Machine Learning is a rapidly evolving technology that is re-inventing traditional applications and simultaneously inspiring new use models. These new applications deployed in the cloud and at on-premises Data Centers must perform in real-time, making it difficult for accelerators to keep pace.
Applications including Cloud Surveillance Analytics, Satellite Imaging, Bioinformatics, Financial Modeling, and Network Security require Machine Learning acceleration as well as other application-specific workloads involving OpenCV, Video Transcoding, and Smart NICs.
ML Application enables developers to optimize and deploy accelerated ML inference. ML Application supports the most prevalent machine learning frameworks including Caffe, MxNet, and Tensorflow, as well as Python and RESTful APIs.
xfDNN Compiler/Optimizer: auto-layer fusing, memory optimization, and framework integration
xfDNN Quantizer: Improves performance with auto model-precision INT8 calibration
Deployable on-premises or through cloud services
Xilinx Alveo™ Accelerator cards are capable of high-performance, energy-efficient real-time DNN inference.
An Alveo™ accelerator card running xDNN processing engines are capable of delivering more than 4,000 images per second of GoogLeNet v1 throughput at low latency without requiring batching.