System Bits: May 14

Faster U.S. supercomputers on the way
The U.S. Department of Energy awarded a contract for more than $600 million to Cray for an exascale supercomputer to be installed at the Oak Ridge National Laboratory during 2021. Cray will provide its Shasta architecture and Slingshot interconnect for what is dubbed the Frontier supercomputer.

Advanced Micro Devices will have a key role in building the supercomputing system, which is expected to achieve performance of greater than 1.5 exaflops. The chip company will supply AMD EPYC CPUs and AMD Radeon Instinct graphics processing units for the project, along with its custom high-bandwidth, low-latency coherent Infinity Fabric, connecting four Radeon Instinct GPUs to one EPYC CPU per node, and an enhanced version of the open-source ROCm programming environment, developed with Cray.

Frontier is expected to be the fastest, most powerful supercomputer system in the world when it is delivered and installed at the federal lab in Tennessee. The second-generation artificial intelligence-based system will be tackling big problems in deep learning, machine learning, and data analytics for applications in human health, manufacturing, and other fields.

Frontier will join ORNL’s Jaguar, Titan, and Summit supercomputing systems in advanced research and development.

The new system will boast more than 100 of Cray’s Shasta cabinets with high-density compute blades powered by the AMD CPUs and GPUs, utilizing AI-optimized and high-performance computing technology. Each node on the system will have a Cray Slingshot interconnect network port for every GPU in the supercomputer.

ORNL’s Center for Accelerated Application Readiness is now accepting proposals from scientists to prepare their codes to run on Frontier, with a deadline of June 8. More information is offered at the Frontier website.

In March, the Department of Energy and Intel announced the building of the Aurora exascale computer, which will be installed at the Argonne National Laboratory in 2021. Argonne is operated by the University of Chicago.

The Aurora contract is valued at more than $500 million, with Cray working as a subcontractor to Intel. Aurora will have more than 200 Shasta cabinets. The Cray software stack will be optimized for Intel architectures and Cray’s Slingshot interconnect technology.

Advances in quantum communication and computing
Researchers at the University of Chicago’s Institute for Molecular Engineering report a pair of breakthroughs in quantum technology – entangling two quantum bits using sound for the first time and creating the highest-quality long-range link between two qubits to date.

“Both of these are transformative steps forward to quantum communications,” said co-author Andrew Cleland, the John A. MacLean Sr. Professor of Molecular Engineering at the IME and Argonne National Laboratory. A leader in the development of superconducting quantum technology, he led the team that built the first “quantum machine,” demonstrating quantum performance in a mechanical resonator. “One of these experiments shows the precision and accuracy we can now achieve, and the other demonstrates a fundamental new ability for these qubits.”

In a study published April 22 in the Nature Physics journal, Cleland’s lab was able to build a system out of superconducting qubits that exchanged quantum information along a track nearly a meter long with extremely strong fidelity—with far higher performance has been previously demonstrated.

“The coupling was so strong that we can demonstrate a quantum phenomenon called ‘quantum ping-pong’—sending and then catching individual photons as they bounce back,” said Youpeng Zhong, a graduate student in Cleland’s group and the first author of the paper.

The other study, published April 26 in Science, shows a way to entangle two superconducting qubits using sound.

A challenge for scientists and engineers as they advance quantum technology is to be able to translate quantum signals from one medium to the other. For example, microwave light is perfect for carrying quantum signals around inside chips. “But you can’t send quantum information through the air in microwaves; the signal just gets swamped,” Cleland said.

The team built a system that could translate the qubits’ microwave language into acoustic sound and have it travel across the chip—using a receiver at the other end that could do the reverse translation.

It required some creative engineering: “Microwaves and acoustics are not friends, so we had to separate them onto two different materials and stack those on top of each other,” said Audrey Bienfait, a postdoctoral researcher and first author on the study. “But now that we’ve shown it is possible, it opens some interesting new possibilities for quantum sensors.”

Meanwhile, researchers report using an IBM quantum computer to turn back time. They undid the aging of one simulated elementary particle by one-millionth of a second. The team acknowledges that this process is unlikely to naturally occur.

“We demonstrate that time-reversing even ONE quantum particle is an unsurmountable task for nature alone,” Valerii M. Vinokur of Argonne National Laboratory told The New York Times. He is part of a research team led by Gordey B. Lesovik of the Moscow Institute of Physics and Technology.