SK vs. Samsung clash in 'PIM'... Focus on 'PIM', the next-generation market in the AI era
Published2024.12.03 16:46AI 모델 발전에 따라 메모리 인텐시브 영역에서 혁신의 병목현상이 발생하고 있는 가운데 HBM에 이은 PIM 반도체에서 차세대 메모리 시장 격돌이 예고되고 있다.
관련뉴스
AI Semiconductor Hidden Card, Take the Lead in 'PIM'
PIM (Process In Memory, intelligent semiconductor) is a chip that adds processor functionality for computational operations to memory semiconductors that perform storage tasks. Related research by the government and industry is still in its early stages. On the 13th, the '2022 PIM Artificial Intelligence Semiconductor Strategy and Technology Symposium' hosted by the PIM Artificial Intelligence Semiconductor Business Division was held for the first time this year at the COEX Conference Hall in Seoul.
2022.12.14by 김예지 기자
[Planning-Chat GPT3③] AI Semiconductor and Data Center Market ‘Active’
Investment in key AI market infrastructure, including domestic and international AI semiconductors and data centers, is being accelerated. Samsung Electronics has been continuously advancing 'HBM-PIM,' which combines memory semiconductors and AI processors into one unit, since 2021. SK Hynix has succeeded in developing 'HBM3,' a next-generation stacked memory optimized for AI inference and learning performance. Data centers provide cloud computing services capable of collecting, storing, and analyzing big data, and are well-suited for training AI models. Domestically, investment in data centers for securing stable infrastructure is accelerating alongside AI semiconductor development.
2023.03.06by 김예지 기자
KAIST develops Korea's first 'triple mode cell' DRAM-PIM semiconductor
과학기술정보통신부는 한국과학기술원(KAIST) 유회준 교수 연구팀이 국내 최초로 DRAM 메모리 셀 내부에 직접 연산기를 집적해 AI 연산을 수행하는 PIM 반도체인 ‘다이나플라지아(DynaPlasia*)’를 개발했다.
2023.03.15by 김예지 기자
.jpg)
Next-generation AI semiconductor PIM, "LLM demand↑, PIM integration is a must"
As AI semiconductor demand increases, the semiconductor industry's future technology agenda is concentrating on AI. PIM semiconductors are emerging as a solution to overcome the computational limitations of conventional GPUs for AI operations.
2023.10.31by 명세환 기자
Next-generation memory in the on-device AI era, PIM-LPDDR6 combination expected in mobile
Generative AI has triggered increased attention on SK Hynix's HBM due to AI server expansion and datacenter infrastructure upgrades. In this context, assessments are emerging that memory is failing to keep pace with advances in CPU and GPU performance used in AI computing, creating bottlenecks, while pressure is intensifying for a paradigm shift toward next-generation memory solutions.
2024.11.19by 권신혁 기자
