If hitting each milestone for the past three years is any indication, IBM should be set up as the forerunner in quantum system deployment.