Cadence has introduced its sixth-generation digital signal processor IP, the Tensilica HiFi iQ DSP, designed specifically to meet the growing demands of voice artificial intelligence and advanced audio processing. The new DSP architecture delivers significantly higher performance while improving power efficiency for edge and embedded systems.
Higher Performance for AI and Audio Workloads
Compared to previous HiFi DSP generations, HiFi iQ offers:
- Up to 2× higher overall compute performance
- Up to 8× improvement in AI and neural network processing
- More than 25% better energy efficiency
These gains are achieved through a redesigned architecture optimized for voice AI pipelines, neural inference, and complex audio algorithms running directly on-device.
Architecture Built for Edge AI
The HiFi iQ DSP integrates advanced vector and SIMD processing, expanded MAC capabilities, and dedicated AI instructions. It supports modern data types and is optimized for running both large and small language models, enabling real-time inference without reliance on cloud resources.
The DSP also integrates seamlessly with Cadence’s AI software ecosystem, including NeuroWeave SDK and popular machine learning frameworks, simplifying deployment of AI models on embedded platforms.
Target Applications
Cadence positions the HiFi iQ DSP for a wide range of applications, including:
- Voice assistants and speech recognition systems
- Automotive infotainment and in-cabin audio
- Consumer audio and immersive sound solutions
- Edge AI and system-on-chip designs
The IP also supports functional safety requirements, making it suitable for automotive and safety-critical designs.
Availability
The Tensilica HiFi iQ DSP IP is expected to be available to customers in 2026, with early access programs beginning ahead of general release.
Source: Cadence / All About Circuits