Lattice Semiconductor Bringing Edge ML into Premium Laptops

At the Consumer Electronics Show (CES) of 2022, Lenovo updated its premium laptop lineup that consists of X1 Carbon, X1 Yoga, and X1 Nano. The most exciting features of these laptops are the inclusion of computer vision technology. Lenovo utilizes computer vision for user detection, privacy control, simple user login, and enhanced energy management.

These features are supported by a Neural Processing Unit (NPU), based on the CrossLink-NX Field Programmable Gated Array (FPGA) from Lattice Semiconductor. Aside from the low-power FPGA, Lenovo uses Lattice Semiconductor’s sensAI (Artificial Intelligence) solution stack. This complementary software solution features ready-to-use Machine Learning (ML) tools, Intellectual Property (IP) cores, hardware platforms, reference designs, and demos, as well as custom design services.

The edge ML market has proliferated in recent years. So much so, that, based on ABI Research’s edge ML market data (MD-AIML-108), edge ML chipset sales are expected to exceed cloud ML chipset sales in 2026. Currently, the growth is mainly led by consumer devices, predominantly in smartphones and wearables. The partnership between Lenovo and Lattice Semiconductor brings edge ML into laptops, a less charted territory.

Since laptops are battery powered and have strict thermal constraint, a low-powered embedded ML-enabled FPGA offers a small energy footprint while delivering always-on computer vision. Lattice Semiconductor claims that its CrossLink-NX FPGAs consume up to 75% less power than similar FPGAs in the market, while featuring small form factor packaging of 4 mm x 4 mm. In addition, an edge ML chipset also helps to process sensitive consumer data directly on-device without sending the data to the cloud, thereby protecting customer privacy and complying with data protection regulations. Finally, the ML inference engine in these laptops is upgradeable via Over-the-Air (OTA) to support future upgrades.

Edge ML chipsets, having previously been found in premium smartphones, true wireless headsets and earbuds, voice control front-end devices, and home security systems, are now increasingly found in mid-range smartphones, other hearables, laptops, webcams, and conferencing systems. Aside from the use cases mentioned earlier, ABI Research foresees on-device natural language processing and contextual ambient sensing as the main drivers for edge ML chipset adoption.

This announcement clearly validates Lattice Semiconductor’s Client Compute AI experience roadmap, which the company announced in November 2021. Since then, the company has introduced several features that allow end users to develop new use cases. The acquisition of face detection and tracking company Mirametrix in November 2021 allows Lattice Semiconductor to upgrade its sensAI solution stack to offer more reference designs for new applications such as user detection, attention tracking, face-framing, and audio detection. The company has also launched an updated neural network compiler and AutoML tools to optimize deep neural networks based on application and dataset targets. Like all popular edge ML software tools and services, sensAI stack is compatible with other widely-used ML platforms, including the latest versions of Caffe, TensorFlow, and TensorFlow Lite.

Offering such a comprehensive set of solutions to the Personal Computer (PC) and laptop Original Equipment Manufacturers (OEMs) is a critical move. There are some vendors in the industry, like Apple, who prefer to introduce their own solutions, but Lattice Semiconductor is targeting other major OEMs. Its solution works well alongside all computing chipsets from major brands, such as Intel, AMD, Qualcomm, and MediaTek. Lattice Semiconductor is able to help customers to reduce time-to-market through its combination of hardware and software, ensuring the deployment is done in the most hardware-optimized and energy-efficient manner. This is also why edge ML software vendors and solution providers are undergoing rapid growth, even though chipset vendors are dominating the market revenue. Software vendors such as Edge Impulse, SensiML, MicroAI, Nota, and Plumerai offer a wide range of software and services targeted at accelerating edge ML deployment, ranging from development platform, model compression techniques, ML acceleration, and ML solutions and applications. For more information on the rise of edge ML software and service vendors, please refer to ABI Research’s report on the Edge AI Ecosystem (AN-5334).