SEMIFIVE strengthens Arm partnership for enhanced SoC platform performance

SEMIFIVE, a provider of advanced design solutions specializing in custom silicon, has announced an expanded collaboration with Arm to elevate its artificial intelligence (AI) and high-performance computing (HPC) System-on-Chip (SoC) platforms. Leveraging Arm Neoverse Compute Subsystems (CSS), SEMIFIVE aims to seamlessly incorporate the high-performance capabilities of Arm’s technology into its SoC platforms, delivering state-of-the-art computing solutions to its clients.

The expansion of this partnership strategically positions SEMIFIVE as a pivotal contributor within the Arm Total Design ecosystem, empowering customers to integrate a comprehensive suite of Arm intellectual properties (IPs), including Cortex-A, Cortex-R, Cortex-M central processing units (CPUs), Mali graphics processing units (GPUs), Ethos neural processing units (NPUs), and an extensive array of subsystem IPs. This collaboration facilitates the rapid development of custom silicon by equipping SEMIFIVE with the tools necessary to accelerate innovation within AI and HPC domains.

Advancing AI Accelerator SoCs for Startups

SEMIFIVE’s Arm-based SoC platforms have empowered numerous AI startups to develop sophisticated accelerator SoCs over the past three years. This success underscores the platform’s efficiency, adaptability, and reliability, particularly in the rapid prototyping and deployment of custom silicon solutions. SEMIFIVE continues to reinforce its market position by bringing an extensive range of Arm-based solutions to a broader customer base.

SEMIFIVE will exhibit its latest advancements in AI and HPC SoC technology at the forthcoming Open Compute Project (OCP) Global Summit in San Jose later this month, offering further insights into how the partnership with Arm is shaping the future of computing.

HyperAccel’s Generative AI Chip “Bertha”: A Step Forward

In parallel with its collaboration with Arm, SEMIFIVE has recently announced a partnership with HyperAccel to develop “Bertha”—a generative AI chip slated for mass production by the first quarter of 2026. Bertha, designed using 4nm process technology, represents a significant advancement in generative AI capabilities.

HyperAccel’s Large Language Model (LLM) Processing Unit, referred to as an LPU, is tailored to efficiently manage transformer-based models, emphasizing cost-efficiency and reduced latency. The LPU is expected to surpass traditional GPUs in performance, offering up to twice the processing power while delivering a 19-fold enhancement in price-to-performance compared to existing high-cost GPU solutions. This innovative chip aims to establish new benchmarks in efficient, high-performance AI hardware.

Strategic Alliance with OPENEDGES for HPC SoC Chiplet Platform

In June 2024, SEMIFIVE formalized a Memorandum of Understanding (MOU) with OPENEDGES Technology to collaboratively develop an advanced chiplet platform that integrates OPENEDGES’ LPDDR6 memory subsystem. This subsystem incorporates a DDR memory controller and physical layer interface (PHY) IPs optimized for the 4nm process node. Through SEMIFIVE’s extensive expertise in SoC design and packaging, this partnership aspires to redefine the high-performance SoC market by making substantial strides in cost efficiency, development flexibility, and performance enhancement.

Pioneering a New Era of Chiplet Platforms

SEMIFIVE’s chiplet platform embodies a novel approach to semiconductor design, prioritizing modularity and scalability. This initiative is poised to generate significant advancements in cost efficiency, performance optimization, and accelerated development timelines, thus positioning SEMIFIVE to drive semiconductor innovation and meet growing demands for custom AI silicon.

Ongoing Collaboration and Future Prospects

The enduring partnership between SEMIFIVE and OPENEDGES, initiated in 2019, has culminated in multiple successful SoC platforms employed across diverse applications, including AI inference, Internet of Things (IoT), and HPC. Their collaborative efforts continue to push the boundaries of innovation in SoC design and semiconductor technology, ensuring that SEMIFIVE remains at the vanguard of advanced custom silicon development.

CEO’s Vision for Continued Innovation

SEMIFIVE’s CEO and co-founder, Brandon Cho, underscored the importance of the collaboration with Arm, emphasizing that the platform’s evolution has been propelled by strategic alliances with leading technology firms such as Arm. Cho articulated the substantial value these partnerships deliver to both established industry leaders and innovative startups, enabling them to unlock the potential of custom silicon. The company remains dedicated to driving breakthroughs in AI and HPC technology, leveraging these partnerships to continue pushing the boundaries of what is technologically possible.

“Our platform has evolved to deliver even greater value, from the world’s most advanced tech companies to the most innovative startups, thanks to our ongoing and broader partnership with Arm,” Cho noted.