Speedata deploys Arteris FlexNoC in Callisto & Andromeda

Speedata has deployed Arteris FlexNoC technology in its Callisto processor and is also using it in its next-generation Andromeda APU.

The agreement ties Arteris' network-on-chip interconnect technology to Speedata's analytics processing unit design for data-centre analytics workloads. Speedata says its processors are built to run Apache Spark SQL, AI data preparation, agentic analytics and batch ETL tasks directly in silicon.

Demand for data analytics infrastructure has grown as companies process larger volumes of structured and unstructured information. That has increased focus on how chips move data between compute, memory and input/output blocks, not just on raw processing speed.

FlexNoC provides the on-chip interconnect that links those parts of the design. That puts Arteris in a semiconductor layer becoming more important as specialised processors proliferate in AI data centres and other systems.

Speedata is developing purpose-built processors for analytics workloads rather than relying on general-purpose compute. Its APU architecture is intended for Spark-based data processing and related tasks that underpin many enterprise analytics pipelines.

Callisto is the current processor named in the deployment. FlexNoC has also been used across Speedata's forthcoming Andromeda APU, suggesting the relationship extends beyond a single product generation.

Large-scale analytics systems often hit bottlenecks when data movement cannot keep pace with processing elements. The problem has become more pronounced as AI-related workloads, data preparation and database processing place heavier demands on memory bandwidth and interconnect design.

Rafi Shalom, Chief Technology Officer, Chief Architect & Co-founder of Speedata, outlined the role of the technology in the company's chip programmes.

"AI inference is transforming every technology stack and industry at speed and scale, with real-time data analytics, accelerated by Speedata APU, as the catalyst. Arteris FlexNoC has been an important enabling technology across both Callisto and our next-generation Andromeda APU, giving our engineering teams the flexibility to focus on innovation in accelerated data analytics while relying on a proven interconnect foundation," said Rafi Shalom, Chief Technology Officer, Chief Architect & Co-founder, Speedata.

Arteris supplies semiconductor design technology used to manage communication across complex chips. Its products include network-on-chip intellectual property and software designed to support system-on-chip integration and hardware security assurance.

The company positions those tools for chip designers working on AI data-centre processors, edge AI devices and physical AI systems. As more developers build domain-specific silicon, interconnect technology suppliers are seeking a larger role in the design stack.

Speedata has built its business around analytics acceleration, particularly for Apache Spark SQL and related workloads. It says its architecture can be introduced into existing data pipelines without code changes, a pitch aimed at organisations seeking performance gains without rewriting established analytics software.

Speedata also says one enterprise AI data preprocessing deployment reduced the number of servers required from 37 to three. It separately claims up to 100 times faster performance and 90% lower total cost of ownership than general-purpose compute for the targeted workloads.

Those figures were not independently verified, but they point to the commercial case for specialised analytics silicon. If enterprises can reduce server counts for data preparation and ETL jobs, they could cut infrastructure footprints as well as operating costs.

K. Charles Janac, President and Chief Executive Officer of Arteris, said the market is placing more value on communication within increasingly complex chip designs.

"Across all types of AI computing and data analytics infrastructure, moving high-performance data efficiently, safely, and securely has become just as important as processing it. As organizations work to extract value from ever-larger datasets, semiconductor innovation increasingly depends on system architectures capable of delivering efficient communication between compute resources. Speedata has developed an innovative approach to data analytics acceleration, and we're pleased our technology is helping enable that vision," said Janac.