Expedera Unveils Origin Evolution NPU IP for Edge AI

Expedera has launched its Origin Evolution NPU IP, a new technology aimed at advancing generative AI capabilities on edge devices. Announced in a press release, the Origin Evolution NPU IP is designed to meet the computational demands of running large language models (LLMs) on resource-constrained devices, such as smartphones and automotive systems.

The Origin Evolution NPU IP features a unique packet-based architecture that enhances efficiency by reducing external memory moves by over 75% for models like Llama 3.2 1B and Qwen2 1.5 B. This architecture supports a wide range of applications, offering scalability from 128 TFLOPS in a single core to PetaFLOPS with multiple cores.

Expedera's solution is compatible with popular neural networks, including Llama3, ChatGLM, and MobileNet, and supports both integer and floating-point precisions. The NPU IP is available now and is production-ready, providing a significant reduction in memory and system power needs while increasing processor utilization.