Amd and xilinx beat the world record in image inference

At today's Xilinx Developer Forum in San Jose, California, Xilinx CEO Victor Peng joined AMD Chief Technology Officer Mark Papermaster to reveal that the two have been working together to connect AMD EPYC CPUs, and the new line of Xilinx Alveo acceleration cards, for high-performance, real-time AI inference processing. In doing so, they revealed a world record for inference performance of 30, 000 images per second.

New image inference record from AMD and Xilinx

The impressive system leverages two AMD EPYC 7551 server CPUs with its industry-leading PCIe connectivity, along with eight of the recently announced Xilinx Alveo U250 acceleration cards. Inference performance is powered by Xilinx ML Suite, which enables developers to optimize and implement accelerated inference, and is compatible with numerous machine learning frameworks, such as TensorFlow. The benchmark was performed on GoogLeNet, a widely used convolutional neural network.

We recommend reading our post about AMD Ryzen Threadripper and AMD EPYC

AMD and Xilinx have shared a common vision on the evolution of computing to a heterogeneous system architecture, and have a long history of technical collaboration. Both companies have optimized drivers and optimized performance for interoperability between AMD EPYC CPUs with Xilinx FPGAs.

AMD EPYC is the perfect CPU platform to accelerate artificial intelligence and high-performance computing workloads. With 32 cores, 64 threads, 8 memory channels with up to 2TB of memory per socket, and 128 PCIe lines, along with the industry's first x86 server security solution with embedded hardware, EPYC is designed to deliver the best capacity memory, bandwidth and processor. Xilinx and AMD see a bright future in their technology collaboration.

Techpowerup font