Looking for Ways to Share New Computing Technology
Samsung Electronics announced on Dec. 12 that it has developed the world’s first large-scale computing system by connecting graphics processing units (GPUs) loaded with processing-in-memory (PIM) chips and is looking for ways to share it with academia and partners.
Vice president Choi Chang-kyu, head of the AI Research Center at Samsung Electronics Advanced Institute of Technology, disclosed the development of the new computing technology through a keynote speech at the 2022 Artificial Intelligence (AI) Semiconductor Future Technology Conference, which was hosted by the Ministry of Science and ICT.
“We have built a large computing system by combining 96 GPUs (MI100) from AMD of the United States, with each of them loaded with a PIM chip. It is the only one of its kind in the world," Choi said.
In October, Samsung announced that it succeeded in demonstrating the performance of a processing in memory (PIM) chip by installing it on a graphics processing unit (GPU) accelerator card from AMD.
PIM refers to the integration of a processor with random access memory (RAM) on a single chip. This technology is expected to contribute to improving the performance of mammoth-sized artificial intelligence (AI).
When Samsung Electronics trained the language model algorithm T5 with this system, its performance improved by 2.5 times and its power consumption dropped by 2.67 times, compared to when PIMs were not used.
Choi explained that the use of AI is increasing in semiconductor production processes. “As the head of the AI research center, I want to make Samsung a semiconductor company that uses AI better than any other company,” Choi added.
“The current method of checking yields has its limits, 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,” Choi said.
Samsung Electronics is also studying ways to combine the know-how contained in the notes of engineers working on site with inspection data and automatically use it in production processes, according to Choi.