How Cadence DSPs Put In-Cabin AI Audio On-Chip in SemiDrive's X10

SemiDrive, an automotive semiconductor company building SoCs and MCUs for the software-defined vehicle, has licensed and integrated Cadence Tensilica HiFi 5s DSPs into its X10 next-generation AI cockpit SoC, where they form the audio and voice subsystem. X10 has secured production collaborations with several leading automakers.

The two companies worked together to bring premium, multi-zone audio and always-on voice into a single-chip cockpit design, without the external audio silicon that architectures of this class have traditionally required.

The Challenge: AI Moves into the Cockpit

The vehicle cockpit has become one of the most demanding compute environments in the automotive industry. Multi-microphone voice capture, far-field recognition, echo cancellation, active noise reduction, immersive multi-zone audio, AI voice assistants, and always-on sensing must all run simultaneously and in real time, alongside high-resolution displays and multi-camera vision.

In traditional designs, graphics and display workloads consume most of the available system bandwidth, leaving little headroom for on-device AI. To deliver a premium AI experience, many next-generation cockpit platforms pair a cockpit SoC with an external AI accelerator, thereby increasing cost, complexity, and thermal load.

Collapsing that architecture into a single chip meant the audio and voice subsystem had to become dramatically more efficient. It needed to deliver production-grade quality across a large microphone and speaker array, run continuously within a tight power and thermal envelope, and coexist with AI, graphics, and safety workloads on the same die.

The Collaboration: Designing the Audio Subsystem Together

Cadence engaged directly with the SemiDrive design team to architect the audio and voice path for X10, aligning the hardware and software needed to develop audio, voice, and AI features in parallel with the rest of the platform.

The result of that joint effort is a design in which Cadence Tensilica HiFi 5s DSPs handle the real-time audio and voice workloads on-chip, eliminating the need for a separate external audio device while supporting a large multi-microphone and multi-speaker array for multi-zone, immersive in-cabin sound.

X10 incorporates four Cadence Tensilica HiFi 5s DSPs, which handle audio and voice processing while CPU, GPU, and NPU resources stay dedicated to application, graphics, and large-model AI workloads. The HiFi 5s DSP also enables AI-based speech and audio AI, allowing the chip to deliver the real time execution of those functions.

For more than two decades, Cadence Tensilica DSPs have powered billions of audio devices worldwide, earning a reputation as one of the industry’s most widely adopted programmable audio DSP architectures. The same strengths translate directly into automotive:

• Deterministic, power-efficient real-time processing for audio, voice, and communications workloads

• A mature, production-proven SDK and software development environment that reduces integration risk

• A broad portfolio of audio, voice, speech, communications, and AI software packages

• Deep ecosystem partnerships spanning OEMs, Tier 1 suppliers, software vendors, and codec licensors

• Developer familiarity and code portability built on a stable, scalable programming model

Customers do not simply license a processor. They gain access to an ecosystem and a Cadence engineering team that works alongside them, which accelerates deployment and reduces risk as vehicle programs grow larger and software complexity rises.

The Result

X10 integrates AI, graphics, connectivity, safety, and real-time signal processing into a single automotive-grade chip, replacing the traditional cockpit-SoC-plus-external-AI-accelerator approach. The platform is designed to meet automotive functional safety requirements, including ISO 26262 ASIL B, as well as cybersecurity requirements, while supporting both a safety-oriented and an open application software baseline.

By consolidating the audio and voice subsystem onto Cadence Tensilica DSPs, the design contributes to meaningful reductions in system cost, power, and thermal complexity, helping bring high-end AI cockpit capability to a broader range of vehicle models.

“The Cadence Tensilica HiFi 5s DSP was purpose-built for exactly this class of workload, combining high-performance audio and voice processing with AI acceleration in a single programmable core. That dual capability lets front-end audio processing and AI-based voice enhancement run together on the same engine, rather than requiring separate hardware blocks. Working alongside the SemiDrive design team, we architected the X10 audio subsystem to take full advantage of that, so premium multi-zone sound and always-on voice fit within the power and area budget of a production cockpit design,” said Casey Ng, Product Marketing Director, Cadence.

"Our goal was to give automakers a richer in-cabin experience without adding cost or complexity to the vehicle. Cadence was the natural partner: the Tensilica audio ecosystem is mature, widely supported by our software partners, and familiar to our engineering teams, which let us develop and tune the audio and voice experience in parallel with the rest of the platform and reach our program milestones on schedule,"  said Mingle Sun, Chief Technology Officer of SemiDrive

AI and DSP, Working Together

As AI capabilities expand in cockpit platforms, the role of DSPs becomes even more critical, not less. AI inference is only one part of the cockpit compute stack. Audio front-end processing, voice enhancement, wake-word detection, acoustic processing, sensor fusion, and other latency-sensitive workloads all depend on dedicated signal-processing engines built for deterministic performance and power efficiency.

The most effective in-cabin AI architectures bring specialized engines together—NPUs for AI inference, CPUs for application control, GPUs for visual workloads, and DSPs for always-on audio, voice, and AI-enhanced signal processing—enabling more natural in-cabin experiences.

Looking Ahead

The automotive industry is entering an era where AI, software, and user experience increasingly define product differentiation. Companies that combine innovative silicon with mature software ecosystems will be best positioned to succeed.

The collaboration between Cadence and SemiDrive on X10 reinforces a belief Cadence has held for many years: Technology leadership is not created by hardware alone. It is created by ecosystems, developer productivity, software availability, and close engineering partnership. Those principles helped make Cadence one of the world’s most widely adopted audio DSP platforms, and they are now helping drive the next generation of automotive cockpit innovation.

Learn more about Cadence Tensilica HiFi DSPs for automotive audio and AI applications.