Do you need a cost-effective, industry-standard server for highly parallel applications requiring performance across widely diverse workloads and legacy applications not optimized well for graphic processing unit (GPU) computing? The HPE Apollo kl20 Server is a 2U server with up to four Intel® Xeon® Phi processors and with support for both Intel Omni-Path and Mellanox® InfiniBand interconnect fabric options for flexibility and increased price-performance and density. An ideal solution for running highly parallel applications optimized for the Xeon Phi Processors, especially for code which has already been optimized for the prior generation of the Intel Phi coprocessor and legacy applications for which recoding would be burdensome or where the environment requires performance across widely diverse workloads not optimized well for GPU computing. Delivers a dense and flexible computing platform optimized for highly vectorized or memory bandwidth bound HPC applications.

Features

Flexible Configurations Tailored to HPC and Deep Learning

Fully integrated and factory tested clusters are delivered with a complete cluster management software stack to simplify deployment and management, and can be custom configured for perfect sizing.

Multiple network topologies are supported in cluster builds: fat tree, hypercube, and enhanced hypercube.

Optimize the server for required workloads by choosing from the available CPUs in the Intel Xeon Phi Processor Family, up to twenty-four HPE 2400 MHz DDR4 DIMMs, optional network adapters, and up to twelve SFF hard drives or solid state drives (SSD) per four-node 2U chassis.

The HPE Apollo kl20 Server is scalable to more than sixty cores per node and greater than 3 TFlops per socket versus Intel Xeon v4.

Integrated on-board 16 GB MC-DRAM memory delivers up to 5x bandwidth per processor vs DDR4 on Intel Xeon v4.

Customized Configuration Compute for Infrastructure Efficiency

Integrated IPMI 2.0 enables standards-based systems-management.

Select the optimal network topology for your target cost model. For example, fat tree can be a strong, general purpose fabric which is effective for small systems, but can be expensive for large systems.

The system is designed for energy efficiency with dual 2000W Titanium Level Power Supplies.

The HPE Apollo kl20 Server is air-cooled for easy integration into standard server environments.

Expertise and Resources for Operational Efficiency

Committed account and technical resource teams.

Enable custom solutions, as one size does not fit all requirements.

Delivers responsive global support.

The HPE Apollo kl20 Server alongside HPE Pointnext services provides expertise in HPC, data management, visualization, systems management, and high performance data analysis (HPDA).

Technical Specifications

Memory

• 6 DIMM slots, DDR4, 2400 MT/s, 384 GB (6x 64 GB) maximum

Management features

• Integrated IPMI 2.0

Storage

• 12 LFF hot plug SATA drives

Form factor

• 2U

Power

• Two 2000W Titanium level power supplies

CPU

• Intel® Xeon® Phi Series

Network

• 1GbE, 10GbE and 100 Gb/s InfiniBand

Boot capabilities

• 1) Direct from HDD 2) Network

System configuration(s)

• Ships standard with one CPU per server, various networking options (1GbE, 10GbE and/or IB 100GbE/OPA 100GbE), up to 6 DIMMs per server and 12 LFF drive slots and integrated fabric.