Data Center
Participate in insightful discussions regarding Data Center topics
98 Discussions

Predict Faster. Predict Smarter. Real-Time Weather Simulation Powered by Intel® Xeon® Processors

JimmyLeon
Employee
0 0 1,193

Author:

Jimmy Leon, Engineering Manager, Intel

Contributors: 

Andrey Ovsyannikov, Software Enabling and Optimization Engineer, Intel

Joseph Robichaux, System and Software Optimization Engineer, Intel

David Trusov, Technical Marketing Engineer, TeamPeople

The demand for more accurate and detailed weather forecasting is surging. Whether it is helping emergency responders anticipate a hurricane’s path or assisting farmers in optimizing irrigation, the accuracy of these simulations depends on raw computational power. Modern numerical weather prediction (NWP) now leverages thousands of cores in a single run, and Intel® Xeon® processors are at the heart of this evolution.

The Intel Xeon processor delivers clear, measurable advantages for NWP workloads, accelerating both simulation throughput and data handling, key factors for producing accurate forecasts on time. Across real‑world NWP benchmarks, the Intel Xeon 6980P demonstrates up to 1.15x faster performance on the Weather Research and Forecasting (WRF) model and up to 1.33x higher performance on the Model for Prediction Across Scales (MPAS) compared to AMD EPYC 9755 processors, underscoring its strength as a high‑efficiency compute platform for modern weather modeling.

Why Intel Xeon Processors for Weather Modeling?

Weather simulation is notoriously "memory-bandwidth bound," meaning the speed at which data moves between the processor and memory often limits the system's performance. Weather modeling applications exhibit a high byte-to-floating-point (BF) ratio, indicating they are heavily dependent on memory bandwidth rather than just raw processor speed or floating-point operations per second (FLOPS). Recent advancements in the Intel Xeon processor family help address these challenges:

  • Memory bandwidth: The latest Intel Xeon 6 processors feature Multiplexed Rank Dual Inline Memory Module (MRDIMM) technology, which can significantly boost the performance of climate models like WRF and MPAS. Intel Xeon 6900P processor-based platforms can deliver up to 1.6 terabytes per second (Tb/s) of theoretical memory bandwidth on a dual-socket platform.(1)
  • Massive parallelism: Models like the WRF system are designed to scale across thousands of cores. For weather simulations, Intel Xeon 6900P processors offer a balanced combination of memory bandwidth and performance cores.
  • Intel® oneAPI HPC Toolkit with Intel® MPI Library: With its optimized compilers, libraries, and tools, the Intel oneAPI HPC Toolkit enables fast communication for multi-node, large-scale simulations. This helps meteorologists to process massive datasets more efficiently, resulting in better predictive accuracy for severe weather events.

Numerical Weather Simulation Competitive Performance

Test case using WRF: The WRF model is an open-source mesoscale numerical weather prediction system designed for both atmospheric research and operational forecasting. It is utilized worldwide by government, academic, and private groups for operational forecasting, research, and specialized applications such as real-time air quality forecasting, hurricane prediction, and climate simulation.

The CONUS 2.5-kilometer (km) benchmark is a high-resolution meteorological modeling test case used to evaluate WRF model performance across the continental United States. It simulates a 6-hour forecast (often focusing on the last 3 hours) at a 2.5 km resolution across the U.S. The Intel Xeon 6980P 128-core processor delivered up to 1.15x improved performance over the 128-core AMD EPYC 9755. The 192-core AMD EPYC 9965 achieved nearly the same performance as the 128-core AMD EPYC 9755.

Picture1.png

Figure 1: Intel Xeon 6980P leads the AMD EPYC 9755 and AMD EPYC 9965 in WRF performance by up to 1.15x. 

 

Test case using MPAS: As a collaborative project to develop atmosphere, ocean, and other Earth-system simulation components, MPAS is used in climate, regional climate, and weather studies.

Focused on the atmosphere, the MPAS-A system uses unstructured Voronoi meshes to support variable-resolution simulations, enabling detailed regional-scale forecasting within a global context. MPAS-A is often used in studies on tropical meteorology, extreme events, and long-term climate simulations. The Intel Xeon 6980P 128-core processor delivered up to 1.33x better MPAS performance than the 128-core AMD EPYC 9755.

Picture2.png

Figure 2: The Intel Xeon 6980P leads the AMD EPYC 9755 in performance by up to 1.33x.

 

The following video compares the Intel Xeon 6980P processor to the AMD EPYC 9755 when running MPAS-A test cases.

 

Intel Xeon Processors: High‑Efficiency Processor for Modern Weather Modeling

The Intel Xeon 6980P processor offers significant performance improvements for NWP by enabling faster simulation speeds and higher memory bandwidth, which are crucial for timely, accurate forecasting.

To help achieve the best performance, use this tuning guide to learn how to tune your Intel Xeon processor high-performance computing (HPC) platform with Intel tools. The guide includes general BIOS and Linux configuration guidelines for all HPC applications as well as build and run instructions (recipes) for a selected set of HPC applications and benchmarks, including WRF and MPAS-A.

Product and Performance Information

WRF

Test by Intel as of December 2024 for Intel Xeon 6980P, January 2025 for AMD EPYC 9755, and October 2025 for AMD EPYC 9965. Your results may vary. Intel technologies may require enabled hardware, software, or service activation.

Hardware Configurations

Intel Xeon 6980P: 1-node, 2x Intel(R) Xeon(R) 6980P 128-Core Processor, 128 cores, 500W TDP, HT On, Turbo On, Total Memory 1536GB (24x64GB DDR5 8800 MT/s [8800 MT/s]), BIOS BHSDCRB1.IPC.3544.P15.2410232346, microcode 0x1000341, 2x Ethernet Controller X710 for 10GBASE-T, 1x 3.5T SAMSUNG MZWLJ3T8HBLS-00007, Ubuntu 22.04.5 LTS, 6.5.0-21-generic. Using physical cores only. Test by Intel as of December 2024.

AMD EPYC 9755: 1-node, 2x AMD EPYC 9755 128-Core Processor, 128 cores, 500W TDP, SMT On, Boost On, Total Memory 1536GB (24x64GB DDR5 6400 MT/s [6000 MT/s]), BIOS 1.1, microcode 0xb002116, 2x Ethernet Controller X710 for 10GBASE-T, 1x 3.5T Micron_7450_MTFDKCB3T8TFR, Ubuntu 24.04.1 LTS, 6.8.0-48-generic. Using physical cores only. Test by Intel as of January 2025.

AMD EPYC 9965: 1-node, 2x AMD EPYC 9965 192-Core Processor, 192 cores, 500W TDP, SMT On, Boost On, Total Memory 1536GB (24x64GB DDR5 6400 MT/s [6000 MT/s]), BIOS 1.1, microcode 0xb002116, 2x Ethernet Controller X710 for 10GBASE-T, 1x 3.5T Micron_7450_MTFDKCB3T8TFR, Ubuntu 24.04.1 LTS, 6.8.0-48-generic. Using physical cores only. Test by Intel as of October 2025.

Software Configuration

WRF (CONUS-2.5km): Intel: App Version: v4.5.2, conus2.5km
AMD: App Version: v4.5.2, conus2.5km.
Intel® Fortran Compiler version 2025.0 release
Intel® MPI Library version 2021.14 release

 

MPAS

Test by Intel as of March 2025. Your results may vary. Intel technologies may require enabled hardware, software, or service activation.

Hardware Configurations

Intel Xeon 6980P: 1-node, 2x Intel(R) Xeon(R) 6980P 128-Core Processor, 128 cores, 500W TDP, HT On, Turbo On, Total Memory 3072GB (24x128GB DDR5 8800 MT/s [8800 MT/s]), BIOS BHSDCRB1.IPC.3544.P15.2410232346, microcode 0x1000341, 1x I210 Gigabit Network Connection, 1x 3.6T INTEL SSDPE2KX040T8, Ubuntu 24.04.2 LTS, 6.8.0-47-generic. Using physical cores only. Test by Intel as of March 2025.

AMD EPYC 9755: 1-node, 2x AMD EPYC 9755 128-Core Processor, 128 cores, 500W TDP, SMT on, Turbo on, Total Memory 1536GB (24x64GB DDR5 6400 MT/s [6000 MT/s]), BIOS 1.1, microcode 0xb002116, 2x Ethernet Controller X710 for 10GBASE-T, 2x Ethernet Controller E810-C for QSFP, 1x 3.5T Micron_7450_MTFDKCB3T8TFR, Ubuntu 24.04.1 LTS, 6.8.0-47-generic. Using physical cores only. Test by Intel as of March 2025.

Software Configuration

Intel® MPI Library 2021 Update 15;1 Intel® oneAPI HPC Toolkit, MPAS Version 7.3 for 720 Timesteps (3 forecast days).

 

(1) Data rate for 8800 MT/s, using a bus width of 64 bit = 8 bytes. BW = Data Rate (MT/s) * Bus Width (bit) /(Bits per Byte (8) * 1000) = 8800 MTs * 64 bits/(8 bit/byte * 1000) = 70.4 GB/s per DIMMc. For 24 channels = 1689.6 TB/s

 

Notices and Disclaimers

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.