Stealth startup Vybium developing European RISC-V AI accelerator
Cette publication existe aussi en Français
European startup Vybium is developing am AI accelerator chip using the open RISC-V instruction set architecture to take on the Nvidia A100 GPU in the data centre.
Vybium is a fabless semiconductor startup still operating in stealth mode and is using RISC-V for European Digital Strategic Autonomy for AI. It was founded by VRULL in Austria and Software Ecosystem Solutions in Albania to develop RISC-V AI accelerators for all verticals.
TO speed up its chip development, it has licensed the NPU IP from Stream Computing and is adding new data types, support for sparsity, and higher-bandwidth memory solutions to the integrated design.
“Vybium has always been about delivering European-developed products that embody the European perspective: providing end-to-end solutions based on RISC-V that address the needs of European industrial companies. The rapid growth of AI/ML across all applications has changed customer priorities. We are addressing these on a timely basis based on the production-proven NPU IP from Stream Computing,” said Dr. Philipp Tomsich, chief technologist and founder of Vrull.
“Empowering Vybium to develop their own silicon by incorporating and innovating on our production-proven NPU IP accelerates the roadmap to European AI/ML accelerators competing with Nvidia. Having a shared technology provides a broader base to support common, open-source software enablement. This is a true win-win,” said Andy Mei, CEO of Stream Computing.
The first generation of Vybium productions will focus on AI/ML accelerator cards and are intended to compete with the Nvidia A100 family. Systems combining RISC-V general-purpose compute with AI/ML for embedded, industrial, and edge use cases are expected to follow the initial designs targeted at AI/ML acceleration in the datacentre, cloud, and enterprise applications.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
