Actions Technology is pleased to announce the launch of a groundbreaking new generation of edge AI audio chips based on Mixed-mode SRAM-based Computation-in-Memory (MMSCIM) technology. We are proud to introduce the ATS323X series chips, targeting the proprietary low-latency wireless audio area. The chip solution is currently being co-developed with brand customers and is set to be implemented in terminal brand products in the near future, promising users a distinctive AI product experience.

The ATS323X series is our first tri-core heterogeneous SoC equipped with an AI-NPU (Actions Intelligence NPU) and our third-generation high-quality, low-latency wireless audio transceiver chip. This series can be widely applied to wireless microphones, wireless gaming headsets, wireless communication headsets, multi-link high-definition conference systems, and Bluetooth transceiver devices, empowering these products with AI capabilities and helping our customers step into the new AI era.

• Breakthrough AI Architecture, Leading Audio Technology Innovation

As our first wireless audio chip utilizing the AI-NPU architecture, the ATS323X features an advanced design integrating MMSCIM and HiFi5 DSP. The MMSCIM core computing capability reaches 100 GOPS@500MHz, with an impressive energy efficiency ratio of 6.8 TOPS/W. Compared to DSP HiFi5 under equivalent conditions, the actual application computing power and energy efficiency ratio can be improved by approximately 16 times and 60 times, respectively, with power consumption reduced by over 90%.

By leveraging the collaborative design of MMSCIM and HiFi5 DSP, our NPU supports fundamental universal AI operators, while the DSP provides specialized operators, creating a highly flexible and energy-efficient AI NPU integrated architecture. The chip supports mainstream deep learning frameworks including TensorFlow, Caffe, TFLite, PyTorch, and ONNX, and is compatible with all current mainstream AI models, providing robust computing power to support personalized AI algorithm development and application integration for our terminal products.

• Outstanding Audio Performance, Innovative Experience Upgrade

The ATS323X delivers exceptional audio performance. The chip supports a full-link 48KHz@32bit high-definition audio path, achieving a DAC SNR of 120dB (noise less than 2uVrms) and an ADC SNR of 111dB (noise less than 3.6uVrms). Our newly upgraded 48K dual-mic AI ENC noise reduction algorithm provides superior noise cancellation compared to previous single-mic solutions, maintaining high fidelity and low latency. In terms of latency control, the ATS323X further reduces end-to-end latency to 9ms under our 2.4G NGPP-Gen3 private protocol mode, meeting the demanding requirements of ultra-low latency applications such as karaoke and gaming.

• Powerful Wireless Performance, Expanding Application Boundaries

The ATS323X supports a transmission power of up to 16 dBm and features a new generation of wireless frequency-hopping technology. The wireless transmission bandwidth has increased by 100% compared to the second-generation chip, reaching 4 Mbps, with a maximum transmission distance of 450 meters. This demonstrates the chip's robust wireless connectivity and interference resistance, supporting flexible networking modes including one-to-multiple, four-to-four, and multiple-to-one configurations.

Amy Liu, Vice President of Actions Technology Group and General Manager of the Low-Latency Wireless Audio BU, stated: "The launch of the ATS323X marks a significant milestone for our company in the edge-side AI audio field. By integrating AI technology with advanced audio processing and wireless transmission technologies, we are committed to providing customers with more intelligent, higher-quality, lower-latency, and more personalized wireless audio solutions, driving rapid industry upgrades."