Increased Total Expansion Capability from Its Previous Generation

The successor to PNYSER48 series, PNYSRA48 series adds a 9th slot to the rear of the chassis, allowing a single-width PCIe x16 card to be installed next to 8 double-width GPU cards. This new design change is ideal for NVMe deployment or high speed networking like 100 Gigabit EDR Infiniband, 100 Gigabit Ethernet, or 100 Gigabit Intel Omni-Path fabric. Furthermore, an internal storage mezzanine slot (for SAS storage deployment without taking up PCIe slots) is also available for integrating device controller solutions. PNYSRA48 series does not require a specialized domed chassis lid for the power connectors of active GPU cards.

 

 

Dual-Root Complex Design

As a dual-root complex GPU server, PNYSRA48 series deploys a pair of Intel® Xeon® Scalable Processors which split the workload of managing 8 GPU cards. The motherboard deploys a set of four PLX PEX8747 PCIe switch chips to provide the connectivity for the 8 GPU card PCIe x16 slots. PNYSRA48 series also comes with an on-board mezzanine board for Intel Omni-Path 100 Gigabit networking fabric.

IPMI 2.0 with Redfish Support

Featuring AMI MegaRAC® firmware, PNYSRA48 series offers support for IPMI 2.0 with iKVM and the DMTF Redfish management protocol. The MegaRAC® SP-X SoC (System-on-Chips) has an AMI generic, user-friendly Graphics User Interface (GUI) called the MegaRAC® GUI. It is designed to be easy to use. It has a low learning curve because it uses a standard Internet browser. You can expect to be up and running in less than five minutes.

Extend Your Scale with Intel® Xeon® Scalable

PNYSRA48 series is built around Intel Skylake CPU architecture - ‘Intel® Xeon® Scalable’. This new product family provides a wealth of new options in scalability - across compute, network and storage - to deliver solutions for a range of applications.

 

CPU Key Features

Intel® Xeon® Scalable CPU architecture brings improved features and functionality at both the host and fabric levels. These enable performance improvements - both natively on the chip and for future extensibility through compute, network and storage peripherals.

Memory Optimization:
Memory speed of up to 2666MHz
Up to 12 DIMMs per socket

Storage:
14x 2.5” SSD/HDD drives, up to 8 SAS drives, up to 4 x U.2 NVMe devices without occupying any additional expansion slots

Networking:
Intel Omni-Path ready

PNYSRA48 System Overview

PNYSRA48 Front View

 

HIGH PERFORMANCE COMPUTING (HPC)

Challenge in HPC:

Intensive and fast algorithmic computing, such as volume rendering, data-parallel vector calculation, multi-layer data linearity transformation, molecular fragments modeling.
GPU Computing as the Solution:

 

DEEP LEARNING

Challenge in Deep Learning:

High data training and validation iterations for linear and non-linear complex features with multiple processing layers and parameters; rapid object detection and parsing.

GPU Computing as the Solution: