SEMIFIVE announces mass production of 14nm AI inference SoC platform with Mobilint

In a significant development for the semiconductor industry, SEMIFIVE, renowned for its design solutions and pioneering platform-based custom silicon solutions, has declared the commencement of mass production for its 14nm AI Inference SoC Platform. This venture is in collaboration with its leading partner, Mobilint, an innovative startup specializing in AI hardware and software solutions based in South Korea. This collaboration marks the third commercialization of SEMIFIVE’s SoC Platform solutions, leveraging Samsung Foundry’s mass production proven FinFET process technologies.

The product at the heart of this announcement, ARIES, is an 80 TOPS custom AI inference chip developed by SEMIFIVE using its proprietary methodology and domain-specific platform architecture. Delivered silicon samples to Mobilint showcase the capability of ARIES, which has been rigorously tested with more than 200 open source deep learning models, including state-of-the-art models. Designed for cutting-edge vision applications, edge servers, and hyperscale data centers, ARIES aims to redefine performance standards in its segment.

Dongjoo Shin, CEO of Mobilint, expressed enthusiasm over the launch, stating, “We are thrilled with the successful launch of ARIES, our first-generation 14nm high-performance AI accelerator chip for Edge and Cloud Computing applications. Thanks to SEMIFIVE’s unique SoC Platform technology coupled with their excellent packaging and bring up expertise, we were able to meet key milestones and technical specifications of ARIES, which will be a crucial solution in this era of intensive AI applications like advanced robotics, smart city infrastructure, and personalized healthcare diagnostics.”

SEMIFIVE’s AI Inference SoC Platform is touted as the ideal approach to realizing custom AI chips such as Application Specific Integrated Circuits (ASICs) for various end applications, including data center accelerators and AI vision processors. The platform is pre-verified and integrated, featuring a quad-core high-performance 64-bit CPU, 8 lanes of PCIe Gen4, and 4-channel LPDDR4 interfaces, among other specifications. This flexibility allows for customer-specific IPs, like an AI Neural Processing Unit (NPU), to be added, facilitating feature differentiation.

Brandon Cho, CEO and Co-founder of SEMIFIVE, highlighted the company’s pride in their collaboration with Mobilint, which underscores the rapid commercialization of AI semiconductors. “As a leading DSP within the Samsung Foundry SAFETM Ecosystem, SEMIFIVE is committed to serving as the gateway for various clients, facilitating the production of AI semiconductors in the most efficient manner possible,” Cho added.

Adding to the discourse, Gibong Jeong, Executive Vice President and head of Foundry Business Development team at Samsung Electronics, pointed out the importance of AI applications as a growth segment for Samsung Foundry’s business. He lauded SEMIFIVE for successfully commercializing Mobilint’s ARIES chip using Samsung’s 14nm FinFET technology.

This partnership and the subsequent launch of the ARIES chip underscore a significant leap forward in the AI and semiconductor industries. By leveraging the strengths of SEMIFIVE’s SoC Platform and Samsung Foundry’s advanced process technologies, Mobilint is positioned to make substantial impacts across various AI-intensive applications. This collaboration not only exemplifies the synergy between design and manufacturing excellence but also signals the burgeoning demand for specialized AI inference solutions in a world increasingly reliant on smart technologies.