Edge AI accelerator chip offers ‘one-click’ compiler

New Products |
By Rich Pell

Edge AI chip startup MemryX has announced that it has begun customer sampling of its MX3 AI accelerator. The MX3, says the company, was created to become the easiest-to-use and most efficient edge AI accelerator.

Users can compile any trained AI model with just a single click. Within minutes, the developer will achieve optimized model performance with high utilization (>50%) and full accuracy using the MX3. This is offered as saving months of software development time and speeds time to market.

The MX3 can be used to accelerate AI processing in nearly any edge computing system, whether the legacy or a new design. The MX3 directly connects to any applications processor, providing scalable AI for a wide variety of industries including automotive, Industry 4.0, robotics, IOT, smart vision systems, edge computing, and the metaverse.

“Reaching the milestone of customer sampling is gratifying, but what’s even better is the feedback from initial customer experience with the MX3,” says MemryX CEO Keith Kressin. “Customers are seeing first-hand how our solution is different from other AI accelerators.”

The company’s at-memory compute and dataflow architecture offers high performance with low latency and low power consumption. Even more importantly, says the company, the architecture was designed to give customers the ability to seamlessly scale performance, power, and latency according to their application requirements.

Hardware kits and SDKs are available to customers upon request (contact MemryX Vice President of Product and Business Development Roger Peene at



Linked Articles