MemryX Inc. is a fabless semiconductor company that specializes in designing edge AI processing solutions. Founded in 2019, the company focuses on accelerating artificial intelligence (AI) processing for edge devices through its proprietary compute-at-memory technology and dataflow architecture. MemryX's flagship product, the MX3 AI Accelerator, is designed to be an efficient and easy-to-use solution for edge AI applications. The MX3 can be integrated into various edge computing systems, offering scalable AI capabilities for industries such as automotive, robotics, IoT, smart vision systems, and edge computing.
The MX3 AI Accelerator features MemryX Compute Engines (MCE) that are tightly coupled with at-memory computing, creating a native dataflow architecture that utilizes up to 70% of the chip with just one click. This design allows developers to compile any trained AI model with a single click, achieving optimized model performance with high utilization (>50%) and full accuracy within minutes. The MX3 supports multiple AI frameworks, including TensorFlow, TensorFlow Lite, ONNX, PyTorch, and Keras, and can run hundreds of unaltered AI models efficiently.
In terms of performance, the MX3 offers more than five TFLOPs of computing power while consuming approximately one watt of power. It supports BFloat16 activations and can handle weights of 4, 8, and 16 bits. The chip can store about 10 million parameters on-die and interfaces with host systems through PCIe Gen 3 I/O and/or USB 2.0/3.x. MemryX offers the MX3 in various form factors, including bare die, single-die package, and modules (mini PCIe or M.2) with one or more MX3 chips.
As of December 2023, MemryX announced the availability of production-level silicon for its MX3 AI Accelerator, with volume production expected to start in 2023. The company has showcased various demonstrations of the MX3's capabilities, including multi-stream AI-enabled video management systems, AI-powered PCs, threat and weapons detection for public spaces, and smart city traffic management solutions.
Key customers and partnerships
MemryX has formed partnerships with several companies to showcase and integrate its AI acceleration technology. In February 2023, MemryX announced a partnership with Edge Impulse, an AI-algorithm software provider, to streamline the development, testing, and deployment of AI models on MemryX's AI accelerator chips. This collaboration aims to simplify the process of creating and implementing AI solutions for edge devices.
The company has also worked with Cachengo to develop an intelligent edge platform for real-time threat detection in public spaces. Additionally, MemryX has partnered with Fracctal to demonstrate smart city traffic management capabilities powered by the MX3 AI Accelerator.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.