Cisco UCS M8 Servers with AMD EPYC? Establish World Record Benchmarks in Performance, Core Density, and Power Efficiency

We are just weeks away from the Summer Olympics in Paris, where world records will be set across a wide variety of sports. Just as athletes across the world have been training and preparing to compete, Cisco and AMD have been hard at work setting – and resetting – our own world records in computing.

But it’s not about winning medals. At a time when workloads like AI are pushing thresholds for reliable, secure, efficient, and high-performance computing, Cisco’s longstanding partnerships with companies like AMD, Intel and Nvidia are about enabling customers to support the next generation of apps while delivering better user experiences. We know that performance, and your choice in compute infrastructure matter more than ever.

What’s new

Cisco’s latest line of UCS M8 platforms powered by 4th Gen AMD EPYC? processors offer customers industry-leading performance and expandability for a wide range of demanding applications like big data analytics, databases, HPC, and AI workloads. These servers represent a significant leap in compute power, memory bandwidth, and I/O capabilities including 100 percent more cores per socket and up to 2.8X more performance the previous generation AMD servers.

For customers, this means faster application response times, improved user experiences, and the ability to handle the most demanding workloads. These benefits are available in our recently announced Cisco UCS C245 and C225 rack server models, available in 1U and 2U configurations, respectively.

Today, I’m excited to announce orderability for UCS x215c – which brings AMD to Cisco’s most popular line of UCS X-Series servers – and brings to customers the industry's first modular blade system with AMD 4th Gen EPYC? processors and built-in support for select next generation AMD EPYC? processors.

In addition to performance, this blade system can support up to 2048 cores in a fully populated chassis (256 cores per blade), making it among the densest computing platforms on the planet. By increasing core density per server, the UCS X215c blade architecture enables customers to consolidate more workloads onto fewer physical machines, resulting in significant savings in power consumption, rack space, data center footprint, and cooling requirements, while improving overall operational efficiency and reducing licensing costs.

The proof is in the pudding – or in this case, the world records

With Cisco’s UCS AMD M8 servers, we’ve added to our trophy case 20 new world records and established world record benchmarks for core density and power efficiency, two critical metrics that enable customers to achieve greater processing power and reduce energy consumption, ultimately lowering their operating costs and enhancing overall performance.

Below is a breakdown on the performance world records and the power efficiency metric using UCS C245 M8 Servers as the test platform. You can get more details about the performance tests and results in this technical brief – and feel free to reach out to me or my team with questions!

• Three world records for integer and floating-point performance, which are good predictors of general-purpose application performance. In addition to the power of 4th Gen AMD EPYC processors, we differentiate our performance through a unique design of hardware and firmware elements. These tests are crucial for ensuring the correctness and reliability of software applications, particularly those involving large mathematical calculations or data manipulation.
• Eight world records in performance tests of SPEChpc? 2021 Benchmark Suites, a set of application benchmark suites using a comprehensive measure of real-world performance for the state-of-the-art HPC systems. This is essential for evaluating the real-world performance of modern HPC systems by utilizing a diverse set of scientific and engineering applications and providing a comprehensive assessment of compute-intensive parallel performance across various hardware configurations and programming models. They offer well-selected science and engineering codes that are representative of HPC workloads and are portable across CPU and accelerators, along with certain fair comparative performance metrics.
• Four world records for SPEC ACCEL performance tests of parallel workload performance. SPEC ACCEL is a comprehensive tool for evaluating the performance of hardware accelerators and their supporting software toolchains to perform computational tasks. By testing applications under OpenCL, OpenACC, and OpenMP 4 target offloading APIs, it provides valuable insights into how well different hardware and software components interact to execute parallel workloads efficiently.
• One world record for Java enterprise middleware performance as measured by the SPECjbb?2015 MultiJVM benchmark. These tests are essential for ensuring optimal user experience by maintaining fast response times and scalability under heavy loads. It helps identify and address performance bottlenecks within the middleware components, like application servers and databases, enabling efficient resource utilization. These results demonstrate the achievable throughput while maintaining strict quality-of-service requirements for response times. These show how well the server can multitask between multiple Java virtual machines (JVMs).
• Four world records for SPEC OMP 2012 performance of shared memory parallel-processing applications. This is an important test suite for assessing shared-memory parallel processing capabilities for objective comparisons of different hardware and software configurations in HPC environments. It includes 14 samples from science and engineering environments, including computational fluid dynamics, molecular modeling, and image manipulation.
• The successful completion of Energy Star and Lot 9 efficiency compliance testing, which measures energy efficiency of platforms like computing systems. Based on the results, we can make the strong case that the UCS C245 is one of the most energy efficient 2U, 2-socket compute systems in the market today.

For more details on the Cisco UCS AMD M8 Servers, visit: https://www.cisco.com/c/m/en_us/solutions/computing/ucs-amd.html

?

Tags: #Cisco #AMD #Compute #UCS #EPYC #CPU #DataCenter

?