Single-GPU Systems Will Beat Quantum Computers for a While: Research | Tom’s Hardware
Things are rarely what they seem, and the world of quantum computing best fits that description. Described as a fundamental shift in our processing power, the development of quantum computing has accelerated incredibly over the past few years.not yet According to a research paper published in the journal of the Association for Computing Machinery:the related quantum computing (what we usually call even the most powerful classical computers draw circles) needs breakthrough discoveries in many areas before it can beat mere graphics cards. is.
The Most Amazing Element in the Paper The conclusion is that many applications will remain suitable for classical computing (rather than quantum computing) for longer than previously thought. The researchers say this is also true for his quantum system, which operates on more than one million physical qubits, and the team simulated its performance as part of the study.
Considering that today’s top system, IBM’s Osprey, is still “only” crammed at 433 qubits (IBM promises to launch a 4,158-qubit system in 2025), one million qubits The timescale to which we are headed is further ahead than expected.
Common drug discovery, materials science, scheduling and optimization problems are still in the sights of quantum computing, according to the researchers, rather than the applications and workloads themselves being the problem. The problem lies in the quantum computing system itself, its architecture and inability to ingest the enormous amount of data that some of these applications, now and in the future, require before finding a solution. This is a simple I/O problem, unlike the one we all knew when HDDs were bottlenecking left and right of CPUs and GPUs before NVMe SSDs became the norm. . Data can only be supplied very quickly.
But the amount of data sent, how fast it gets to its destination, and how long it takes to process are all factors in the same equation. In this case the equation is for quantum dominance. That is the moment when quantum computers offer performance beyond what is possible with conventional systems. And for workloads that require processing large datasets, it seems quantum computers will need to watch out for GPUs such as his Nvidia’s A100.
Quantum computing may need to solve big computational problems with small amounts of data, but classical ones have the enviable task of handling “big data” problems. This is a hybrid approach to quantum computing that has gained momentum in recent years.
According to the blog post (opens in new tab) According to Microsoft’s Matthias Troyer, one of the researchers involved in the study, this is a perfect fit between chemistry and materials science, but ultimately it’s about drug design, protein folding, weather and climate prediction. This means that workloads such as are better suited to traditional systems. Bill against the “Big Computing, Small Data” philosophy.
While this might feel like an ice bucket challenge flop to the hopes of quantum computing, Troyer was quick to emphasize otherwise. The world today boils down to problems of chemistry and materials science,” he said. “Better and more efficient electric vehicles rely on discovering better battery chemistries. More effective and targeted cancer drugs rely on computational biochemistry.”
But there’s another element in the researchers’ paper that’s hard to ignore. It is that current quantum computing his algorithms alone seem insufficient to guarantee the desired “quantum supremacy” result. It’s a matter of simple performance, not system engineering complexity of quantum computers. In general, quantum algorithms do not provide sufficient acceleration. For example, Grover’s algorithm offers his quadratic speedup over traditional algorithms. But that alone isn’t enough, according to researchers.
“These considerations help distinguish hype from practicality in the search for quantum applications, and can guide algorithm development,” reads the paper. It shows that we need to focus on quadratic speed, ideally exponential speedup, and carefully consider I/O bottlenecks.”
Yes, yes, quantum computing still has a long way to go. But IBM and Microsoft around the world are working tirelessly to make it possible. Many of the problems facing quantum computing today are the same ones faced in the development of traditional hardware. Today’s CPUs, GPUs and architectures got off to a much earlier and more impactful start. But they still had to undergo the same design and performance iterations that quantum computing would eventually do, within its own brave new timeframe. Services (AWS), and the fact that it was written by scientists from the Institute for Scalable Parallel Computing in Zurich (all parties with a vested interest in the development and success of quantum computing) further enhances the potential of that goal. Raise.