Supercomputing event SC21 is being held in St. Louis this week, and two major semiconductor players, Nvidia and Xilinx, to the event as an opportunity to showcase new high-performance computing use cases that leverage their latest products.
Nvidia
Nvidia announced a partnership with Atos to create the Excellence AI Lab (EXAIL) to enable European researchers from various fields to leverage high performance computing and AI on projects ranging from climate research to healthcare and genomics to hybridization with quantum computing,
Atos will develop an exascale-class BullSequana X supercomputer with Nvidia’s Arm-based Grace CPU, Atos BXI Exascale Interconnect and NVIDIA Quantum-2 InfiniBand networking platform, which was announced just last week.
“If you are looking at the new use cases and applications being developed, AI is going everywhere,” said Gilad Shainer, vice president of networking at Nvidia. “Every company in the world is going to use AI, and that means every company is going to need access to high-performance computing. We’ll see more users, we’ll see more apps and we’ll see more jobs sharing the same HPC infrastructure.”
In the field of researching climate change, Atos and Nvidia also plan to run new AI and deep learning models at Germany’s Jülich Supercomputing Center on what they described as Europe’s fastest supercomputer. These models can be used to predict the evolution of extreme weather events and their changing behavior due to global warming, and they will benefit greatly from exascale-class computing, according to Nvidia.
The JUWELS Booster system, based on Atos’ BullSequana XH2000 platform, with over 2 exaflops of AI and 3,744 NVIDIA A100 Tensor Core GPUs and NVIDIA Quantum InfiniBand networking, will help provide deeper understanding of climate change and more accurate long-term predictions of events, such as hurricanes, extreme precipitation, and heat and cold waves.
Nvidia also said EXAIL will harness Atos’ advanced computing solutions and NVIDIA Clara to help healthcare researchers and providers accelerate drug discovery and design advanced diagnostic solutions using embedded, edge, data center and cloud platforms.
In the fast-emerging quantum computing field, the Atos Quantum Learning Machine, a quantum software development and simulation appliance, will use Nvidia GPUs to help dramatically increase the speed and scale of quantum simulations, which will in turn speed the research in quantum algorithms, quantum information science, new quantum processor architectures and hybrid quantum-GPU system architectures.
Xilinx
Xilinx announced its U55C data center accelerated computing card at SC21, which Nathan Chang, HPC product manager at Xilinx, described as addressing the growing need for accelerated computing via smaller form factor, more power efficient cards, rather than expanding use of power-hungry servers.
“Scale and compute is no longer about adding servers,” he said. “It’s a lot of waste power. This paves the way for accelerator cards.” The U55C fits in a single-slot, making its dual-slot U280 predecessor look like a relative space hog. It also has double the high-bandwidth memory of the U280 and lower maximum power draw of 150 watts.
Xilinx highlighted three examples of how the U55C is being put to work in real-world use cases:
- CSIRO, Australia’s national research organization, which also has the world’s largest radio astronomy antenna array, is using Alveo U55C cards for signal processing in the Square Kilometer Array radio telescope. Deploying the Alveo cards as network-attached accelerators with HBM allows for massive throughput at scale across the HPC signal processing cluster. The Alveo accelerator-based cluster allows CSIRO to tackle the massive compute task of aggregating, filtering, preparing and processing data from 131,000 antennas in real time. The 460Gbps of HBM2 bandwidth across the signal processing cluster is served by 420 Alveo U55C cards fully networked together across P4-enabled 100 Gbps switches. “Out in the desert where they’re deploying, you’ve got power and space concerns, and you need to get as many cards into a single node as possible,” Chang said.
- Ansys, whose LS-DYNA crash simulation software is used by nearly every automotive company in the world, also is using Alveo cards. The design of safety and structural systems hinges on the performance of models as they mitigate the costs of physical crash testing with computer-aided design finite element method (FEM) simulations. FEM solvers are the primary algorithms driving simulations with hundreds of millions of degrees of freedom, these enormous algorithms can be broken out into more rudimentary solvers like PCG, Sparse matrices, ICCG. By scaling out across many Alveo cards with hyperparallel data pipelining, LS-DYNA can accelerate performance by more than 5X in comparison to x86 CPUs. This results in more work per clock cycle in an Alveo pipeline with LS-DYNA customers benefiting from game changing simulation times.
- TigerGraph, provider of a leading graph analytics platform, is using multiple Alveo U55C cards to cluster and accelerate the two most prolific algorithms that drive graph-based recommendation and clustering engines. “Graph databases help you get rid of your silos and make data a first-class citizen by looking at the relationships between data sets,” Chang said. “Now there’s a need for real-time processing.” The Alveo U55C accelerates query times and predictions for recommendation engines from minutes down to milliseconds. By utilizing multiple U55C cards to scale up analytics, the superior computational power and memory bandwidth accelerates graph query speeds up to 45X faster compared to CPU-based clusters. The quality of scores also increases by up to 35%, resulting in greater confidence dramatically lowering false positives to low single digits.