Run larger neural networks in edge devices with powerful, power-efficient intelligence.
Fast, flexible, made for more complex (and more compelling) tasks
Ergo 2 provides the performance to handle more complex use cases, including those requiring transformer models, larger neural networks, multiple networks operating simultaneously and multimodal inputs.
|Performance – Max inferences per second (IPS)|
In addition to standard CNNs, RNNs, and LSTMs, Ergo 2 enables developers to leverage more powerful networks such as RoBERTa, GANs, and U-Nets, delivering responsive, accurate results, processed on-device for better power-efficiency, privacy, and security.
The Ergo 2 chip has been designed with a pipelined architecture and unified memory, which improve its flexibility and overall operating efficiency. As a result, it supports higher-resolution sensors and a wider range of applications, including:
- Language processing applications like speech-to-text and sentence completion
- Audio applications like acoustic echo cancellation and richer audio event detection
- Demanding video processing tasks like video super resolution and pose detection
Ergo 2 works with large networks, high frame rates, and multiple networks, enabling intelligent video and audio features for devices such as:
- Enterprise-grade cameras for security, access control, thermal imaging, or retail video analytics applications
- Industrial use cases including visual inspection
- Consumer products including laptops, tablets, and advanced wearables
Like its predecessor, Ergo 2 has a 7 mm by 7mm footprint and requires no external DRAM, making it well suited for use in compact devices.
Powerful, yet power-efficient
Ergo 2 offers significantly better power-efficiency than alternatives, running inferences on 30 fps video feeds in as little as 17 mW of compute power. This means it can operate within a variety of power-constrained devices and situations without external cooling, making it ideal for small devices and more versatile product packaging.
ResNet‑50 Power Efficiency (IPS/W)
Processing power to handle serious on-device ML features
Reducing or eliminating the need to send data to the cloud for inference reduces latency, improves power-efficiency, saves cloud costs and minimizes the opportunity for sensitive data to be intercepted.
Ergo 2 provides product developers with locally processed machine learning without trading away features or performance due to a lack of processing power or the power constraints of the device.
Development tools and model zoo to get you started quickly
Ergo 2 is supported by a full suite of development tools to ensure successful integration into your device. Perceive provides an ML toolchain that compresses and optimizes neural networks to run on Ergo 2, a software development kit for building embedded applications using Ergo 2, and a model zoo with example networks and applications to accelerate the development process.Download the Product Overview
NVIDIA Jetson Xavier NX: https://developer.nvidia.com/embedded/jetson-benchmarks (Feb 27, 2022)
Qualcomm Snapdragon 888: https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets/documents/tech_summit_2020-ai_-_jeff.pdf
Hailo Hailo-8: https://hailo.ai/developer-zone/benchmarks/