News

Users can achieve equal performance at a lower cost – as much as 40% lower than the competition.

Cornelis Networks is the leading independent provider of purpose-built, open-source, scale-out interconnects for high-performance computing (HPC), artificial intelligence (AI), and data analytics. The Cornelis Networks Omni-Path 100Gbps high-performance networking fabric delivers class-leading message rate, latency, and scalability, while allowing customers to deploy solutions which enable faster time to solution and improved workload scalability combined with leading price/performance.

To highlight the capabilities of Omni-Path, this paper analyzes the performance of Ansys Fluent, an industry leader in Computational Fluid Dynamics (CFD) provided by Ansys. Ansys Fluent is a versatile CFD tool that includes well validated physical modeling capabilities to deliver fast, accurate results across a wide range of CFD and multiphysics applications.  It is widely used in automotive, aerospace, academia, oil & gas, marine, and Formula 1 racing applications. Typical workload sizes range from two million to 500 million cells.

Imagine getting the most simulation throughput out of your investment by using a combination of the right software on the right networking fabric in your environment.  When using Ansys Fluent with Omni-Path, users will see comparable and even indiscernible differences in performance versus NVIDIA^® InfiniBand HDR.  Since users often find Cornelis Omni-Path to be 25-40% less expensive than NVIDIA InfiniBand, depending on the exact cluster configuration, it makes both operational and financial sense to choose Omni-Path.

Cornelis Networks ran the standard benchmark suite released with Ansys Fluent 2022 R1 on up to 16 dual-socket Intel^® Xeon^® Platinum 8358 processor nodes, each installed with Cornelis Omni-Path and NVIDIA InfiniBand HDR. Intel MPI Library 2019.12 as packaged with Fluent were used for both fabrics, and the Cornelis Omni-Path Express (OPX) libfabric provider was used for Omni-Path. Complete configuration is given at the conclusion of the paper.

Figure 1 shows the Solver Rating (i.e., jobs per day) when running on 16 nodes normalized by the cost of building a 16-node cluster comprised of 16 PCIe adapters, 16 copper cables, and one managed edge switch¹. Fluent customers can achieve up to 1.55X higher job throughput per fabric cost when using Cornelis Omni-Path compared to NVIDIA InfiniBand HDR. This ultimately means that customers running Cornelis Omni-Path can deploy an equally performant cluster for less money or redirect unused budget to purchase additional Ansys software licenses or compute power.

The new Cornelis OPX provider for libfabric unleashes additional performance on existing Omni-Path 100 Gbs hardware and is the provider of choice for Cornelis Networks’ next generation CN5000 fabric. Thanks to a streamlined code path designed specifically for libfabric, it provides small message latency and message rate advantages over the traditional PSM2 provider. OPX is a drop-in replacement for PSM2 and can be evaluated for performance side-by-side with PSM2. Obtain OPX either through the Cornelis Omni-Path Express Suite (OPXS) software download at www.cornelisnetworks.com or build your own using libfabric version 1.16.1 or newer from https://github.com/ofiwg/libfabric.

Cornelis Omni-Path 100 is ready to ship with minimal lead time; contact your vendor of choice today to start experiencing 1.55X higher Ansys Fluent performance with Cornelis Omni-Path 100!

System Configurations

Tests performed on 2 socket Intel Xeon Platinum 8358 CPU @ 2.60GHz. Intel Hyper-Threading Technology disabled. Intel Turbo Boost Technology enabled with ACPI driver. Rocky Linux 8.4 (Green Obsidian). 4.18.0-305.19.1.el8_4.x86_64 kernel. Memory: 32x32GB, 512 GB total, 3200 MT/s.

NVIDIA InfiniBand HDR: MLNX_OFED_LINUX-5.4-2.4.1.3, mlx5_core version 5.4-2.4.1. FI_PROVIDER=mlx, UCX 1.15.0 from hpcx-v2.16-gcc-mlnx_ofed-redhat8-cuda12-gdrcopy2-nccl2.18-x86_64.

Cornelis Omni-Path: Omni-Path Express Fabric Suite (OPXS) 10.11.1.1.1, FI_PROVIDER=opx, libfabric 1.18.0.

The following data are used to generate the plot in Figure 1. These benchmarks scaled with greater than 70% efficiency at 16 node scale. The solver ratings are an average of five runs. The normalized standard deviation for the tests ranged from 0% to 3%.

Legal Disclaimer

You may not use or facilitate the use of this document in connection with any infringement or other legal analysis concerning Cornelis Networks products described herein. You agree to grant Cornelis Networks a non-exclusive, royalty-free license to any patent claim thereafter drafted which includes subject matter disclosed herein.

No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.

All product plans and roadmaps are subject to change without notice.

The products described may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request.

Cornelis Networks technologies may require enabled hardware, software, or service activation.

No product or component can be absolutely secure.

Your costs and results may vary.

Cornelis, Cornelis Networks, Omni-Path, Omni-Path Express, and the Cornelis Networks logo belong to Cornelis Networks, Inc. Other names and brands may be claimed as the property of others.

Copyright © 2024, Cornelis Networks, Inc. All rights reserved.

¹ MSRP Pricing obtained on 7/11/2023 from https://store.nvidia.com/en-us/networking/store. Mellanox MCX653105A-HDAT $1628 per adapter. Mellanox MQM8700-HS2F managed HDR switch, $25555. MCP1650-H002E26 2M copper cable – $281. Cornelis Omni-Path MSRP pricing as of 7/11/2023. Cornelis 100HFA016LSN 100Gb HFI $880 per adapter. Cornelis Omni-Path Edge Switch 100 Series 48 port Managed switch 100SWE48QF2 – $19750. Cornelis Networks Omni-Path QSFP 2M copper cable 100CQQF3020 – $147. Exact pricing may vary depending on vendor and relative performance per cost is subject to change.