With NVIDIA superchips and software, supercomputers, sensors, and a developing quantum-computing ecosystem offer unprecedented possibilities.

Accelerated computing is helping researchers tackle today’s major issues, as seen at ISC, Europe’s annual assembly of supercomputing professionals.

To replicate new energy sources, some people are creating digital twins. Some researchers employ AI and high-performance computing to look into the human brain.

Others, said Ian Buck, vice president of accelerated computing at NVIDIA, are driving HPC to the edge with incredibly sensitive instruments or speeding simulations on hybrid quantum systems, at an ISC special talk in Hamburg.

Providing ten AI Exaflops

For example, at Los Alamos National Laboratory (LANL), a new supercomputer named Venado will give 10 exaflops of AI capability to boost studies in fields like materials research and renewable energy.

With NVIDIA GPUs, CPUs, and DPUs in the system, named after a mountain in northern New Mexico, LANL researchers aim for 30x speedups in their computational multi-physics applications.

NVIDIA Grace Hopper Superchips will be used by Venado to perform tasks up to three times quicker than previous GPUs. It also has NVIDIA Grace CPU Superchips, which deliver double the performance per watt of regular CPUs on a wide range of non-accelerated tasks.

BlueField is gaining traction.

The LANL system is the latest in a long line of systems that use NVIDIA BlueField DPUs to offload and speed communications and storage activities from host CPUs.

Similarly, the Texas Advanced Computing Center is expanding the NVIDIA Quantum InfiniBand network on Lonestar6 with BlueField-2 DPUs. It will evolve into a cloud-native supercomputing development platform, capable of hosting numerous users and apps with bare-metal speed while securely isolating workloads.

“For next-generation supercomputing and HPC clouds, that’s the architecture of choice,” Buck added.

In Europe, it is exascale.

NVIDIA and SiPearl are working together in Europe to enhance the ecosystem of developers working on exascale computing on Arm. The study will aid users in the region in porting programmes to SiPearl’s Rhea and future Arm-based CPUs, as well as NVIDIA accelerated computing and networking capabilities.

The University of Tsukuba’s Center for Computational Sciences is using NVIDIA H100 Tensor Core GPUs and x86 CPUs on an NVIDIA Quantum-2 InfiniBand infrastructure. Climate science, astrophysics, big data, AI, and other fields will be tackled by the new supercomputer.

The new machine will join the 71 percent of supercomputers on the recent TOP500 list that use NVIDIA technology. Furthermore, NVIDIA GPUs, networks, or both are used in 80% of new systems on the list, and NVIDIA’s networking technology is the most prevalent interconnect for TOP500 systems.

NVIDIA solutions are chosen by HPC users because they provide the best application performance for both traditional supercomputing workloads like simulation, machine learning, and real-time edge processing, as well as new workloads such as quantum simulations and digital twins.

Boosting Your Power With Omniverse

Buck demonstrated the capabilities of these technologies by demonstrating a virtual fusion power station being built in NVIDIA Omniverse by researchers from the UK Atomic Energy Authority and the University of Manchester. The goal of the digital twin is to mimic the full power station in real time, including its robotic components and even the behavior of the fusion plasma at its core.

NVIDIA Omniverse, a 3D design collaboration and world simulation platform, allows project participants to collaborate in real time while utilizing various 3D apps. They want to use NVIDIA Modulus, a platform for constructing physics-informed AI models, to improve their work.

Buck stated, “It’s really sophisticated work that’s setting the road for tomorrow’s clean renewable energy sources.”

Medical Imaging using Artificial Intelligence

Buck also highlighted how researchers used NVIDIA Cambridge-1, a supercomputer dedicated to AI advancements in healthcare, to construct a library of 100,000 synthetic pictures of the human brain.

King’s College London researchers utilized MONAI, an AI framework for medical imaging, to create lifelike pictures that can aid researchers in understanding how illnesses like Parkinson’s evolve.

“This is a fantastic example of HPC+AI bringing genuine value to the scientific and research community,” Buck added.

HPC on the Cutting Edge

HPC activity is increasingly extending beyond the supercomputer facility. Observatories, satellites, and new types of lab gear all require real-time data streaming and visualization.

For example, at Lawrence Berkeley National Lab, researchers are utilizing NVIDIA Clara Holoscan to view life in real time at the nanoscale scale, a task that would take several days on a CPU.

NVIDIA is creating Holoscan for HPC, a highly scaled version of our imaging software that can speed up any scientific discovery, to help deliver supercomputing to the edge. From Jetson AGX modules and appliances to quad A100 servers, everything will run on accelerated platforms.

“We’re excited to see what researchers develop with this software,” Buck added.

Quantum Simulations at a Higher Speed

Buck reported on the growing acceptance of NVIDIA cuQuantum, a software development kit for accelerating quantum circuit simulations on GPUs, in yet another supercomputing vector.

It is already being used in research by dozens of organizations in a variety of sectors. It’s included into leading quantum software frameworks, allowing users to take advantage of GPU acceleration without having to write any new code.

AWS just announced that cuQuantum is now available in its Bracket service. It also showed how cuQuantum can speed up quantum machine learning workloads by up to 900x while lowering expenses by 3.5x.

“Modeling quantum computers on GPU supercomputers is crucial to moving us closer to practical quantum computing,” Buck added. “We’re really thrilled to be at the vanguard of this work,” he added.

Watch the complete session below to learn more about accelerated computing for HPC.Quantum Computing