PFN and SEMIFIVE Join Forces to Release Cutting-Edge AI Inference Chips

PFN and SEMIFIVE Collaborate on Next-Generation AI Chips



In a groundbreaking partnership, Preferred Networks, Inc. (PFN) and SEMIFIVE Inc. announced their collaboration to develop the MN-Core L2000 Series of next-generation AI inference chips. This advancement is set to supersede the existing MN-Core L1000 Series and aims to deliver over double the memory bandwidth, significantly enhancing AI inference capabilities. The collaboration emphasizes high-performance design, efficiency, and speed, critical in today’s AI-driven landscape.

The Future of AI Infrastructure



The MN-Core L2000 Series is designed to tackle limitations faced by previous generations, with an emphasis on better memory bandwidth, which is a pivotal factor for generative AI performance. PFN has been utilizing its MN-Core architecture since 2016, refining it through various applications in its own services. The L2000 Series will incorporate 3D-stacked structures that combine logic chips and DRAM (Dynamic Random Access Memory). This innovative design enables higher efficiency and enhances the speed of AI inference, presenting a new avenue for AI infrastructure development.

At the forefront of this venture, PFN is accountable for processor architecture and logical designs, leveraging the insights gained from its decade-long experience in chip development. The collaboration with SEMIFIVE will enhance the Series through the contributions of SEMIFIVE’s expertise as a turnkey solution provider. This partnership aims to overcome various technical challenges tied to weighty chip designs, including heat dissipation, power consumption, and manufacturing yield.

Technical Innovations and Challenges



SEMIFIVE's objective to utilize a 3D-IC design closely integrates with the logic chips' requirements for optimal power management and data transfer efficiencies. The transition towards 3D design structures represents a shift in AI chip competitions, prioritizing factors beyond mere computational speed. As the CEO of SEMIFIVE, Brandon Cho, explained, 3D-IC technology is instrumental in breaking through the performance limitations currently faced by AI silicon.

As the MN-Core L2000 moves through its developmental phases, notable milestones have been set, including a tape-out scheduled for the first quarter of 2027. The implications of this project extend beyond technical advancements; it positions PFN and SEMIFIVE at the helm of the innovative AI chip market, setting benchmarks for future AI architectures.

The upcoming MN-Core L2000 Series will be displayed at the SC26 event in Chicago, Illinois, where attendees can learn more about its capabilities and the advantages it presents over previous models.

Collaborating for Future Success



Both companies see this collaboration as a pivotal step towards reshaping the AI chip landscape. PFN aims to leverage its unique technology to fortify its contributions to the field of AI, while SEMIFIVE’s extensive design capabilities will help bring the latest advancements in AI chips to fruition.

The integration of Samsung Foundry's technology into this project also hints at a robust future for 3D-IC design and the development of high bandwidth, energy-efficient solutions necessary for the next generation of AI applications.

In conclusion, the MN-Core L2000 series symbolizes a significant advancement in AI technology, combining innovative design, increased efficiency, and the promise of vast improvements in AI inference capabilities. As the launch approaches, the tech world watches closely for the results of this promising collaboration between PFN and SEMIFIVE.

Topics Consumer Technology)

【About Using Articles】

You can freely use the title and article content by linking to the page where the article is posted.
※ Images cannot be used.

【About Links】

Links are free to use.