Cadence has unveiled its sixth-generation Tensilica Hi-fi iQ DSP, designed to address the growing computational demands of AI-driven audio and voice processing. This new architecture promises substantial performance gains—twice the raw compute power and eight times the AI efficiency compared to its predecessor—while also delivering over 25% better energy efficiency across most workloads.
The Tensilica Hi-fi iQ DSP is tailored for next-generation applications, including immersive audio codecs like Dolby MS12 and Eclipsa Audio, as well as advanced voice AI models. It supports formats such as FP8 and BF16, enabling seamless execution of small language models (SLMs) and large language models (LLMs) directly on the DSP. This positions it as a versatile solution for edge AI and physical AI applications, where low latency and power efficiency are paramount.
Key improvements over the previous generation include
- Compute Performance: 2x greater raw compute performance.
- AI Efficiency: 8x higher AI performance for complex workloads.
- Energy Savings: More than 25% reduction in power consumption, enhancing battery life in portable devices.
- Enhanced Auto-Vectorization: Simplifies programming and accelerates development timelines.
- Immersive Audio Support: Optimized for codecs like Dolby MS12, Opus HD, and others, delivering up to 40% better performance than previous DSPs.
The DSP also integrates advanced signal processing capabilities for keyword spotting (KWS), active noise cancellation (ANC), beamforming, and automatic speech recognition (ASR). These features are critical for modern voice AI applications, where seamless natural language processing (NLP) is essential. Additionally, it supports multi-stream and multichannel audio playback, enabling realistic 3D spatial sound experiences.
Cadence emphasizes the DSP’s compatibility with its existing ecosystem, including support for TensorFlow Lite for Micro (TFLM), LiteRT, and ExecuTorch environments. This allows developers to leverage established software libraries, compilers, and frameworks, ensuring quick deployment and market readiness. The Tensilica Hi-fi iQ DSP is also targeted for ISO 26262 functional safety certification, making it suitable for safety-critical applications in automotive and other industries.
Availability is expected to begin with lead customers and partners in the first quarter of 2026, with general availability following in the subsequent quarter. Future updates may include support for cache-coherent multicore configurations, further enhancing its flexibility and performance.
The introduction of this DSP underscores Cadence’s commitment to pushing the boundaries of audio and AI processing, catering to the evolving needs of industries where voice and audio interfaces are becoming increasingly prevalent. As AI continues to integrate deeper into consumer electronics, solutions like the Tensilica Hi-fi iQ DSP will play a pivotal role in shaping the next generation of immersive experiences.
