Semidynamics: From RISC-V with AI to AI with RISC-V
How to enhance AI compute with a scalable RISC-V NPU architecture
By Anne-Françoise Pelé, EETimes Europe | May 15, 2025
In just two years, Semidynamics has “evolved from RISC-V with AI to AI with RISC-V”, Semidynamics’ chief sales officer Volker Politz said at this week’s RISC Summit Europe 2025 in Paris.
In 2023, the Barcelona, Spain-based startup came out of stealth mode with a family of fully customizable 64-bit RISC-V cores, designed to process large amounts of data for machine learning, artificial intelligence, and high-performance computing. Also in 2023, Semidynamics released the customizable Vector Unit, delivering up to 2048 bits of computation per cycle for unprecedented data handling, and the Gazzillion Misses technology “to hide the memory wall and keep the machine working”, Politz said. Gazzillion Misses provides up to 128 cache misses per core and avoids idle times waiting for main memory to service the data. Semidynamics then released the RISC-V Tensor Unit to handle matrix multiplication required by AI. It integrates the company’s 64-bit fully customizable RISC-V core and Vector Unit, which is constantly fed with data by the Gazzillion Misses technology, so that there are no data misses.
To read the full article, click here
Related Semiconductor IP
- All-In-One RISC-V NPU
- 64-bit in-order RISC-V Customisable IP Core
- Fully-coherent RISC-V Tensor Unit
- Fully-custom RISC-V Vector Unit
- Multi-core capable 64-bit RISC-V CPU with vector extensions
Related News
- SiFive’s New RISC-V IP Combines Scalar, Vector and Matrix Compute to Accelerate AI from the Far Edge IoT to the Data Center
- Semidynamics Announces Cervell™ All-in-One RISC-V NPU Delivering Scalable AI Compute for Edge and Datacenter Applications
- Caspia Technologies Collaboration to Enhance Security Verification in Siemens' Questa One With Caspia's Generative AI Security Platform
- Axiomise Partners With Bluespec to Verify Its RISC-V Cores
Latest News
- Qualitas Semiconductor Successfully Demonstrates Live UCIe PHY IP at AI Infra Summit 2025
- Silicon Creations Announces 1000th Production FinFET Tapeout at TSMC and Immediate Availability of Full IP Library on TSMC N2 Technology
- Intel and NVIDIA to Jointly Develop AI Infrastructure and Personal Computing Products
- Comcores MACsec IP is compliant with the OPEN Alliance Standard
- Sofics joins GlobalFoundries’ GlobalSolutions Ecosystem to Enhance Chip Robustness, Performance and Design Efficiency