A lone Nvidia GPU surpasses the physical horsepower of a quantum computer at least in these apps

A group of researchers from Microsoft and the Scalable Parallel Computing Laboratory in Zurich have offered a hard reality check to those who extol the world-altering potential of quantum computers, discovering that off-the-shelf GPUs can sometimes do better than machines from the frontiers of the world. physics.

Drug discovery, material sciences, scheduling, traffic congestion, supply chain, and weather forecasting are all commonly cited applications that vendors claim quantum computing is suitable for.

But in an article published in the journal of the Association for Computing Machinery, Torsten Hoefler, director of the Scalable Parallel Computing Laboratory, together with former Microsoft researcher Thomas Hner and Microsoft’s Matthias Troyer concluded that unless dramatic improvements in the hardware and software, even future quantum systems are unlikely to achieve practical speeds in many of these workloads.

For a quantum system to be worthwhile, it must be able to perform a task faster than a conventional system, and to test this, the team compared a hypothetical quantum system with 10,000 qubits of error correction, or about a million qubits. physical. a classic computer equipped with a single Nvidia A100 GPU.

To be clear, no such quantum system exists today. The most advanced quantum computers currently available scale to a few hundred physical qubits. IBM’s Osprey system, for example, packs 433 qubits. And while IBM says it’s on track to deliver a 4,158-qubit system in 2025, even that falls well short of the system envisioned by Hoefler and his co-authors. On the other hand, as high performance computing (HPC) systems go, the conventional system considered in this paper is downright anemic.

“For our analysis, we set a break-even point of two weeks, which means that a quantum computer should be able to perform better than a classical computer on problems that would take no more than two weeks for a quantum computer to solve.” ” Troyer explained in a blog post published on Monday.

The comparison, according to the authors, revealed a clear problem with most of today’s quantum algorithms. A quadratic speedup, such as that allowed by Grover’s algorithm, is not sufficient to gain an advantage over conventional systems. Instead, “super quadratic or ideally exponential accelerations” are needed.

This is not the only problem facing quantum architectures. Input and output (I/O) bandwidth is another limiting factor.

“Our research has revealed that applications that rely on large datasets are better served by classical computing because bandwidth is too low on quantum systems to allow for applications such as searching databases or training machine learning on large datasets,” Troyer explained.

He added that this means that workloads such as drug design, protein folding, as well as weather and climate forecasting are better suited to conventional workloads given the current state of the technology.

This doesn’t mean that quantum computing is useless; it simply means that, at least for the foreseeable future, the applications for quantum systems are likely to be narrower than marketers would have you believe.

In general, quantum computers will be practical for “big computation” problems on small data, not big data problems

“In general, quantum computers will be practical for ‘big computation’ problems on small data, not big data problems,” the researchers wrote.

One such workload likely to benefit from quantum systems is the chemical and materials sciences. This is because many of these workloads rely on relatively small datasets.

“If quantum computers only benefit chemistry and materials science, that would be enough,” Troyer stressed. “Many problems facing the world today boil down to problems in chemistry and materials science. Better and more efficient electric vehicles are based on research into better battery chemistries. More effective and targeted cancer drugs are based on computational biochemistry.”

Cryptanalysis using Shor’s algorithm presents similar challenges, the researchers note. However, not all algorithms capable of exponentially accelerating are necessarily suitable for quantum systems. The team notes that while linear algebra has exponential speed, this is negated by I/O bottlenecks as soon as the array is loaded into memory.

“These considerations help separate hype from practicality in the search for quantum applications and may guide algorithmic developments,” the paper reads. “Our analysis shows that there is a need for the community to focus on super-square speeds, ideally exponential accelerations, and I/O bottlenecks need to be carefully considered.”

#lone #Nvidia #GPU #surpasses #physical #horsepower #quantum #computer #apps

Leave a Comment