Samsung Unveils Large-Scale Computing System Featuring Almost 100 AMD Instinct MI100 GPUs With PIM Technology
PIM chips integrate a processor with RAM onto an individual chip. This technological advancement is anticipated to assist with the large-scale performance of AI workloads. In October, Samsung revealed that the company demonstrated how a PIM chip’s performance could be combined with an AMD graphics processing unit. When the company trained the language model algorithm T5 with this next-gen breakthrough, the performance increased over twice as much while lowering the power consumption in a system without PIMs present. Now we have the world’s first large-scale computing system that houses 96 AMD Instinct MI100 GPUs that are infused with PIM ‘Processing-In-Memory’ technology, allowing for a boost in the AI algorithms. Artificial intelligence for science and technology has found acceptance in today’s world, as well as showing broad support in the semiconductor production processes. According to Choi, the company is also exploring methods to integrate the knowledge possessed by the company’s engineers with inspection data and find a use for it in production. — Choi Chang-kyu, Vice President and Head of the AI Research Center, Samsung Electronics Advanced Institute of Technology The Artificial Intelligence Semiconductor Future Technology Conference brought organizations to the Korea conference site, such as the AI Semiconductor Forum, the Next Generation Intelligence Semiconductor Foundation, the PIM Artificial Intelligence Semiconductor Project Group, as well as demonstrations in the industry, such as DEEPX, Eyenix, ETRI, Furiosa AI, Nextchip, Nepes, Openedges Technology, SAPEON Korea, and Telechips. The current method of checking yields is limited, as testing can only be done every three to six months when wafers are put into and taken out of the fab. We need to move on to a stage where yields can be predicted using AI sensors and inspection data. News Source: Business Korea
