New Xilinx Virtex UltraScale+ FPGA Optimized for Networking and Storage Acceleration
May 27, 2020 -- Xilinx announced the addition of the Xilinx Virtex UltraScale+ VU23P FPGA to its UltraScale+ FPGA portfolio. In a blog posted below, Xilinx touts its performance and notes that the new addition is optimized for networking and storage solutions.
The exponential growth of data demands smart and adaptable network and data center solutions that deliver maximum throughput, high data-processing capability and the flexibility to adapt to evolving connectivity standards. To meet all of these requirements and more, we unveiled today the newest member in our UltraScale+ FPGA portfolio, the Xilinx Virtex UltraScale+ VU23P FPGA, delivering breakthrough performance for networking and storage applications. To quote “The Boss,” this new device is truly “Born to Run,” purpose-built with a unique mix of resources that enables highly efficient packet processing and scalable data bandwidth. The new VU23P FPGA delivers unparalleled performance with a powerful feature-set optimized for networking and storage acceleration.
Image courtesy of Xilinx.
The VU23P’s impressive specs include the highest ratio of look-up tables and embedded memory (block RAMs) to DSP slices in the Virtex UltraScale portfolio, enabling high-throughput processing within a fixed size and power envelope. It comes in packages as small as 35mmx35mm, delivering a lot of compute power in a small form factor, making it ideal for use in high-density server environments, including on SmartNICs. The device’s 58G PAM4 transceivers support up to 200G SmartNIC and network systems, with the latest connectivity capabilities—PCIe Gen 4—for maximum I/O bandwidth.
Here are a few of the applications that can reap dramatic benefits from VU23P FPGA acceleration:
- SmartNICs: By offloading processing to FPGA-based SmartNICs, users can significantly boost server performance and reduce system-level TCO while enabling rapid feature innovation. SmartNICs with a Virtex UltraScale+ VU23P FPGA can enable up to 2x100G applications, depending on workloads. The small package options enable flexibility in building diverse solutions for use in high-density server farms.
- NMVe-oF: NVMe over Fabrics, also known as NVMe-oF or non-volatile memory express over fabrics, is a protocol specification designed to connect hosts to storage across a network fabric using the NVMe protocol. The Virtex UltraScale+ VU23P enables increased acceleration of a variety of types of storage workloads including data compression, decompression, de-duplication and sequencer functionality while improving overall storage utilization and minimizing the burden on the CPU. With its small package sizes and PCIe Gen 4 compatibility, this new Virtex UltraScale+ FPGA offers unparalleled performance per square millimeter, creating flexibility in deployments and reducing overall TCO of storage systems.
- Converged-access fronthaul: In wired communications, converged access fronthaul gateways allow all services to be carried across a single interface without compromising performance. The new VU23P FPGA enables different device “personalities” to be selected for each deployment within the network infrastructure, supporting a variety of standards including fronthaul over Ethernet, fronthaul over OTN and fronthaul over PON. Furthermore, the integrated PCIe Gen 4 IP provides a path to virtualize baseband processing for existing radios.
Smart designers of next-gen networking and storage systems will run (not walk) to beat out the competition in benefiting from this ground-breaking device. To learn more, check out www.xilinx.com/vu23p
About Xilinx
Xilinx develops highly flexible and adaptive processing platforms that enable rapid innovation across a variety of technologies – from the endpoint to the edge to the cloud. Xilinx is the inventor of the FPGA, hardware programmable SoCs, and the ACAP, designed to deliver the most dynamic processor technology in the industry and enable the adaptable, intelligent, and connected world of the future. For more information, visit www.xilinx.com.
Source: Xilinx
