Technology
PUBLISHED
In this article you'll discover
- Researchers from across the US have presented a novel benchmark for assessing just how useful a quantum computer might be.
- While it shows we're still a long way off from solving useful problems with quantum computers, the benchmark is already helping companies make improvements to their machines.
- “It gives you the summary performance of the computational power of different quantum computers and allows us to track progress towards useful quantum computing,” the experts told IFLScience.
Quantum computers could change the way we simulate everything. Once we figure out how to get them running at scale, problems beyond even today's supercomputers might one day become routine to solve. Despite being a long way from that lofty goal, we do have working quantum computers today, and now we might have our first good way of assessing how useful they are.
This new benchmark is called QUOPS, or the Quantum Universal Operations Performance System. It’s a special set of programs that can run on any kind of quantum computer, something that's particularly important because there are many, many ways to build one – which is why assessing their performance up to now has been a tad tricky.
With a regular computer you might talk about its processing speed, number of cores, or amount of RAM, all different ways of assessing how good it will be at running certain programs. But the comparison is markedly more complex for quantum computers due to the wildly different approaches researchers have taken in constructing them.
“[QUOPS] is designed to be a benchmark that allows you to compare all sorts of different quantities across different technologies and using different kinds of architectures,” Dr. Timothy Proctor at Sandia National Laboratories in California, told IFLScience.
“The purpose of this benchmark is really to be a flagship benchmark for the field. It gives you the summary performance of the computational power of different quantum computers and allows us to track progress towards useful quantum computing.”
Current quantum computers are, as mentioned, not at supercomputer-beating levels yet. But having a benchmark like QUOPS could be a very useful tool for researchers to quantify exactly how wide that gap is and point towards clearer ways of improving their designs.
“I think we'll see (and we've already seen) that this is going to allow people to really assess these larger fault tolerance systems and make decisions about the best way to progress,” Nicholas Harrigan, product marketing manager for quantum at NVIDIA, told IFLScience.
QUOPS has already been used on some of the world's cutting-edge quantum computers, including those belonging to Google, IBM, and quantum computing firm Quantinuum. The benchmark outputs two separate scores, the QUOPS, which is a measure of the largest size of quantum operation that the computer can perform, and the QUOPS/s, which is the speed at which the machine can perform quantum operations.
Google's Willow achieved 216 QUOPS at 2.0 × 10⁷ QUOPS/s, IBM's Boston achieved 204 QUOPS at 3.1 × 10⁵ QUOPS/s, Quantinuum's H2-1 achieved 1320 QUOPS at 353 QUOPS/s, and Quantinuum's Helios-1 achieved 1504 QUOPS at 303 QUOPS/s. As you can see, there are massive disparities between machines in terms of both the size and speed of their quantum calculations.
To put the numbers into context, the researchers included prospective QUOPS scores for some potentially useful quantum problems. These included factoring a large number called RSA-2048, which is thought to be impossible for classical machines, and figuring out the reaction mechanism of an enzyme called FeMo-co. The researchers concluded that quantum computers would need to reach QUOPS values of 2.5 × 108 QUOPS for RSA-2048 and 3.4 × 108 QUOPS for FeMo-co, an increase of about five orders of magnitude.
For the team at Quantinuum, initial lower scores with QUOPS led to improvements, as it showed some of the limitations of the Helios-1 system – limitations that might have been less obvious without such a tool.
“We realized there are other things that could be done. And we played around with different components of this implementation, tweaked different parts, and were able to improve the score to some extent,” co-author Charlie Baldwin, R&D manager at Quantinuum, told IFLScience.
QUOPS is certainly not the last word on benchmarking in the world of quantum computers, and the team is very clear that what works today to assess quantum computers might not work tomorrow.
“I think a really important thing with benchmarks in general is that it can be very easy for them to misdirect because the field moves on,” Proctor told IFLScience. “You can end up targeting something that is no longer relevant because of things you've discovered over time.”
QUOPS has in-built flexibilities, with what Proctor describes as “lots of parts that can be easily swapped out.” Hopefully this would keep it as an important benchmark for a long way yet.
A paper presenting QUOPS is available on ArXiv.
1