Advantech FPGA Platforms for Artificial Intelligence Acceleration
FPGAs are becomingly increasingly well regarded for their ability to run the deep neural networks behind AI more efficient than CPUs or GPUs, yet still retain the software re-programmability that sets them apart from ASIC options. Advantech‘s VEGA Series of Xilinx FPGA-based PCI Express cards have been designed to accelerate AI deep learning workloads, offering a range of integration options to suit customer needs. In particular, the range focuses on the needs of video-based, real-time AI inference applications at the edge where fastest and more efficient video processing is critical.
Lowest Latency AI Inference
Ultra-Low Power Consumption
Flexibility for Reconfiguration
Optimized for Video Intelligence
Cloud-based Artificial Intelligence
The Advantech VEGA 4000 Series are based on Xilinx’s Virtex UltraScale+ VU9P FPGAs which are also deployed widely in public cloud installations thanks to their high performance coupled with excellent power efficiency. FPGAs are considered by many as the optimum engines for cloud-based AI deep learning applications, and with VEGA 4000 Series, customers are able to add this capability to their cloud edge appliances and private or hybrid cloud installations.
The Advantech VEGA 4000 Series are based on Xilinx’s Virtex UltraScale+ VU9P FPGAs which are also deployed widely in public cloud installations thanks to their high performance coupled with excellent power efficiency. FPGAs are considered by many as the optimum engines for cloud-based AI deep learning applications, and with VEGA 4000 series, customers are able to add this capability to their cloud edge appliances and private or hybrid cloud installations.
AI Video Content Analysis
The Advantech VEGA 500 Series are based on the Xilinx Zynq Ultrascale+ ZU7EV MPSoC which adds dedicated video encoding/decoding blocks and ARM processor cores to the programmable logic array for extra scalability. That makes the VEGA 500 series especially suitable for applications that require a lot of video processing as part of the workflow. This is especially relevant when used in parallel with power efficient Inference Engines running in the Programmable logic.
The Advantech VEGA 500 Series are based on the Xilinx Zynq Ultrascale+ ZU7EV MPSoC which adds dedicated video encoding/decoding blocks and ARM processor cores to the programmable logic array for extra scalability. That makes the VEGA 500 Series especially suitable for applications that require a lot of video processing as part of the workflow. This is especially relevant when used in parallel with power efficient Inference Engines running in the Programmable logic.
Get Ready for a Multi-Codec World
Reconfigurable FPGA-based video processing is becoming more appealing to OTT infrastructure users as the range of codecs required for OTT applications grows and the corresponding need for acceleration escalates. In addition to the established video codec standards like H.264/AVC and H.265/HEVC, OTT service providers now need to consider additional use cases for codecs like VP9 and potentially even AV1 in the near future. FPGAs offer a flexible and future proof way to accelerate multiple live codecs and can be upgraded over time.
Reconfigurable FPGA-based video processing is becoming more appealing to OTT infrastructure users as the range of codecs required for OTT applications grows and the corresponding need for acceleration escalates. In addition to the established video codec standards like H.264/AVC and H.265/HEVC, OTT service providers now need to consider additional use cases for codecs like VP9 and potentially even AV1 in the near future. FPGAs offer a flexible and futureproof way to accelerate multiple live codecs and can be upgraded over time.
Tailored for AI Inference at the Edge
Video-based AI applications that can make use of this inferencing acceleration include self-driving cars, retail analytics, and surveillance / security zone tracking. Advantech have been working closely with both Xilinx and several 3rd party vendors to provide the accelerated Deep Neural Network libraries that can underpin these applications using commonly available AI frameworks such as Mxnet, Caffe and Tensorflow.
Object and Lane Detection for Autonomous Driving
Object and Face Recognition for Safety and Security
Live Video Analytics for Mobile Advertising
Heat Map Analytics for Retail Intelligence
The Right Solution for Your FPGA Video Accelerator
The VEGA range of FPGA video accelerator boards are exclusively based on Xilinx FPGAs. Xilinx has established a leadership position in FPGA-based machine learning inferencing, including use of “DeePhi” inferencing optimization that substantially increases performance per watt. With a range of physical formats available, infrastructure builders can integrate Advantech VEGA accelerators into edge appliances and scale-out data center servers to deploy accelerated AI services in the cloud or closer to end-users on accelerated AI edge platforms for real-time inferencing.
Get Started with the VEGA 4000/500 Series
Do not hesitate to contact us for more information about Advantech Edge-to-Cloud AI Solutions. For companies with special requirements, Advantech provides AI video solutions supported by a global team of experts with broad experience on integrating, customizing & designing FPGA-based video products. Contact us today to discuss your unique product needs.
Do not hesitate to contact us for more information about Advantech Edge-to-Cloud AI Solutions. For companies with special requirements, Advantech provides AI video solutions supported by a global team of experts with broad experience on integrating, customizing & designing FPGA-based video products. Contact us today to discuss your unique product needs.
