Author: Nadezhda Ten, System and Software Optimization Engineer, Intel
Contributor: Contributor: Konstantin Nektyagaev, Software Enabling and Optimization Engineer (HPC), Intel
Intel® Xeon® processors have been widely adopted in high-performance computing (HPC) centers for years, powering many of the world’s top supercomputers and enterprise HPC clusters at Argonne National Laboratory in the United States, University of Cambridge in the United Kingdom, and Imperial College London in the United Kingdom. HPC's ability to process immense datasets and perform complex calculations enables advancements and efficiencies across many sectors, including scientific research on climate modeling and drug discovery, engineering product design simulations and testing, fraud detection, genomics, and even energy seismic data analysis.
Intel Xeon processors excel in HPC due to their combination of high core counts, double-precision floating-point performance using Intel® Advanced Vector Extensions 512 (Intel® AVX-512), and high memory bandwidth with advanced memory controllers that efficiently feed data-intensive scientific workloads. Additionally, HPC customers using Intel Xeon technology benefit from our mature software ecosystem with extensive compiler optimizations, broad application compatibility, and well-established development tools that have been refined over decades. This Intel oneAPI software tool is the key to unlocking the full hardware, delivering a critical performance differentiation that generic, open-source tools cannot match.
For example:
- The Intel® C/C++ (icx) and Fortran (ifx) Compilers are co-designed with new hardware, providing the only path to utilize features like Intel® APX. This doubles the general-purpose registers, drastically reducing memory loads and stores to accelerate general-purpose code far beyond what a generic GCC compiler can achieve.
- The Intel® oneMKL library provides a mature, drop-in replacement for BLAS/LAPACK that is highly optimized for AVX-512, often outperforming open-source alternatives like OpenBLAS by 4x or more on Intel hardware.
- Intel® oneDNN is the essential software bridge to the built-in Intel® AMX AI accelerators, providing optimized cblas_gemm kernels for FP16, BF16, and INT8 that are orders of magnitude faster for AI workloads.
- The Intel® MPI Library is a production-proven, tuned implementation that ensures lower latency and higher scalability in multi-node clusters, a mature advantage over generic Open MPI.
- Finally, tools like Intel® VTune Profiler provide deep, hardware-level insight into these new features, allowing developers to diagnose and fix complex bottlenecks at a microarchitectural level.
The latest Intel Xeon 6 processors continue this HPC leadership. With enhanced architecture delivering improved performance-per-watt, higher core densities, and support for Multiplexed Rank Dual Inline Memory Module (MRDIMM) technology for increased memory capacity and bandwidth, Intel Xeon offers the reliability and software compatibility that HPC centers depend on.
HPC systems can cost millions to build — that’s why HPC performance prediction helps prevent costly mistakes and enables informed decision-making in critical environments where both financial investment and scientific progress are at stake.
Comparative Analysis with AMD EPYC Processors
The Intel Xeon 6980P processor currently demonstrates the best SPEChpc 2021 performance in 2-socket systems without GPU accelerators in tiny and small test suites.(1) Developed by SPEC, SPEChpc 2021 is a set of application benchmark suites for measuring the performance of modern HPC systems. It serves as a tool for vendors to stress test their systems, for developers to compare programming models, and for end-users to make informed procurement decisions for HPC resources. The final SPEChpc scores represent the tested system's performance compared to a reference machine. A higher score indicates better performance, meaning the system completed the benchmarks in less time.
With approximately 33% fewer cores, Intel Xeon 6980P with 128 cores delivers 6.9% more peak performance than AMD EPYC 9965 with 192 cores in SPEChpc 2021 tiny test suites (see Figure 1). With the same core count of 128, Intel Xeon 6980P demonstrates approximately 10% improvement in peak performance over AMD EPYC 9755. The Intel Xeon 6980P also shows 9% improvement between peak and base performance (see Figure 1).
Figure 1: Intel Xeon 6980P with 128 cores delivers 6.9% more peak performance than AMD EPYC 9965 with 192 cores in SPEChpc 2021 (2,3,4)
In SPEChpc 2021 small test suites (see Figure 2), Intel Xeon 6980P shows 7% more base performance and approximately 15% more peak performance than AMD EPYC 9965.
Figure 2: Intel Xeon 6980P shows 7% more base performance and approximately 15% more peak performance than AMD EPYC 9965.(5,6)
SPEC CPU is a standardized benchmark suite that measures CPU performance. SPEC CPU floating point rate specifically measures how fast a processor can execute floating point operations across multiple parallel threads or processes, using real-world applications such as weather modeling, computational fluid dynamics, and scientific simulations. It serves as a standardized predictor of real-world HPC application performance. Higher scores indicate better performance.
Intel Xeon 6980P with 128 cores achieves the highest scores for 2-socket SPEC CPU floating point rate performance to date, outperforming both AMD EPYC 9965 with 192 cores and AMD EPYC 9755 with 128 cores in both base and peak scores. (see Figure 3).(7)
Figure 3: SPEC CPU 2017 floating point rate 2-socket performance.(8,9,10)
Intel Xeon Processors: Better Performance with Fewer Cores
These results reveal that having more cores doesn't necessarily translate to better performance. Intel Xeon 6980P demonstrates both efficiency and potential through its optimization headroom. This makes Intel Xeon 6980P the better choice for HPC applications, offering similar or better performance with fewer cores and better optimization potential.
Learn more about Developer Tools for Intel Xeon Processors.
Endnotes
- Based on SPEChpc 2021 (tiny, small) published results as of October 2025.
- Based on SPEChpc 2021 tiny suite published results on Intel Xeon 6980P test by Intel published September 2025.
- Based on SPEChpc 2021 tiny suite published results on AMD EPYC 9965 test by Supermicro. published October 2024.
- Based on SPEChpc 2021 tiny suite published results on AMD EPYC 9755 test by Supermicro, published October 2024.
- Based on SPEChpc 2021 small suite published results on Intel Xeon 6980P test by Intel, published September 2025.
- Based on SPEChpc 2021 small suite published results on AMD EPYC 9965 test by Supermicro, published October 2024.
- Based on SPEC CPU 2017 floating point rate published results as of October 2025.
- Based on SPEC CPU 2017 floating point rate published results on Intel Xeon 6980P test by xFusion, published April 2025.
- Based on SPEC CPU 2017 floating point rate published results on AMD EPYC 9965 test by ASUSTek Computer Inc. published April 2025.
- Based on SPEC CPU 2017 floating point rate published results on AMD EPYC 9755 test by xFusion, published October 2025.
Notices and Disclaimers
Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.
Performance varies by use, configuration, and other factors. Learn more on the Performance Index site.
Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or component can be absolutely secure.
Your costs and results may vary.
Intel technologies may require enabled hardware, software, or service activation.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.