Beijing, China, May 26, 2020 – Xilinx, Inc. (NASDAQ: XLNX), the global leader in adaptive and intelligent computing, today announced UltraScale+ FPGAs optimized for networking and storage acceleration The newest member of the product family, the Virtex® UltraScale+™ VU23P FPGA, uniquely combines multiple resources to achieve more efficient packet processing and scalable data bandwidth, and is dedicated to breakthrough performance for networking and storage applications. In today’s world where the exponential growth of data places high demands on intelligent and flexible network and data center solutions, the new VU23P FPGA provides the industry with the maximum throughput, powerful data processing capabilities and flexibility advantages required to enable It can adapt to evolving connectivity standards and meet current and future needs.
Figure: Xilinx launches new Virtex UltraScale+ FPGA optimized for network and storage acceleration
The VU23P has an impressive set of features that enable the highest look-up table and embedded memory (block RAM) to DSP slice ratio in the Virtex UltraScale portfolio, enabling high-throughput processing at the same size and power consumption . It is packaged in a small 35mmx35mm package, providing super computing power in a small size, making it ideal for high-density server environments such as SmartNICs. The device’s 58G PAM4 transceivers can support up to 200G SmartNICs and networking systems with state-of-the-art PCIe Gen 4 connectivity for maximum I/O bandwidth. With VU23P FPGA acceleration, a variety of applications can benefit from the following significant benefits:
SmartNICs: By offloading processing to FPGA-based SmartNICs, users can dramatically increase server performance and reduce system-level total cost of ownership (TCO) while enabling rapid feature innovation. SmartNICs with Virtex UltraScale+ VU23P FPGAs enable applications up to 2x100G depending on the workload. Small package options provide flexibility to build diverse solutions for high-density server clusters.
NMVe-oF: NVMe over Fabrics, also known as NVMe-oF or the Fibre Channel Non-Volatile Memory Standard, is a protocol specification designed to connect hosts to storage over a network fabric using the NVMe protocol. The Virtex UltraScale+ VU23P provides enhanced acceleration for many types of storage workloads, including data compression, data decompression, data deduplication, and sorter functions, while improving overall storage utilization and minimizing CPU load. With a small form factor and PCIe Gen 4 compatibility, the new Virtex UltraScale+ FPGAs deliver unmatched performance per area, enabling flexible deployment and lowering storage system total cost of ownership.
Converged Access Fronthaul: In wired communications, using a converged access fronthaul gateway facilitates carrying all services over a single interface without degrading performance. The new VU23P FPGAs give different devices different “personality” and can be selected for each deployment in the network infrastructure, and it supports multiple types of standards, including Ethernet fronthaul, OTN fronthaul and PON fronthaul. In addition, the integrated PCIe Gen 4 IP also provides a way to virtualize the baseband processing of existing radios.
As a breakthrough device, the VU23P FPGA will enable a new generation of networking and storage system designers to quickly gain a head start on the competition.
Xilinx is committed to enabling rapid innovation in many different technologies, from endpoint to edge to cloud, by developing highly flexible and adaptive processing platforms. Xilinx is the inventor of FPGAs, hardware programmable SoCs, and ACAPs to deliver the industry’s most dynamic processor technology for an adaptive, intelligent, and connected future world.