Origin delivers full accuracy and predictable performance without needing hardware-specific optimizations or changes to your trained model. Its patented packet-based execution architecture provides single-core performance up to 128 TOPS and sustained utilization rates of 70-90%, measured in silicon running common AI workloads such as ResNet. This best-in-class performance and utilization enables users to run AI models with less power consumption than alternative solutions.
Expedera’s Origin IP platform includes a full TVM-based software stack that enables developers to work efficiently as they deploy their networks on the target hardware. The software stack provides a choice of two flows – an offline compilation flow based on the TVM compiler and an interpreted flow for on-device or online use. TVM enables a strategy of writing once and deploying everywhere. It supports all popular frameworks, including TensorFlow, TensorFlow Lite, PyTorch, and ONNX, allowing developers to continue working with their favorite frameworks.
The software stack automatically handles tasks such as packetization and debugging while providing user-friendly options such as mixed precision quantization—using your tools or ours—custom layer and custom network support, and multi-job APIs.
Origin E1
The Origin E1 processing cores are individually optimized for a subset of neural networks commonly used in home appliances, edge nodes, and other small consumer devices. The E1 LittleNPU supports always-sensing cameras found in smartphones, smart doorbells, and security cameras.
