Samsung, one of the , has that it has partnered with NAVER to build semiconductor chips that are optimized for hyperscale AI models. It’s a wide-reaching collaboration with NAVER, which is the leading global internet firm with top-notch AI experience.
The collaboration will leverage Samsung’s advanced products like computational storage (), CXL (), PIM (), and PNM (Processing Near Memory), and pool their hardware and software experience together to drastically improve performance during massive AI workloads.
The companies claim that recent trends around have led to exponential growth in data volumes. However, the current computational technologies have performance and efficiency limitations to properly handle those massive workloads. So, NAVER and Samsung have decided to work together and make AI-optimized semiconductor chips and other solutions.
Electronics will contribute with its hardware and manufacturing expertise, while NAVER will bring its expertise around the development and verification of AI-based algorithms. NAVER will improve HyperCLOVA, a hyperscale model with over 200 billion parameters, and its compression algorithms for better efficiency.
The post appeared first on .
The collaboration will leverage Samsung’s advanced products like computational storage (), CXL (), PIM (), and PNM (Processing Near Memory), and pool their hardware and software experience together to drastically improve performance during massive AI workloads.
The companies claim that recent trends around have led to exponential growth in data volumes. However, the current computational technologies have performance and efficiency limitations to properly handle those massive workloads. So, NAVER and Samsung have decided to work together and make AI-optimized semiconductor chips and other solutions.
Electronics will contribute with its hardware and manufacturing expertise, while NAVER will bring its expertise around the development and verification of AI-based algorithms. NAVER will improve HyperCLOVA, a hyperscale model with over 200 billion parameters, and its compression algorithms for better efficiency.
The post appeared first on .