WHITE PAPER
Enabling big data processing and AI-powered everything, everywhere
mrcy.com 4
Nothing is more symbolic of the current state-of-
the-art big data processing capability than Intel's
Xeon Scalable processor with on-die AI acceleration.
This data center processor uniquely features Deep
Learning Boost, which extends Advanced Vector
Extensions-512 (AVX-512) to accelerate inference
applications like speech recognition, image
recognition, language translation, object detection,
and other pattern manipulations. Each processor has
a new set of embedded accelerators (Vector Neural
Network Instructions, or VNNIs) that speed up dense
computations characteristic of convolutional neural
networks (CNNs) and deep neural networks (DNNs).
Mercury embeds these
processors in fog and
edge applications with
optional BuiltSECURE
systems security
engineering and
extreme environmental
protection packaging.
Miniaturization is created with system-in-package, wafer
stacking and other 2.5D and 3D fabrication technologies
Performance cooling – Effective and efficient conduction,
air, liquid and hybrid cooling technologies for embedded
processing are used to enable small form factor Xeon-powered
packages to operate reliably, at full throttle, delivering
unrestricted processing performance across wide temperature
ranges. Platform fuel and refrigerants are used in liquid-
cooled solutions and advanced air management approaches
are used in air-cooled solutions instead of the traditional,
less efficient air-blowing approaches used in data centers.
Next-generation air, liquid and hybrid cooling for reliable full-throttle big
data and AI processing at the edge
Unrestricted fabrics – Many open system architectures
have performance bottlenecks, especially between
their module interconnects. New backplane channel
and enhanced interconnect technologies enable
unrestricted switch fabric performance, eliminating
limitations to processing power and scalability across
large systems and broad temperature ranges.
Ethernet signal transmission "eye" taken from an enhanced
OpenVPX subsystem backplane channel - A wider eye produces
a reduced bit error rate
To avoid vendor lock and quicken technology adoption
and refreshes, Mercury technologies are compliant
with open systems standards. When combined, they
allow Intel Xeon Scalable and other cloud computing
capabilities to be physically deployed in a wide
spectrum of environments and form factors.