In the financial services industry, enterprises that need to perform calculations for trading, market and credit risk, regulatory capital, limits, and initial margin can rely on the CompatibL Platform for their solution needs. As a leading Independent Software Vendor (ISV) in this space, CompatibL team lends their expertise to financial institutions and enables them to optimally execute their risk management applications, both on-prem and in the cloud. Part of that scoping effort from CompatibL includes the right sizing of hardware to ensure suitability for the specific needs of their customers.
While AMD announced its latest 5th Generation EPYC processors, codenamed “Turin”, on October 10, 2024, during Advancing AI Day, CompatibL evaluated CompatibL Platform on the latest AMD 4th Gen EPYC-based instances available in the cloud, namely Amazon Web Services (AWS).
CompatibL conducted their tests on a portfolio of 20,213 over-the-counter (OTC) derivatives and followed the workflow depicted in Diagram 1 below.
Diagram 1. Calculation workflow. Retrieved from the CompatibL case study on their website (https://www.compatibl.com/insights/amd-collaborates-with-compatibl/).
Within AWS, there are four AMD-based offerings that utilize 4th Gen EPYC processors: M7a (general purpose), C7a (compute optimized), R7a (memory optimized), and HPC7a (HPC optimized). M7a, C7a, and R7a instances are novel in their configuration – every vCPU is a physical CPU core, meaning that there is no Simultaneous Multi-Threading (SMT) enabled. For previous generations based on 3rd Gen EPYC processors (i.e., M6a, C6a, and R6a instances), every vCPU is a thread of a CPU core and SMT is enabled. This difference has implications for the levels of performance users can expect from the latest AMD-based instances, as each vCPU now leverages the full capabilities of a single CPU core.
Given its deep understanding of risk management calculations, CompatibL decided to select AWS M7a instances for its evaluation, driven by CompatibL Platform’s need for 4GiB of memory per core.
Diagram 2. Instance size options for the AWS AMD M7a instance type. Retrieved from the AWS website, under “product details” (https://aws.amazon.com/ec2/instance-types/m7a/).
CompatibL summarized its findings as follows:
- The AVX-512 instructions available in the 4th Gen AMD EPYC processors improved the performance of these compute-intensive calculations. In addition to achieving lower latencies, no degradation in CPU clock speed or instructions per second was observed.
- As the workload was parallelized across different M7a machine sizes, from m7a.2xlarge with 8 CPUs to m7a.12xlarge with 48 CPUs, full utilization of all CPU cores and nearly linear scaling were observed. In other words, as more CPUs were added to compute the same workload, the total run time reduced by the same proportion, which is critical when enterprises are looking to maximize compute resources and minimize cost overheads.
- 4th Gen AMD EPYC-based instances in AWS were well-suited for the cloud-native workload that CompatibL trialed, posing no challenges when deploying AWS Elastic Kubernetes Service (EKS) on the M7a instances of different sizes.
- No code changes were required to migrate their calculation workflow in CompatibL Platform from Intel to AMD for the experiments.
- AWS M7a instances are a robust solution for performing complex quant calculations and handling large portfolios, the likes of which were tested via CompatibL Platform.
Kumaran Siva, Corporate Vice President, Strategic Business Development, AMD, echoed the same sentiment: “As financial services embrace digital transformation, the 4th Gen AMD EPYC processors are designed to meet the demands of intensive calculations in cloud-native environments. CompatibL’s results showcase how our processors’ support for AVX-512 and all-physical core mode can accelerate financial analytics, helping firms optimize workloads and improve efficiency in real-time risk calculations."
AMD continues to engage with Independent Software Vendors in key verticals, showcasing that customer applications transition seamlessly to AMD EPYC processors with compelling performance, including in the cloud.
Beyond the ease of application migration and performance advantages, enterprises modernizing their data center infrastructure from legacy Intel-based servers can leverage 4th Gen AMD EPYC processors to lower their total cost of ownership (TCO), improve power efficiency, consolidate their data center footprint, reduce software licensing costs, and help achieve their sustainability goalsSP5TCO-073A.
The full white paper from CompatibL includes technical information and thorough explanations on their testing methodology. AMD recommends reviewing the comprehensive case study here on the CompatibL website for further details.
End Notes
SP5TCO-073A: As of 06/18/2024, this scenario contains many assumptions and estimates and, while based on AMD internal research and best approximations, should be considered an example for information purposes only, and not used as a basis for decision making over actual testing. The Server Refresh & Greenhouse Gas Emissions TCO (total cost of ownership) Estimator Tool compares the selected AMD EPYC™ and Intel® Xeon® CPU based server solutions required to deliver a TOTAL_PERFORMANCE of ~9020 units of VMmark3 matched pair performance based on the published scores (or estimated if indicated by an asterisk) for Intel Xeon and AMD EPYC CPU based servers. This estimation reflects a 5-year time frame. This analysis compares a 2P AMD 48 core EPYC_9474F powered server with a VMmark 3.1 score of 26.95 @ 26 tiles, https://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/vmmark/2024-05-14-Supermicro-AS-21...; compared to a 2P Intel Xeon 64 core Platinum_8592+ based server with a VMmark 3.1 score of 27.52 @ 28 tiles, https://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/vmmark/2024-04-16-Fujitsu-PRIMERGY...; versus legacy 2P Intel Xeon 28 core Platinum_8280 based server with a VMmark 3.1 score of 9.02 @ 9 tiles, Environmental impact estimates made leveraging this data, using the Country / Region specific electricity factors from the '2020 Grid Electricity Emissions Factors v1.4 – September 2020', and the United States Environmental Protection Agency 'Greenhouse Gas Equivalencies Calculator'. Results generated by: AMD EPYC™ Server Refresh & Greenhouse Gas Emission TCO Estimation Tool - version 1.51 PRO. VMmark is a registered trademark of VMware in the US or other countries. For additional details, see https://www.amd.com/en/claims/epyc#SP5TCO-073A