This platform looks to empower developers helping them to create custom silicon solutions more efficiently and has been designed with ML in mind. The SoC platform offers dedicated functionality, which can be accessed using SensiML’s development tools. This integration allows developers to reduce development time and maximise the ML capabilities for edge applications.
According to the company, by leveraging these tools, engineers can efficiently deploy advanced ML models on custom silicon tailored to specific edge use cases, ensuring scalable and effective solutions such as keyword spotting.
The SoC platform delivers a 10x improvement in power efficiency and performance compared to microcontrollers (MCUs) with general-purpose Neural Processing Units (NPUs) and allows developers to optimise performance while reducing power consumption, which is crucial for battery-operated edge devices.
"With our chipIgnite ML custom silicon platform, developers can create solutions that are perfectly tailored to their edge applications, offering significantly improved power efficiency and performance compared to existing solutions," said Mohamed Kassem, CTO at Efabless. "This creates numerous opportunities for specialised ML edge applications that require both optimal performance and reduced power consumption.”
In addition to the launch of the new platform Efabless announced that it has joined forces with SensiML to deliver an open source enabled hardware and software solution for ML edge processing in IoT applications.
SensiML’s AutoML platform enables embedded developers, regardless of their data science experience, to create ultra-efficient sensor inference algorithms that run autonomously on resource-limited edge devices.
Similarly, Efabless equips developers with easy-to-use, open-source tools to design optimised custom SoCs without requiring deep expertise in IC design. According to Efabless and SensiML they are eliminating two of the biggest barriers to IoT innovation by providing a seamless path from development to deployment.
Key benefits include:
- Leverage chipIgnite ML custom silicon platform to achieve dramatically faster performance compared to traditional MCU-based solutions.
- Reduce power consumption by 10x, enabling longer battery life.
- Possible to customise silicon design to meet the specific requirements of a user’s edge ML applications. Profiling and optimisation of ML inference workloads can be accomplished in pre-hardware simulation to assist in sizing inference models appropriately.
- Benefit from a complete development path, from data to silicon, powered by Efabless and SensiML.
- Open-source hardware and software development tools provide transparency, customisation, and a cost-effective path to ML at the edge products.
"We’re excited to collaborate with Efabless to offer a comprehensive development pathway for intelligent edge devices," said Chris Rogers, CEO of SensiML. "By combining our strengths, this joint platform tackles the complex hardware and software challenges developers face, enabling the creation of truly differentiated IoT edge sensing products and applications."
Efabless has already taped out the chipIgnite ML, marking a major milestone in the platform's development. A design kit will be available for early evaluation starting in November 2024.
The first shuttle for prototyping is scheduled for April 2025, with full-scale production expected to follow.
