OmniXtend’s capabilities are based on an open-source architecture supported by a full spectrum of hardware, including ubiquitous Ethernet physical layer and programmable Ethernet switches, such as Barefoot’s Tofino™ P4-programmable switch.
Thanks to compatibility with the Tofino switch, architects in data centers – where main memory is central to the network – can implement innovative architectures that may support further advancements in areas including CPU micro-architecture and purpose-built compute acceleration in data-centric devices.
By using OmniXtend, system designers can take advantage of the many benefits of memory-centric architecture. Heterogeneous systems should all reside in the same memory domain and share memory in a coherent way. We can bring a new, open approach to design and enable machine-learning accelerators, custom ASICs, FPGAs, CPUs, and GPUs, all able to share one memory pool through the switch. We’re very, very excited to have achieved this next-generation milestone in OmniXtend.
For more information, visit the OmniXtend project blog and the open source repository.
