Edge AI Processor

ARA-1 Edge AI Processor

The Kinara Ara-1 AI processor, with its patented Polymorphic Dataflow Architecture, enables applications to run multiple AI models with zero overhead context-switching while minimizing data traffic and significantly reduce power consumption. The processor delivers a significant advantage in performance/$ and performance/Watt over general-purpose GPU’s, making it ideal for applications in like Smart Cities, Smart Retail, Automotive, and Industry 4.0.

Unrivaled Software Tools

The Kinara Software Development Kit (SDK) combined with Ara-1 provides unmatched ease of use, flexibility, and insights to deploy AI at the edge. Deep Vision provides Linux and Windows drivers to support runtime communication between most Linux- or Windows-based host systems and Ara-1.

Increase Performance by Offloading Inferencing

While the host system performs all pre- and post-processing functions, the Ara-1 Edge AI Processor handles the application’s end-to-end inferencing requirements. It is optimized for a batch size of ‘1’ and delivers both real-time responsiveness and low latency. For simultaneous multi-model support, the Ara-1 Edge AI Processor provides the ability to run multiple models simultaneously without sacrificing performance. Ara-1 processor scales from endpoints to edge servers – connect multiple Ara-1 processors to a host for a linear increase the AI performance.

Latency Optimized Design

Optimized to deliver results for a batch size of ‘1’, the Deep Vision ARA-1 Edge AI processor delivers results real time to meet the needs of edge applications

Zero Load on host

The unique hardware design ensures that compiled models place no load on host processors, freeing them from inference application demand

Bigger, Better Deep Learning Applications

Kinara ARA-1 Edge AI Processor can effectively run multiple models without a performance penalty, generating results faster and more accurately.

A New Standard in Flexibility

The processor provides incredible flexibility in implementing highly scalable, low-latency, low power applications.