SiFive becomes first RISC-V designer to integrate Nvidia NVLink for AI data centers

SiFive Brings Nvidia NVLink to RISC-V CPU Cores

SiFive has joined the growing list of chip companies embracing Nvidia NVLink Fusion interconnect technology, becoming the first RISC-V chip designer to integrate the high-speed chip-to-chip connectivity solution 1

1

2

2

. The company, which designs RISC-V CPU cores and processors for applications ranging from edge devices to data centers, licenses its blueprints to customers who can incorporate them into their own silicon—a business model similar to Arm's approach 1

1

.

High-Speed Connections Between Chips Enable AI Infrastructure

Nvidia NVLink enables customers to abstract an entire rack of CPUs and GPUs as a single unified accelerator, delivering up to 3.6 TB/s of chip-to-chip bandwidth 1

1

. These high-speed connections between chips are essential in AI data centers where thousands of Nvidia's AI chips must work together seamlessly to process massive amounts of data 2

2

. Patrick Little, SiFive CEO, emphasized that customers will be able to connect a RISC-V CPU to Nvidia's AI chips with the same speeds as CPU chiplets using Intel or Arm technology 2

2

.

NVLink Fusion Interconnect Technology Gains Industry Momentum

Until recently, Nvidia kept NVLink proprietary. However, with the introduction of NVLink Fusion last year, Nvidia extended support for the interconnect fabric to the broader ecosystem 1

1

. Major players including Intel, Arm, Fujitsu, and Qualcomm have already thrown their weight behind the technology. Intel even plans to release client systems that use NVLink Fusion to connect its CPU chiplets to Nvidia's GPU dies 1

1

. Nvidia officially supports two NVLink Fusion configurations: combining partner CPUs with Nvidia's GPUs, and pairing Nvidia's Grace or Vera CPUs with custom accelerators 1

1

.

Custom CPUs and Heterogeneous Co-Designed Systems

SiFive is focusing on enabling customers to build custom CPUs using its cores and System on a Chip (SoC) reference designs for integration with Nvidia's AI infrastructure 1

1

. "The adoption of NVLink Fusion reflects a broader industry shift toward heterogeneous, co-designed systems where open CPU architectures and advanced interconnects work together to define the future of AI data center computing," Patrick Little stated 1

1

. He added that integrating NVLink Fusion with SiFive's high-performance compute subsystems enables an open and customizable CPU platform that pairs seamlessly with Nvidia's infrastructure to deliver exceptional efficiency at data center scale 1

1

.

RISC-V Gains Traction as Open Alternative to Arm

RISC-V, an open-standard instruction set architecture, has drawn fresh interest from companies like Google and Meta as more details have emerged about Arm's ambitions to become a chip designer in its own right 2

2

. SiFive chip designs featuring the technology won't likely hit the market until 2027 or later, but the partnership represents a multi-generational commitment from both companies 2

2

. SiFive reports having a number of datacenter design licenses with several already taped out, though the company hasn't announced new ones yet 1

1

.

Implications for UALink and Competing Interconnect Standards

The growing momentum behind NVLink Fusion comes as AMD prepares to bring UALink-based AI rack systems to market with the MI455X-based Helios 1

1

. UALink was envisioned as an open alternative to NVLink with backing from Intel, AMD, Amazon, Meta, Microsoft, and Arm. However, a lack of UALink switches forced AMD to tunnel the protocol over standard Ethernet. Further complicating matters, Broadcom, once a strong backer of the standard, is now pushing its own interconnect fabric called Scale Up Ethernet 1

1

. The widespread industry adoption of NVLink Fusion by major chip designers raises questions about the viability of competing interconnect standards in AI data center computing.