Collaboration with main Japanese AI firm will produce cutting-edge AI accelerator chips
Samsung Electronics, a world chief in superior semiconductor expertise, immediately introduced that it’s going to present turnkey semiconductor options utilizing the 2-nanometer (nm) foundry course of and the superior 2.5D packaging expertise Interposer-Cube S (I-Cube S) to Preferred Networks, a number one Japanese AI firm.
By leveraging Samsung’s modern foundry and superior packaging merchandise, Preferred Networks goals to develop highly effective AI accelerators that meet the ever-growing demand for computing energy pushed by generative AI.
Since beginning mass manufacturing of the trade’s first 3nm course of node making use of Gate-All-Around (GAA) transistor structure, Samsung has strengthened its GAA expertise management by efficiently successful orders for the 2nm course of with additional upgrades in efficiency and energy effectivity.
The cooperation with Preferred Networks marks the primary achievement for Japanese corporations within the area of large-sized heterogeneous built-in bundle applied sciences and Samsung plans to speed up its main international superior bundle market offensive.1
The 2.5D Advanced Package I-Cube S expertise included within the turnkey options, is a heterogeneous integration bundle expertise, with a number of chips in a single bundle to boost inter-connection velocity and scale back bundle dimension. The use of the silicon interposer (Si-interposer) is essential in attaining ultra-fine redistribution layer (RDL) and stabilizing energy integrity for optimum semiconductor efficiency. GAONCHIPS, a specialised system semiconductor improvement firm, designed the chip.
“We are excited to lead in AI accelerator technology with Samsung Electronics’ 2nm GAA process. This solution will significantly support Preferred Networks’ ongoing efforts to build highly energy-efficient, high-performance computing hardware that meets the ever-growing computing demands from generative AI technologies, especially large language models.,” stated Junichiro Makino, VP and Chief Technology Officer (CTO) of Computing Architecture at Preferred Networks.
“This order is pivotal as it validates Samsung’s 2nm GAA process technology and Advanced Package technology as an ideal solution for next-generation AI accelerators,” stated Taejoong Song, Corporate VP and the top of Foundry Business Development Team at Samsung Electronics. “We are committed to closely collaborating with our customers ensuring that the high performance and low power characteristics of our products are fully realized.”
Preferred Networks, headquartered in Tokyo, develops superior software program and {hardware} applied sciences by vertically integrating the AI worth chain from chips to supercomputers and generative AI basis fashions. It offers options and merchandise for varied industries resembling manufacturing, transportation, healthcare, leisure and training. The firm is among the main gamers within the international AI market, attaining first place thrice previously 5 years on the Inexperienced5002 checklist of supercomputers.
Based on this collaboration, Samsung and Preferred Networks plan to showcase groundbreaking AI chiplet options for the next-generation knowledge heart and generative AI computing market sooner or later.3
1 Heterogeneous integration packaging expertise integrates several types of chips – resembling reminiscence, logic and sensors – in a single bundle.
2 Inexperienced500: the rating of the TOP500 supercomputers on the earth primarily based on their performance-per-watt effectivity
3 Chiplet Solution: A expertise of assembling right into a single bundle with different completely different chips performing completely different fuctions by means of the packaging applied sciences
