Administrator Guide

2
Copyright © 2019 Dell Inc. or its subsidiaries. All Rights Reserved. Dell, EMC and other trademarks are trademarks of Dell Inc. or its subsidiaries
Copyright © 2019 Dell Inc. or its subsidiaries. All Rights Reserved. Dell, EMC and other trademarks are trademarks of Dell Inc. or its subsidiaries
Figure 1b: Dell EMC R740 PowerEdge server
Recent Optimizations
In part 1 of this blog, we presented the Intel
Acceleration Stack and the Intel distribution of
OpenVINO both of which are part of the system
stack for the Intel PAC. Within the PAC, the Deep
Learning Accelerator (DLA), part of the PAC
hardware stack, accelerates specific deep learning
models, e.g., AlexNet, GoogleNet, ResNet, etc.
Intel has released an updated ResNet-50 binary [2],
and also released new versions of OPAE and
OpenVINO. With these new improvements,
developers can derive increased acceleration in
their deep-learning inference workloads with minimal changes to the system stack (Fig 2).
New ResNet-50 Binary
ResNet-50 is a deep learning model for image classification, allowing applications to describe an
image with only 3.57% error. Like any other deep neural network, ResNet-50 has input, output,
and hidden layers which describe its underlying algorithm through a network of interconnected
neurons that propagate information from one layer to the next (Fig 3.)
As new research in deep learning architectures and models continue to evolve, FPGAs are
uniquely positioned to incorporate these research advancements down into the hardware without
necessitating a new silicon spin. Intel has further optimized the ResNet-50 model for low-bit
(FP11) precision inferencing, enabling increased performance of vision applications. Further, this
new model optimization at the hardware level, combined with the many software optimizations,
provides seamless application integration while significantly increasing performance.
Figure 3: ResNet-50 model architecture.
Figure 2: System stack