Today we announced that the U.S. Department of Energy’s Argonne National Laboratory will expand its AI infrastructure by deploying SambaNova Suite, our platform that combines hardware and software for fine-tuning and inference in scientific research.
Argonne will deploy the platform for use by the scientific community as part of its AI Testbed at the Argonne Leadership Computing Facility (ALCF) operated by the DOE Office of Science. It will join existing SambaNova DataScale systems to provide advanced AI capabilities to researchers and scientists in areas including energy and climate research, life sciences, and national security.
“Inferencing large language models and foundation models is crucial to our efforts to apply AI to complex scientific problems,” said Rick Stevens, Argonne’s associate laboratory director for Computing, Environment and Life Sciences. “The addition of the new SambaNova system aligns with our mission to explore how novel AI accelerator platforms can benefit and advance science.”
SambaNova Suite is a fully integrated hardware-software platform that combines our DataScale SN40L, with software and models to enable organizations to train, fine-tune, and deploy AI workloads. Powered by the world’s most efficient AI silicon, our Reconfigurable Dataflow Unit, it is optimized for low latency and high throughput inference.
“A dramatic advancement of AI workloads going into production for inferencing has begun,” said Marshall Choy, SVP of Product at SambaNova Systems. “Argonne National Laboratory is leading the way in delivering fast inference services for AI for Science, we are pleased to support their efforts through our longstanding partnership and this new system deployment."
The system’s capabilities will support the development of large foundation models like Argonne’s AuroraGPT, which is being built to enable autonomous scientific exploration across disciplines, including biology, chemistry, materials science, and climate modeling. AuroraGPT is being trained on Argonne’s Aurora exascale system, one of the world’s most powerful supercomputers, which earlier this year broke the exascale barrier of 1 quintillion calculations per second.
Beyond advancing the use of AI for science, the systems in Argonne’s AI Testbed are allowing the ALCF and its user community to gain insights into how AI accelerators could be integrated with next-generation supercomputers to boost performance and efficiency.
“Our AI Testbed enables the ALCF user community to leverage novel AI technologies for innovative research projects involving large language models, large-scale data analysis, and the development of trustworthy AI,” said Michael Papka, director of the ALCF. “With the deployment of the new DataScale SN40L system, we're extending advanced AI inference capabilities beyond our traditional ALCF user base. By making trained AI models more accessible, we aim to empower a wider community of researchers to explore new directions in generative and agentic AI workloads for science and engineering.”
“Inference is one of the largest workloads for our AI Testbed systems,” added Venkat Vishwanath, AI and machine learning lead at the ALCF. “Being able to rapidly evaluate AI models and adjust parameters for improved performance is crucial for driving progress in AI-driven science across many research areas, including drug discovery, climate science, and brain mapping.”
The RDU allows for fast switching between models. It also consumes about one-tenth the power of a typical GPU-based system. Both capabilities will factor in Argonne’s ongoing evaluation of how AI can be used to enhance scientific research.
“The ability to switch between different AI models instantly and fine-tune them using domain-specific datasets can help streamline the process of testing and validating their performance,” Vishwanath said. “By reducing the time needed for each inference cycle, we can accelerate the evaluation of AuroraGPT and other large-scale models.”
The system also gives the lab a new platform to continue its explorations into energy-efficient technologies for next-generation supercomputers and data centers.
“Both supercomputers and AI model development and evaluation have substantial energy demands,” Vishwanath said. “One of our goals with the ALCF AI Testbed is to determine how novel AI accelerators like the SN40L could be integrated with future supercomputers to enhance energy efficiency.”