Cadence Expands Tensilica IP Portfolio with Enhanced Intelligence and Edge AI
November 15, 2023
The holy grail of Edge AI is in the meeting of computer vision, audio/visual processing, low-latency algorithmic processing and energy efficiency. Now, in a recent announcement Cadence Design Systems, has made public an expansion of its Tensilica HiFi and Vision DSP families that accomplish just that goal.
The release outlines the introduction of four new DSPs based on the Tensilica Xtensa LX8 processor platform. According to the company, the new HiFi and Vision DSPs manage ever-growing requirements for system-level performance and AI across a wide range of applications that improve performance through multiple algorithms and even improve energy efficiency.
The expanded portfolio reportedly includes:
- Tensilica HiFi 1s DSP
- Tensilica HiFi 5s DSP
- Tensilica Vision 110 DSP
- Tensilica Vision 130 DSP
“With lengthy SoC design cycles and rapidly changing AI algorithms, our SoC customers require the utmost flexibility. On-device and edge AI systems feature numerous diverse sensors, so any AI inference solution must be adaptable to the end application requirements,” said David Glasco, VP, research and development, Tensilica IP at Cadence. “Designing with a system-level perspective is crucial, and these SoCs must be energy-efficient, flexible and future proof to accommodate new neural networks. Cadence continues to make significant investments in our Tensilica product line, and our latest HiFi and Vision DSPs reflect that commitment.”
Xtensa LX8 Platform
The new HiFi and Vision DSPs leverage the Xtensa LX8 platform, according to the announcement, which includes native AMBA AXI support for lower latency and L2 cache for increased system performance.
Cadence says that branch prediction enables a 5 to 20 percent reduction in cycles for both the HiFi and Vision DSPs. Supported by Cadence’s new NeuroWeave Software Development Kit (SDK), the DSPs can be paired with the Cadence Neo Neural Processing Units (NPUs) to offload AI inference workloads.
HiFi 1s and HiFi 5s DSPs
The new Tensilica HiFi DSPs add more capabilities for lightweight imaging and AI so they can serve as additional compute resources beyond just audio, stated the release. Individually and collectively, the following features reportedly improve performance:
- Double-precision floating-point acceleration provides up to 30X better out-of-box performance of popular codecs, enabling easier handling for computation requiring expanded dynamic range and precision
- Hardware/software co-design enhances auto-vectorization, greatly reducing the need for hand-optimized code and providing source code compatibility across HiFi DSPs
- New 8-bit operations in the HiFi 1s DSP boost imaging and AI performance by up to 2X at the kernel level, compared to the HiFi 1 DSP
- L2 cache increases system performance by up to 50 percent
“Hailo is developing powerful AI-centric vision processors serving the growing edge AI market. ln our recently announced Hailo-15 device targeting the IP camera market, we've chosen to integrate Cadence’s Tensilica DSP among the computational subsystems complementary to our proprietary neural engine,” said Guy Kaminitz, VP VLSI at Hailo. “Cadence’s Tensilica DSPs offer the performance and energy efficiency as well as the flexibility required for such heterogeneous devices.”
Vision 110 and 130 DSPs
The 128-bit Tensilica Vision 110 DSP and 512-bit Vision 130 DSP reportedly offer numerous performance improvements to address growing sensor and AI workloads, offloading from the CPU or GPU while providing industry-best energy efficiency. Both DSPs deliver AI performance enhancements that enable up to 3X improvement for certain AI workloads and up to 5X improvement on specific kernels, according to the release. Additional benefits include:
- Up to 20 percent higher frequency and up to 2X improvement for 16-bit, 32-bit and 64-
- bit floating-point performance
- Better system performance through iDMA enhancements previously available only on the
- flagship high-performance Vision DSPs
- Up to 3X improvements in fast Fourier transform (FFT) for radar applications
- Reduction in code memory footprint due to code size improvement
- Common SIMD and VLIW architecture and instruction set with Vision DSP predecessors
- ensure easy software migration
“Kneron continues to develop market-leading SoCs for numerous on-device and edge AI applications, and selecting an efficient processor to complement our high-performance NPU is crucial,” said Albert Liu, CEO of Kneron. “We continue to use Cadence Tensilica Vision DSPs, which provide the computational throughput and flexibility needed to keep up with today’s complex AI workload requirements. Additionally, Tensilica's extensive framework of software libraries and strong partner ecosystem enable rapid deployment for fast time to market.”
Tensilica HiFi and Vision DSPs support Cadence’s Intelligent System Design strategy, enabling SoC design excellence. The HiFi 3z DSP, HiFi 4 DSP, ConnX 110 DSP, ConnX 120 DSP, MathX 110 DSP and MathX 130 DSP have also been upgraded to the Xtensa LX8 platform.
The Vision 110 DSP and Vision 130 DSP, as well as the upgraded ConnX 110 DSP, ConnX 120 DSP, MathX 110 DSP and MathX 130 DSP, are available now. The new HiFi DSPs are expected to be in general availability in December 2023.