Neural Network Inference at Dramatically Lower Latency Compared to GPUs with Zebra by Mipsology on Xilinx Alveo U50 Accelerators

TOKYO, Oct. 15, 2019 /PRNewswire/ -- Mipsology announced Zebra software support for the Xilinx Alveo U50 Data Center accelerator card. Zebra is the industry's premier "plug & play" software for accelerating AI inference and delivering the best performance on FPGAs.

Neural network inference can now be deployed faster than ever with Zebra on the Alveo U50 Data Center accelerator card. The latest Alveo accelerator card has a small form factor and a power of just 75W. Accelerating the Zebra software stack with Alveo accelerator cards delivers best-in-class performance transparently, replacing CPUs or GPUs for inference in a data center or at the edge.

Zebra delivers a range of critical capabilities:

    --  Fastest CNN inference - Zebra sets a new standard of performance.
    --  Supports all neural networks - Zebra accelerates all layers of any CNN.
    --  Easy to use - Zebra is a "plug & play" solution, no changes or
        retraining of the NN are required.
    --  No changes to the software environment - No new languages, frameworks or
        tools. Zebra works from within TensorFlow, PyTorch, ONNX, Caffe and
        MXNet.
    --  Scalable, flexible and adaptable - Zebra can replace GPUs or complement
        CPUs anywhere.

"We are excited to enable customers with Zebra support for the Alveo U50, the latest of the Alveo cards and the first half-height, half-length accelerator built with PCI Express 4.0 and HBM2 memories. Leveraging Zebra's AI compute capabilities, gives our customers unprecedented ability to easily accelerate neural networks and realize significant total cost of ownership benefits," said Ramine Roane, vice president of Software and AI Solutions Marketing, at Xilinx.

Zebra Software Support for Xilinx Alveo U50 Data Center Accelerator Card

"We are very pleased to have Zebra running now on the Xilinx Alveo U50. Zebra delivers better inference performance for AI neural networks. We are excited about this new board for its size, the large HBM bandwidth and ability to unlock Zebra's performance and ease of use in data center and edge environments. All with less than 75W, no cooling required!" said Ludovic Larzul, Founder and CEO, at Mipsology.

Mipsology demonstrated Zebra live on the Xilinx Alveo U50 and Alveo U250 cards at the Avnet Acceleration Seminar in Tokyo Japan running 4 live camera streams at low latency.

About Avnet
Avnet is a global technology solutions provider with an extensive ecosystem delivering design, product, marketing and supply chain expertise for customers at every stage of the product lifecycle. We transform ideas into intelligent solutions, reducing the time, cost and complexities of bringing products to market. For nearly a century, Avnet has helped its customers and suppliers around the world realize the transformative possibilities of technology. Learn more about Avnet at www.avnet.com.

About Mipsology
Mipsology is a groundbreaking startup focused on state-of-the-art acceleration for deep learning inference. The team devised Zebra - the first technology to accelerate the computations of inference for neural networks on FPGA and conceal the hardware to AI users. Learn more about Mipsology at www.mipsology.com

View original content:http://www.prnewswire.com/news-releases/neural-network-inference-at-dramatically-lower-latency-compared-to-gpus-with-zebra-by-mipsology-on-xilinx-alveo-u50-accelerators-300938408.html

SOURCE Mipsology