IBM Quantum with the new Qiskit Runtime delivers more than 120x speedup of quantum workloads

A few days ago, IBM was pleased to announce to the public that their team has achieved speedup up to 120x in simulating molecules, all of this was thanks to several improvements, one of them was the ability to run quantum programs with Qiskit Runtime, entirely on the cloud.

To understand Qiskit we should first take a look at ML and QML

Machine Learning is probably one of the hottest topics in the past years and has entered our daily lives, it compromises everything from deep learning to smart NLP systems. It takes the training data set and uses it to train the model and the model then can make predictions on the new data. One of the biggest benefits of Machine Learning is that if you have the right algorithm, a researcher doesn’t have to worry about how the model works, the algorithm is trained to find the most optimal parameters and deliver the best result.

With the development of quantum computing, the interest in QML or Quantum Machine Learning has increased dramatically. The most general definition of QML that can be explained in simple words is that QML is a combination of quantum algorithms with Machine Learning algorithms, and here Qiskit comes into play.

What is Qiskit?

Qiskit’s logo, Qiskit is an Open-Source framework, supported by IBM

Qiskit is one of the most popular Quantum toolsets out there, created by IBM for their plan to jump into the quantum space. Qiskit is an open-source SDK, that is not technically specific and can operate on different operating systems, and can run on different hardware. The biggest news is that Qiskit offers a new module specific to Machine Learning, and have just released the latest version of the Qiskit QML module. Although Qiskit runs independently, IBM is always there to support this program by offering online tutorial lessons on how to learn quantum computing, and also offers summer school for young learners, starting from July this year. For more info about Qiskit read our full aritcle.

Qiskit fulfils IBM’s expectations of achieving x100 speedup

Last fall, IBM demonstrated their ambitious plan to speedup up to 100x of the quantum workloads on their IBM quantum roadmap, for scaling quantum technology. And not only that, on April 11, 2021, they have announced that they had not only met their goals, but also went a bit higher and achieved 120x speedup in simulating molecules. All these advancements thanks to several improvements, most notably they got the ability to run quantum programs completely on the cloud with their Qiskit Runtime.

The advancements IBM have made

IBM does still make mainframes (an ancient 1970’s ICL mainframe pictured), but the real buzz from IBM is about their Quantum Computing capabilities and their room-sized machines which could allow users to perform some tasks such as Molecular Simulation much more effectively.

Since quantum programs have millions of interactions between quantum and classical computing, it was critical to creating systems that accelerate the execution of quantum programs natively, and not just only quantum circuits. These systems that are built to execute the quantum programs are required to have greater effective capacities and they require many improvements across the whole stack, including system software, cloud service design, control hardware, and sometimes even quantum hardware.

IBM Quantum in 2017 had predicted that a quantum computer can simulate the behaviour of the LiH molecule and the process of modelling the Lithium Hydride molecule would take about 45 days with the technology available at the time. But IBM’s engineers with the aid of Qiskit went further than this, and in these four years managed to reduce the prediction to only 9 hours. Read the full article.

Now, we can solve the same problem in just nine hours — a 120x speedup.

IBM Representative

The main improvements were algorithmic. They managed to reduce the iterations of the algorithm that was required to receive a final answer by two and sometimes ten times. These improvements into the system have removed approximately 17 seconds per iteration. Also, improved chip performance had led to ten times decrease in the number of tries and the repeated circuit runs with each iteration of the algorithm. And at last with the improved control system, like better readout and qubit reset performance has reduced the amount of time per execution from 1000 microseconds down to 70 microseconds.

Qiskit Runtime role in these advancements

Qiskit Runtime played a big role in reducing the time to perform the tasks of computing with its unique design created for quantum computers, so instead of building up latencies as the code passes between the cloud-based quantum computer and the user’s device, the developers can run their program in the Qiskit Runtime environment. In this situation, all of the hard work is handled by the IBM hybrid cloud. These new software architectures allow the users to maximize the time of computing and reduce the waiting time.

What does this speedup mean for developers?

The speedup is going to allow the developers to work and experiment with quantum applications. The Qiskit Runtime is going to allow the developers to try the new and powerful quantum kernel alignment algorithm. This algorithm searches for an optimal quantum kernel, and with this kernel, it can perform machine learning tasks. And with the aid of this algorithm, they proved that quantum computers will make a tremendous impact over classical computers in the area of machine learning.

IBM’s expectations for the future

IBM representatives say that their goal is to find practical quantum computing use cases, then deliver them to their large developer base. They also hope that their Qiskit Runtime is going to allow users to take the advantage of the 127-qubit named IBM Quantum Eagle, listed for a release date this year, and the 1.121 qubits called Condor slated for release in 2023.