SmartNICs enable FPGA computation development efforts

Napatech announced the immediate availability of Link Programmable SmartNICs for compute-intensive workloads. Built upon Napatech’s hardware platform, the Link Programmable family of cards enables OEMs and end users to develop and deploy their own custom FPGA computation solutions based on industry-standard tools.

Link Programmable SmartNICs are ideally suited for computationally intense and time-critical workloads, such as computer vision, video encoding, high-frequency trading, cryptocurrency mining, satellite image processing, data mining, audio/video DSP, encryption/decryption, compression/decompression, and high-speed network data plane acceleration. The platform has superior design in performance per watt and cooling technologies. Using FPGAs for CPU inline and look-aside acceleration can enable up to 90x higher performance than general-purpose CPUs on key workloads and over 4x higher inference throughput than GPU-based solutions.

Key benefits:

• Fully programmable with immense flexibility: Provides increased performance, taking advantage of hardware parallelism with substantial cost savings over custom silicon.
• PCI-SIG certified and linkable: With focus on power, thermal and quality, 8+ SmartNICs can fit a standard COTS server, offering terabits/second, linked via onboard interconnects and/or via PCIe for massive parallel processing capacity.
• A flexible, configurable platform based on industry standard tools: Built around Xilinx UltraScale+ XCVU5P, the flagship platform is fully supported by the Xilinx Vivado Design Suite for rapid prototyping and product maturing. The platform is flexible and can be configured with various types of memory banks and FPGA chips – for example, XCVU7P or XCVU9P.
• New possibilities: The solution opens up Napatech cards to customers and users who wish to promote their own solutions. It also makes Napatech SmartNICs available to a new ecosystem of software and IP providers who have emerging applications and use cases that benefit from being deployed on a high-performance, multi-hundred GigE solution.