/Nvidia SDK simulates quantum computing circuits on GPU systems (via Qpute.com)
Java's popularity hinges on future developers, adaptability

Nvidia SDK simulates quantum computing circuits on GPU systems (via Qpute.com)


Nvidia has dipped a toe in the quantum computing waters, but don’t expect the company to dive into building its own quantum systems any time soon.

Nvidia built a development platform with CalTech for simulating quantum circuits on GPU-accelerated systems. The companies used a newly minted SDK called cuQuantum to simulate quantum circuits that run on its Nvidia A100 Tensor Core GPUs.

The testing generated a sample from a full circuit simulation of the Google Sycamore circuit in a little over nine minutes on Nvidia’s Selene supercomputer — something that would normally take days and millions of CPU cores, the company said during its recent GTC 2021 conference.

However, Nvidia CEO Jensen Huang estimated that for quantum computers to solve meaningful, real world problems would require systems containing several million physical qubits, which allows such systems to deliver adequate error correction.

“The research community is making fast progress, doubling physical qubits every year,” Huang said. “But even with that progress, we likely will not achieve that milestone [solving real world problems] until 2035 to 2040. But in the mean time we can best help the quantum researchers trying to create the computer of the future by creating the fastest [classical] system today.”

Development of applications for quantum systems has stalled of late, with one of the primary reasons being a lack of horse power available on classical systems responsible for simulating quantum circuits, according to Huang. He believes cuQuantum SDK and Nvidia A100 Tensor Core GPUs can remedy that.

One analyst somewhat disagreed with that assertion, but is nonetheless encourage by the power of Nvidia’s latest GPUs and Selene supercomputer.

“I don’t know if the lack of hardware power is necessarily slowing things down,” said Paul Goodson-Smith, senior quantum and AI Analyst for Moor Insights & Strategy. “We have enough quantum simulators out there right now from IBM and AWS to help things along, but with the new A100, the DGX systems and the SDK, they [Nvidia] are taking steps in the right direction.”

Another analyst said the new offerings look promising but without more technical details and proven use cases from Nvidia, he reserved judgement on their competitive chances in the market.

“There are a lot of companies building digital simulators out there now,” said Bob Sorensen, SVP and chief analyst for quantum computing with Hyperion Research. “But if they can do the right tweaking on the GPUs, it looks like they could be a pretty effective products. I can see right away a number of applications for them, especially AI and machine learning.”

Huang stated strongly that Nvidia has no intention of producing a full-blown quantum system, but said he believes that ultimately, GPU-accelerated platforms are the most suited best for quantum circuit and algorithm development testing.  He said there are about 50 teams around the world among vendors, academia and national laboratories building or researching quantum circuit algorithms and applications and Nvidia is working with many of them.

“We will deal with anyone doing quantum circuit simulation,” Huang said. “You need simulators to test and validate the hardware as you are building out solutions. Testing is a critical part of simulating quantum environments.”

Nvidia is working with both vendors and large organizations that are building simulators and algorithms that take advantage of quantum systems, although company officials declined to identify them.

The cuQUantum SDK consists of libraries and tools that accelerate quantum computing workflows. Developers can use the development kit to speed up quantum circuit simulations based on state vector, density matrix and tensor network methods by several orders of magnitude, Nvidia said.

Nvidia took an agnostic approach with the new SDK, allowing users to select the set of tools that best fit their use case. As one example, the state vector method — which is a set of data describing exactly where an object is located in space, and how it is moving — offers high-fidelity results, although its memory requirements also grows substantially depending on the number of qubits users have at their disposal.

As Editor At Large with TechTarget’s News Group, Ed Scannell is responsible for writing and reporting breaking news, news analysis and features focused on technology issues and trends affecting corporate IT professionals. He has also worked for 26 years at Infoworld and Computerworld covering enterprise class products and technologies from larger IT companies including IBM and Microsoft, as well as serving as Editor of Redmond for three years overseeing that magazine’s editorial content.


This is a syndicated post. Read the original post at Source link .