AMD EPYC Gen 4 vs Intel Xeon 5th Gen: A Battle for Server Supremacy
The recent availability of 5th Gen Intel Xeon processors allows testing of 4th Gen AMD EPYC CPUs on critical workloads. AMD EPYC 9654 beats competitors in data centre performance, power efficiency, and cost. AMD EPYC CPU hold over 300 performance world records, and Intel’s latest release maintains our lead. Over 250 server designs and 800 cloud instances comprise AMD EPYC.
SPECpower
Modern data centers manage power utilization to meet growing needs, save costs, and satisfy environmental goals. The System Under Test’s power and performance are used to compare volume server class computers’ energy efficiency in the SPECpower ssj 2008 benchmark. SPECpower was the first industry benchmark for single- and multi-node server power and performance.
Consumers can compare server and configuration energy efficiency using the SPECpower ssj 2008 benchmark. This benchmark applies to hardware, IT, OEM, and government companies. The SPECpower ssj 2008 statistic assesses SUT power efficiency as “overall ssj ops/watt”. It is the ratio of the overall throughput ssj ops, which is the sum of all target load scores, to their average watt power consumption.
AMD EPYC 9754
AMD EPYC 4th Gen excels at power efficiency. A dual-socket system with 128-core AMD EPYC 9754 processors outscored Intel Xeon Platinum 8592+ by 2.25x.
AMD EPYC 9554
Industry benchmarks for compute-intensive jobs include the SPEC CPU 2017 test, which stresses the processor, memory subsystem, and compiler on various computer systems. This blog discusses SPECrate 2017 Integer and Floating Point, two of SPEC CPU 2017’s four suites’ 43 benchmarks.
On SPECrate 2017, 32-core AMD EPYC 9374F machine surpasses Intel Xeon Scalable 8562Y+ by ~1.14x (base) and ~1.06x (64 cores). 2P 64-core AMD EPYC 9554 systems outperform Intel Xeon Scalable 8592+ in SPECrate 2017 and similar.
Top performer: SPECrate 2017 ranked 2P 64-core Intel Xeon Scalable 8592+ behind 2P 96-core AMD EPYC 9654. SPECjbb2015 was considered simulates a corporate IT infrastructure that handles point-of-sale requests, online transactions, and data-mining to test server-side Java applications. Because Java is so popular, JVM providers, hardware manufacturers, Java application developers, academics, and academia should evaluate this benchmark.
AMD EPYC 9554 64-core processor
64 cores A dual-socket 64-core AMD EPYC 9554 machine performs comparably to an Intel Xeon Scalable 8592+ on SPECjbb 2015 MultiJVM-maxJOPS.
Top of stack: A dual-socket 96-core general purpose AMD EPYC 9654 machine outperforms a 64-core Intel Xeon Scalable 8592+ system 1.48x on SPECjbb 2015 MultiJVM-maxJOPS.
Decision Support System
The decision support benchmark TPC Benchmark H (TPC-H) evaluates systems that handle complex business queries by running sophisticated queries over big datasets. This benchmark’s queries and data manipulations affect several industries.
TPC-H’s Composite Query-per-Hour Performance Metric (QphH Size) evaluates query processing. When handling concurrent user requests, database sizes, query stream processing, and query throughput are considered. A system’s price-performance ratio, or “bang for the buck,” is likewise assessed by TPC-H.
A 2P 64-core AMD EPYC 9554 system outperforms a 2P Intel Xeon Scalable 8592+ system by ~1.14x and ~1.22x at SF 10000 in price-performance TPC-H.
Business Uses
Key SAP ERP logistics module is SAP Sales and Distribution (SAP SD). SAP Application Performance Standard units’ database efficiency is measured using the SAP-SD 2-Tier benchmark to assess hardware performance. SAPS is hardware-independent and uses the Sales and Distribution (SD) benchmark to evaluate SAP systems. Recently, a 2P system with 96-core AMD EPYC 9654 processors outscored a 64-core Intel Xeon Platinum 8592+ by ~1.53x. For further information, refer to my blog post “4th Gen AMD EPYC CPUs Empower SAP SD 2-Tier Performance.”
In four generations of general-purpose AMD EPYC processors, “top of stack” CPU SKUs enhanced performance. It also indicates that AMD EPYC 2nd Gen and later outperform Intel Xeon processors.
The Virtualization Infrastructure
VMmark 3 benchmarks virtualized server performance, power efficiency, and scalability on physical hardware under intense loads. Server capabilities and virtualization solutions are compared for strengths and limitations.
AMD EPYC 9654
A 2P 64-core AMD EPYC 9554 machine outperforms a 2P Intel Xeon Scalable 8592+ system 1.29x on VMmark 3.
In VMmark 3, a 2P 96-core AMD EPYC 9654 system outperforms a 2P 64-core Intel Xeon Scalable 8592+ system by ~1.60x.
Powerful computing
HPC predicts major climate events and enhances transit and infrastructure safety, making it part of almost every modern activity. Optimizing material use, simplifying designs, and lowering development costs lowers prices. Rapid virtual prototyping reduces physical testing and speeds market entry.
Demand for HPC workload performance rises. Performance increases enable faster simulations, shorter product development cycles, more scenario testing, and finer model tweaks, making products more effective and efficient. Explore 4th Gen AMD EPYC CPUs HPC dominance.
AMD EPYC 9374F
CFD models and studies fluid behavior like water flowing around a boat hull or air around a vehicle or aircraft using numerical analysis. Uses include industrial and consumer goods. Memory bandwidth limits computationally demanding CFD activities.
Altair AcuSolve: Without CFD, Altair AcuSolve allows organizations investigate ideas using flow, heat transfer, turbulence, and non-Newtonian material analysis. In AcuSolve’s acus-in test, a 2P 32-core AMD EPYC 9374F machine surpasses an Intel Xeon Scalable 8562Y+ system by ~1.46x.
Ansys CFX: High-performance CFD software provides fast, reliable, and accurate solutions for a variety of CFD and Multiphysics applications. In several CFX tests, a 32-core AMD EPYC 9374F machine outperformed an Intel Xeon Scalable 8562Y+ system by ~1.48x.
Ansys Fluent: Advanced physics modelling and industry-leading accuracy characterize this fluid simulation package. A 2P 32-core AMD EPYC 9374F machine outperformed a 2P Intel Xeon Scalable 8562Y+ system performance ~1.25x in chosen Fluent tests.
OpenFOAM CFD software is free and open-source. Academic and commercial institutions use it. In OpenFOAM tests, a 2P 64-core AMD EPYC 9554 system outperformed an Intel Xeon Scalable 8592+ and a top-of-stack general purpose system by ~1.14x. A 2P 96-core AMD EPYC 9654 system outscored a 64-core Intel Scalable 8592+ system by ~1.22 in FEA Clear.
An explicit Finite Element Analysis (FEA) numerical simulation evaluates structures and materials under dynamic situations like impacts, explosions, and crashes. The auto industry uses FEA to predict crash behavior and evaluate occupant safety. Cell phone manufacturers use FEA to simulate drop tests for durability. Simulations save manufacturers time and money by testing designs online instead of prototyping.
These simulations involve complex digital replicas of items like cars and phones. To reproduce dynamic events like impacts, models solve differential equations over time. Model part interactions are assessed for deformations or failures. The calculations demand lots of memory and processing power. Due to model interconnectivity, compute nodes must communicate to exchange information about model elements’ effects.
AMD EPYC 9374F Benchmark
A prominent explicit simulation programme is Ansys LS-DYNA. Complex short-duration events in automobile, aircraft, construction, military, manufacturing, and biotechnology are simulated. In some LS-DYNA tests, the 2P 32-core AMD EPYC 9374F machine outperformed the Intel Xeon Scalable 8562Y+ system by ~1.50x.
Altair Radioss measures impact or crash structures. It benchmarks hardware using common usage issues. The 2P 32-core AMD EPYC 9374F system outperformed the Intel Xeon Scalable 8562Y+ workstation by ~1.25x in some Radioss tests.
AMD EPYC 9654 96-core processor
Molecular Dynamics
Newton’s equations are solved using molecular dynamics to analyze atoms and molecules. Molecular systems can be studied by analyzing material behavior, protein folding, and chemical processes. For hundreds to millions of particles, GROMACS simulates Newtonian motion using modular dynamics. A 2P 64-core AMD EPYC 9554 system provides ~1.24x the average performance of an Intel Xeon Scalable 8592+ system, outperforming competing datacenter CPUs. A top-of-the-stack 2P 96-core AMD EPYC 9654 machine outperforms Intel CPU systems by ~1.63x in GROMACS tests.
AMD EPYC 9554 64-core processor
Quantum chemistry
Quantum chemistry illuminates bonding and reactions through molecule structure, energetics, and reactivity. In quantum chemistry and solid-state physics, CP2K duplicates atomic systems. 2P 64-core AMD EPYC 9554 system outscored Intel Xeon Scalable 8592+ by ~1.27x in H2O-dft-ls test, whereas top-of-stack The AMD EPYC 9654 outperformed the Intel processing system by ~1.62 (Weather Forecast).
The conus2.5km test showed a 2P 64-core AMD EPYC 9554 system outperforming a 2P Intel Xeon Scalable 8592+ system by ~1.30x, and a top-of-the-line 2P 96-core AMD EPYC 9654 system outperforming the Intel system by ~1.50.
Conclusion
This blog revealed that 4th Gen AMD EPYC processors outperformed 5th Gen Intel Xeons in critical tasks. The comparisons above show that 4th Gen AMD EPYC CPUs perform well and reliably in industry-critical tasks. 4th Gen AMD EPYC CPUs excel in virtualized environments, molecular dynamics, and quantum chemistry simulations. Leading operating systems, hypervisors, and cloud services support AMD EPYC CPUs. The AMD Documentation Hub offers BIOS and OS tunings for different workloads to improve performance.
AMD delivers innovative technology beyond processors:
- AMD Instinct Accelerators: These accelerators assist scientists tackle complex challenges in various industries through exascale discoveries. They power scientific research, data analytics, machine learning, and other demanding applications with high-performance computing.
- AMD Pensando DPUs enable cloud, computing, networking, storage, and security software-defined programming. These technologies improve productivity, performance, and scalability over present infrastructures wherever data is. Programmable infrastructure pieces provide data centre agility.
- AMD FPGAs, hardware-adaptive SoCs, and ACAP processing platforms are adaptable and adaptive. These technologies boost endpoint, edge, and cloud infrastructure innovation.
- Developers can adapt and optimize solutions for specific applications and use cases, enhancing processing efficiency.
