Samsung Electronics has detailed plans to create an AI megafactory in partnership with Nvidia, aiming to combine AI throughout its manufacturing chain.
The ability will use greater than 50,000 Nvidia graphics processing items (GPUs) to embed AI capabilities into all features of semiconductor manufacturing, from design and course of administration to gear operations and high quality assurance.
Entry deeper trade intelligence
Expertise unmatched readability with a single platform that mixes distinctive knowledge, AI, and human experience.
Samsung expects the AI megafactory to allow steady evaluation and optimisation of manufacturing environments in actual time.
It plans to scale the AI manufacturing unit throughout its world manufacturing community utilizing Nvidia accelerated computing infrastructure.
The corporate may even deploy digital twin know-how utilizing Nvidia Omniverse libraries for simulating fabrication operations, anomaly detection, predictive upkeep, and manufacturing optimisation previous to implementation in bodily environments.
Websites together with Samsung’s facility in Taylor, Texas, are set to be included on this rollout.
Each companies are working with digital design automation (EDA) companions to develop GPU-accelerated EDA instruments match for next-generation design necessities.
Samsung’s proprietary AI fashions, that are constructed on Nvidia accelerated computing and the Megatron framework, are deployed throughout greater than 400 million of its units. They’re additionally built-in into its manufacturing methods for functions together with real-time translation and multilingual processing.
For robotics and automation, Samsung is leveraging platforms such because the Nvidia RTX PRO 6000 Blackwell Server Version and Jetson Thor.
The mixing connects digital simulation knowledge with real-world operations, supporting clever activity execution and security controls in robotic methods.
In course of optimisation, Samsung reported a twentyfold improve in computational lithography efficiency utilizing Nvidia cuLitho and CUDA-X libraries for optical proximity correction.
This permits for speedy AI-driven sample prediction and correction throughout chip manufacturing, lowering improvement cycles.
Samsung, together with Nvidia and South Korean telecom operators, academia and analysis establishments, is advancing AI-RAN communication know-how. This strategy embeds AI computation inside cell networks to assist robots, drones and industrial gear performing inference duties near the community edge.
2024’s proof-of-concept mixed Samsung’s software-based community with Nvidia GPU know-how for preliminary demonstrations.
Their partnership started greater than 25 years in the past with Samsung DRAM in Nvidia graphics playing cards and now consists of work on superior reminiscence similar to HBM4.
Samsung’s HBM4, constructed with 10nm-class DRAM and a 4nm logic base die, achieves processing speeds as much as 11Gbps, which exceeds present trade requirements.
Just lately, Nvidia solid a partnership with Oracle to develop an AI supercomputer at Argonne Nationwide Laboratory for the US Division of Power (DOE).
Join our day by day information round-up!
Give your enterprise an edge with our main trade insights.
