THE INVESTOR

메뉴열기
July 18, 2024

Rebellions' AI chip to undergo IBM quality testing

PUBLISHED : September 26, 2023 - 09:32

UPDATED : September 26, 2023 - 09:32

  • 본문폰트 작게
  • 본문폰트 크게
  • 페이스북
  • sms

From left: Kaoutar El Maghraoui, a principal research scientist at IBM Research AI, John Rozen, a program director at AI Hardware Center, and Oh Jin-wook, chief technology officer and co-founder of Rebellions, pose for a photo at IBM Research's AI Hardware Center in Albany, New York. (Rebellions)

Korean AI semiconductor startup Rebellions said Monday that its chips were set to undergo quality testing at data centers of IBM Research's AI Hardware Center in New York.

IBM Research's AI Hardware Center conducts exhaustive quality testing for enterprise-grade data center processors. It evaluates the reliability of the chip until it reaches an acceptable level of performance.

Earlier this year in April, Rebellions gained global attention when its Atom chip outperformed Nvidia's GPUs and Qualcomm's NPUs in the MLPerf Inference v3.0, an industry-standard benchmark for AI chips' inference tasks.

The Atom chip, an inference-only processor, is specifically designed for AI inference, where trained AI models generate predictions from new data. Such chips are ideal for real-time response applications, such as chatbots.

Established in 2019, IBM Research's AI Hardware Center is dedicated to AI-semiconductor developments, especially in AI training and inference. It has collaborated with top-tier firms in various AI sectors, including chip design and software, to cultivate a comprehensive AI ecosystem. Among its 15 global collaborators are industry leaders like Samsung Electronics, Synopsys, and Applied Materials.

Dr. Oh Jin-wook, chief technology officer and co-founder of Rebellions, who was previously affiliated with the IBM Thomas J. Watson Research Center, played a key role in brokering this collaboration.

"As generative AI applications surge, our collaboration with IBM is key to delivering reliable products for growing demands in fields like LLM, notably around inference," said Oh.

By Moon Joon-hyun (mjh@heraldcorp.com)

  • 페이스북
  • sms
최상단으로
COPYRIGHT HERALD CORPORATION. ALL RIGHTS RESERVED.