Faster, Smarter AI: Alphawave Semi Shakes Up AI with Modular Chiplet Design

Introduction

Alphawave Semi is making significant strides in the semiconductor industry by developing a compute chiplet based on the ARM Neoverse N3 Compute SubSystem (CSS).

This strategic move marks Alphawave Semi’s shift from primarily an IP business to producing high-performance custom silicon solutions.

The development comes in the wake of their acquisition of RISC-V developer SiFive, positioning Alphawave Semi at the forefront of advanced AI computing technology.

Why is this a big deal?

There are a couple of key things that make Alphawave Semi’s high-performance AI compute chiplet a big deal:

Shift towards chiplet-based design: This chiplet represents a move by Alphawave Semi from just licensing intellectual property (IP) to designing and manufacturing physical chiplets. Chiplet technology allows for more modular and efficient chip design, potentially leading to faster innovation and better performing chips.
Focus on high-performance AI: The chiplet is specifically designed for the demanding world of AI computing, which is a rapidly growing field. This focus positions Alphawave Semi to cater to a specific need in the market.
Faster development and lower costs: By using pre-designed chiplets with features like ARM Neoverse N3 cores and high-speed connectivity, companies can develop new custom AI chips faster and potentially at a lower cost compared to traditional chip design methods.

Here’s how this can benefit different players in the tech industry:

Chip designers: Alphawave Semi’s chiplet provides a building block that can be easily integrated into custom AI chips, saving them development time and potentially reducing costs.
AI hardware manufacturers: These companies can leverage the chiplet to create high-performance AI hardware for various applications like data centers, autonomous vehicles, and research.
End users (businesses and researchers): Faster development of powerful AI hardware can lead to advancements in various AI applications, impacting fields like healthcare, finance, and self-driving cars.

Background

Alphawave Semi has been a notable player in the semiconductor IP market, known for its high-speed connectivity solutions. Their recent acquisition of SiFive, a prominent RISC-V developer, highlighted Alphawave’s commitment to expanding their technological capabilities and product offerings. This acquisition is part of a broader strategy to leverage emerging technologies and integrate them into cohesive, high-performance computing solutions.

The ARM Neoverse N3 Compute SubSystem (CSS) provides an advanced platform that meets the demands of modern high-performance computing (HPC) and artificial intelligence (AI) applications. Renowned for its efficiency, scalability, and performance, ARM’s Neoverse architecture makes an ideal choice for developing next-generation compute solutions.

Alphawave Semi is making a shift in the chip industry by developing a new chiplet specifically designed for high-performance artificial intelligence computing. This chiplet utilizes the ARM Neoverse N3 cores, which are considered mid-range for performance but offer scalability.

Here are some key points about the chiplet:

Designed for high-performance AI computing
Uses ARM Neoverse N3 CPU cores
Includes ARM Coherent Mesh Network (CMN) for efficient performance
Manufactured on leading-edge process nodes for optimal efficiency
Designed to be adaptable for various applications including HPC, data centers, and AI/ML

This move is significant for Alphawave Semi. It marks a shift from their traditional business model of intellectual property licensing. Now, they are venturing into custom silicon design. This chiplet is a key component of their new platform.

A New Era in Chiplet Technology

Alphawave Semi’s foray into chiplet technology represents a crucial step towards enhancing performance and efficiency in AI computing.

The chiplet-based custom silicon design platform integrates various IPs, including I/O extension chiplets, memory chiplets, and compute chiplets.

Leveraging Alphawave Semi’s expertise in PCI Express 6.0 and 7.0 connectivity, alongside advanced packaging capabilities, the new compute chiplet promises robust performance and flexibility.

Levering their expertise in packaging and high-speed connectivity solutions, Alphawave Semi’s chiplet offers several advantages for AI computing:

Modular Design: The chiplet integrates seamlessly with other chiplets such as IO extension and memory chiplets, enabling a flexible custom silicon design platform.
Advanced Packaging: Alphawave Semi utilizes advanced packaging techniques to combine the compute chiplet with high-speed connectivity solutions like PCIe Gen 6.0 and 7.0, offering superior data transfer rates between the chiplet and other components.
Scalability: The chiplet’s design with the ARM Neoverse N3 cores and the ARM Coherent Mesh Network (CMN) allows for efficient scaling of performance based on specific needs.
Faster Time-to-Market: By using pre-designed chiplets, companies can develop new custom AI chips quicker compared to traditional methods.

Overall, Alphawave Semi’s chiplet is designed to be a versatile and high-performance solution for demanding workloads in AI, data centers, and HPC environments. This strategic partnership with ARM positions Alphawave Semi as a strong contender in the growing chiplet-based chip design space.

Key Features of the Neoverse N3 Compute Chiplet

High-Performance Core

The compute chiplet revolves around the mid-range Neoverse N3 CPU core cluster and the ARM Coherent Mesh Network (CMN).

It is designed to be available on industry-leading process nodes, ensuring optimal performance and energy efficiency.

This setup aims to cater to the demands of high-performance computing (HPC), data centers, AI/ML applications, and 5G/6G infrastructure.

Advanced Packaging and Connectivity

Utilizing its advanced packaging techniques, Alphawave Semi incorporates PCIe Gen 6.0 and 7.0, Universal Chiplet Express (UCIe), and 112/224G Ethernet subsystems.

This comprehensive integration ensures that the ARM-based compute chiplet delivers high-speed I/O capabilities necessary for modern digital infrastructure.

Strategic Partnerships and Ecosystem Development

Alphawave Semi’s new chiplet development benefits significantly from its strategic partnership with ARM. Mohit Gupta, Senior VP & GM of Custom Silicon & IP at Alphawave Semi, emphasized the importance of this collaboration.

“Our ARM-based compute chiplet is a critical component in Alphawave Semi’s custom silicon platform. It showcases our IP, SoC, and packaging capabilities and our successful strategic partnership with Arm,” Gupta said.

Alphawave Semi joined the ARM Total Design program last year, which fosters a rapidly growing ecosystem around custom silicon. Eddie Ramirez, Vice President of Go-to-Market, Infrastructure Line of Business at ARM, highlighted the collaboration’s significance.

“Alphawave Semi’s new advanced compute chiplet exemplifies how leading companies leverage Neoverse CSS to optimize performance and power efficiency, accelerating market entry for next-generation AI and HPC workloads,” Ramirez stated.

Implications for the Semiconductor Industry

Enhanced Performance and Scalability

The integration of ARM’s Neoverse CSS platform with Alphawave Semi’s proven IP is set to drive innovation across the chiplet ecosystem.

This development simplifies creating critical custom silicon for AI, HPC, and other demanding digital infrastructure applications.

The chiplet design approach allows for enhanced performance, scalability, and energy efficiency, meeting the growing computational demands of various industries.

Market Disruption and Competitive Edge

The Aum processor combines high performance, efficiency, and scalability. It aims to disrupt the market and challenge established players.

This move highlights Alphawave Semi’s commitment to advancing semiconductor technology. It also shows their dedication to staying competitive in the rapidly evolving AI and HPC sectors.