Built on Ceva’s XC20 architecture, already in design with two Tier-1 infrastructure OEMs for 5G-advanced and pre-6G processors, these DSPs are intended to enable faster, more efficient data processing while reducing latency and increasing throughput.
Both DSPs come with AI support, allowing customers to apply Machine Learning to optimise modem algorithm performance and network efficiency for UE and infrastructure and future-proof their designs for evolving wireless standards.
The Ceva-XC21 5G IoT DSP is designed for low-power, cost- and size-optimised cellular IoT modems, NTN VSAT terminals, eMBB and uRLLC applications and, as the successor to the widely adopted Ceva-XC4500 DSP, the XC21 offers a size reduction of up to 48% compared to the Ceva-XC4500, while delivering the same performance at 63% of the area.
The Ceva-XC23 DSP targets regenerative NTN satellite payloads, high-end user equipment (UEs) and infrastructure baseband processing, including baseband units (BBUs), distributed units (DUs) and radio units (RUs). It provides high-performance capabilities for 5G-advanced and pre-6G applications with performance boost of up to 2.4 times and a 2.3X efficiency improvement compared to the Ceva-XC4500.
"Our latest vector DSPs represent a significant leap forward in performance and efficiency for advanced 5G and 6G applications. These new DSPs offer a robust, future-proof platform to develop highly efficient next-generation modems and infrastructure ASICs. With AI support and programmability, they not only improve modem performance but also enable advanced AI and ML workloads, ensuring optimal network performance and efficiency towards 6G," said Guy Keshet, vice president and general manager, Mobile Broadband Business Unit at Ceva.
The Ceva-XC21 and Ceva-XC23 DSPs are built on top of the scalable, multi-thread Ceva-XC20 architecture.
Key features include:
Ceva-XC21: 256b SIMD vector size, single/dual thread options, and 64/32 16x16-bit MAC options
Ceva-XC23: 512b SIMD vector size, dual thread design, and 128 16x16-bit MACs for demanding applications
Both DSPs feature:
- 5th generation Ceva-XC Communication Vector DSP Architecture
- AI support for 8-bit neural networks
- Dual threading with Dynamic Vector Threading (DVT)
- Enhanced 5G ISA for accelerated 5G channel processing
The Ceva-XC21 and Ceva-XC23 DSPs ensure easy migration for existing Ceva-XC customers, with code compatibility with other Ceva-XC20 products and ISA compatibility with the Ceva-XC4500.
Both devices will be available for general licensing by the end of the first quarter.
