“The frequent updates with new APIs and math operators available to OEMs through the Google Play store will improve the AI performance.”
In 2019, Qualcomm announced the first GPU updateable drivers in the mobile industry. Qualcomm Snapdragon Elite Gaming features enabled smartphone users to update their GPU drivers more regularly, just like desktops and reap the possible benefits, including improved performance and new features. Now Qualcomm technologies have taken another step towards expanding its mobile platform. Qualcomm has announced an exciting collaboration of developing Snapdragon mobile platforms’ updatable drivers and making it more powerful in conjunction with Google.
This fall, Google will launch regular Neural Networks API (NNAPI) updates through Google Play Services to Qualcomm Snapdragon mobile platforms. The NNAPI drivers will be available to many devices, including the latest and best Snapdragon processors, which will automatically update in the background. Often it happens that driver updates are only available for the latest models, but this time it’s not the case; the NNAPI drivers will be compatible with previous Snapdragon processors, as well. This will allow a broad spectrum of mobile users to benefit from the driver updates. Regular, routine updates to Google Assistant and Google Maps are made possible because of this change.
“The NNAPI gives up to 100% more compute with 2-3 times better latency at lower power.”
In an I/O presentation about advancements in machine learning, Google developers stated that the NNAPI could boost performance equivalent to two extra cores on the phone. The NNAPI gives up to 100% more compute with 2-3 times better latency at lower power and thermal output. For many years Qualcomm Technologies has been supporting and advocating for community alignment in the Android ecosystem. Qualcomm has been making updates on the Qualcomm Neural Processing SDK regularly. The Qualcomm Neural Processing SDK for AI was created to help developers save time and effort in optimising the effectiveness of trained neural systems on Snapdragon devices. It does this by providing model conversion and execution tools and APIs for the core to match the desired user experience with the performance. The Qualcomm SDK can operate one or more Caffe/Caffe2, ONNX or TensorFlow-trained neural network models. The frequent updates with new APIs and math operators available to OEMs through the Google Play store will improve the AI performance. Still, various mobile applications will be enhanced because of these frequent updates, making them faster and more responsive.
The brand-new 6th Gen Qualcomm® Artificial Intelligence (AI) Engine includes features such as a fused AI accelerator, boosts AI Engine performance to an astonishing 26 TOPS, is a step towards Qualcomm’s dreams of making AI ubiquitous. The re-engineered Qualcomm HexagonTM 780 Processor combines artificial intelligence to give enhanced experience in photography and connectivity. Qualcomm has been rolling out new technologies for IoT devices to further its vision of Global AI acceleration. It’s AI Model Efficiency Toolkit (AIMET) on GitHub to collaborate with other prominent AI researchers and provide a simple library plugin for AI developers to use for cutting-edge model efficiency performance. This open-source project aims to assist in the migration of the ecosystem toward integer inference to boost performance per watt.
It will be exciting to see Snapdragon-powered devices take flight with their updatable AI drivers to give users a smooth and enhanced performance. Its capability to run applications faster will be put to the test in the coming months.