Samsung Electronics says a chip simulation that once took about 18 hours now finishes in 41 seconds. The result comes from its expanded work with NVIDIA on accelerated computing and AI physics.
The figure was shared on November 8 in an NVIDIA blog post on Korea’s physical AI efforts. The post described three Korean companies building AI systems for the physical world. It also covered new projects with Kakao Mobility and LG Electronics.
NVIDIA said Korean companies are using accelerated computing, AI and simulation in chipmaking, robotics and self-driving cars. The company said these efforts are moving physical AI from development into real-world production.
Samsung Reports Faster Chip Engineering Work
Samsung Electronics is expanding its collaboration with NVIDIA to push semiconductor work forward. The effort covers materials research, process development and device engineering, using several NVIDIA software tools.
The biggest speed gain came from a chip-scale thermal and mechanical analysis built with NVIDIA PhysicsNeMo. In one AI homogenization case, Samsung combined AI-predicted material properties, coarser grids and GPU acceleration.
The result was a speedup of about 1,600 times. A fine-grid solve on a 32-core CPU took about 18 hours, while the GPU-accelerated coarse-grid solve took 41 seconds.
In materials research, Samsung’s sMOL platform ran selected quantum chemical calculations up to 5 times faster end to end. That work used NVIDIA cuEST.
Samsung is also using NVIDIA cuISS and cuDSS in technology computer-aided design. These tools speed up math involving sparse linear systems. The work also extends to GPU acceleration for process and optical simulations.
The company is exploring NVIDIA cuEquivariance to speed up tensor operations. These operations are used in both training and inference for atomistic AI models, which study materials at the level of atoms.
NVIDIA said its Agent Toolkit could later connect these simulation tools and AI models. That could help future engineering agents run analyses and compare design options.
Kakao Mobility and LG Electronics Expand Physical AI Work
Kakao Mobility plans to use NVIDIA technology to develop a level 4 autonomous driving system. The project pairs Kakao Mobility’s own driving software and its experience in Seoul’s Gangnam district with NVIDIA’s vehicle computing platform.
The company plans to use NVIDIA Hyperion, a reference platform for vehicle sensors and compute powered by NVIDIA DRIVE AGX Thor. It will also use NVIDIA Halos to support safety during development and testing.
Early development will center on Kia purpose-built vehicles, with Kakao Mobility’s software integrated with NVIDIA technologies. The goal is to support future level 4 mobility services in Korea.
NVIDIA and LG Electronics showed LG’s CLOiD robot handling manipulation and manufacturing tasks in a real factory setting. The robot ran on NVIDIA Isaac GR00T technology.
The companies built what NVIDIA calls a physical AI data flywheel. It links real-world data collection, NVIDIA Cosmos-based data augmentation, GR00T model training, NVIDIA Jetson AGX Thor onboard computing and simulation-based testing.
NVIDIA said the setup created reusable workflows for developing and checking CLOiD’s skills. The teams are now using the work to refine a repeatable process for training, evaluating and improving robots for manufacturing.
