Samsung Electronics cut simulation time roughly 1,600-fold in one chip engineering case by combining artificial intelligence, a simpler computational grid and graphics processors, NVIDIA said in a Nov. 9 blog post.
NVIDIA, which supplies computing platforms used in the collaborations, described the result in “Korea Advances Physical AI Across Industries With NVIDIA,” published Nov. 9. The company said its Korean partners are moving AI systems that interact with the physical world from development into production.
The company said Samsung’s simulation took 41 seconds, compared with about 18 hours for a calculation using a finer grid on a 32-core central processing unit. The faster calculation combined AI predictions of material properties with a coarser grid and was sped up by graphics processors.
NVIDIA presented the result as one case within Samsung’s development of tools for analysing heat and mechanical behaviour at chip scale using NVIDIA PhysicsNeMo. The comparison changed both the computational grid and the computing hardware, alongside the use of AI-predicted properties.
Samsung’s broader collaboration covers materials research, manufacturing process development and device engineering, NVIDIA said. It includes work with NVIDIA cuISS and cuDSS to speed up calculations involving sparse linear systems, mathematical problems used in engineering simulations, alongside process and optical simulations.
In materials research, Samsung’s sMOL platform completed selected quantum chemical calculations up to five times faster from start to finish using NVIDIA cuEST, the company said. Samsung is also exploring cuEquivariance to speed up calculations used to train and run AI models of atoms.
NVIDIA said its Agent Toolkit could eventually connect simulation tools and AI models so engineering agents could run analyses and compare designs. It described that capability as a future possibility.
In autonomous driving, NVIDIA said Kakao Mobility plans to combine its driving software and experience operating in Seoul’s Gangnam district with NVIDIA’s vehicle computing technology. The planned work targets a level 4 autonomous driving system.
Kakao Mobility plans to use the Hyperion sensor and computing reference platform, powered by DRIVE AGX Thor, and Halos technology for safety during development and validation, NVIDIA said. Initial development is planned around Kia purpose-built vehicles, with future mobility services in Korea as the aim.
NVIDIA also said it and LG Electronics demonstrated robotic handling and manufacturing tasks with LG’s CLOiD robot in an actual manufacturing environment. The work connected real-world data collection, Cosmos-based data expansion, GR00T model training, Jetson AGX Thor onboard computing and simulation-based evaluation.
The teams are using that work to develop a repeatable process for training, evaluating and improving manufacturing robots, NVIDIA said.
