AMD EPYC 9004/9005 on Windows Server: Installation & Support Guide

Table of Contents

This comprehensive guide delves into the essential support statements and detailed installation procedures for running the Windows Server operating system on systems powered by AMD EPYC 9004 and AMD EPYC 9005 series processors. These cutting-edge processors are designed with high core counts, often exceeding 64 cores per CPU socket, which introduces unique considerations for operating system deployment and ongoing support. Understanding these nuances is crucial for ensuring stable and optimized server performance.

Windows Server Compatibility and Evolution¶

The robust AMD EPYC 9004 and AMD EPYC 9005 series processors are fully supported across a range of modern Windows Server releases. This compatibility ensures that organizations can leverage the exceptional performance and efficiency of these processors within their existing or planned Microsoft server environments. The supported versions demonstrate Microsoft’s commitment to enabling next-generation hardware.

The following Windows Server releases offer official support for AMD EPYC 9004 and AMD EPYC 9005 series processors:

• Windows Server 2019: This long-term servicing channel release remains a popular choice for many enterprises, providing a stable and feature-rich platform. Specific build requirements are necessary for high-core count systems, as detailed later in this guide. Its continued support highlights the sustained relevance of this operating system.
• Windows Server 2022: As a more recent release, Windows Server 2022 offers enhanced security features, improved hybrid cloud capabilities, and better performance optimizations for modern hardware. It builds upon its predecessors to deliver a more efficient and resilient server environment. This version is particularly well-suited for demanding workloads.
• Windows Server 2025: The upcoming iteration of Windows Server signifies the continuous evolution of Microsoft’s server operating system to meet the demands of future data center technologies. Its inclusion in the support list underscores forward compatibility and ongoing advancements in hardware integration. Planning for this version can help future-proof infrastructure.

Understanding AMD EPYC Processor Ordering Part Numbers (OPN)¶

AMD EPYC 9004 series processors represent a broad portfolio of server CPUs, each designed to meet diverse workload requirements ranging from general-purpose computing to high-performance computing (HPC) and virtualization. Similarly, the AMD EPYC 9005 series further expands upon this foundation, offering even more specialized capabilities and higher performance metrics. Identifying the specific processor model within these series is essential for proper system configuration and support.

Each AMD EPYC processor is uniquely identified by an Ordering Part Number (OPN), which provides detailed information about its specifications, core count, clock speed, and thermal design power (TDP). System administrators and IT professionals must consult the official AMD product documentation to accurately determine the characteristics of their installed processors. This ensures that the chosen Windows Server configuration and any subsequent optimizations are precisely aligned with the hardware’s capabilities. Understanding the OPN is a fundamental step in server deployment and troubleshooting.

Installing Windows Server on AMD EPYC 9004 or 9005 Series Processors¶

When deploying Windows Server on systems equipped with AMD EPYC 9004 or 9005 series processors, it is always recommended to use the absolute latest available installation media image. This practice helps to ensure that the initial operating system installation includes the most recent drivers, critical updates, and performance optimizations. After the initial installation is successfully completed, a mandatory step involves updating the system to the very latest Windows Update release. This ensures all current patches and features are applied.

For systems that feature AMD EPYC 9004 or 9005 processors with configurations exceeding 64 cores, specific Windows Server operating system media images are required for a successful initial installation. Older installation media may not possess the necessary kernel enhancements or hardware abstraction layer (HAL) support to properly recognize and utilize such a high core count during the setup phase. Adhering to these minimum build requirements prevents common installation failures.

The minimum required OS media build versions are:

• Windows Server 2019: Build 17763.3532 (October 2022) or any later release.
• Windows Server 2022: Build 20348.859 (July 2022) or any later release.

If you are working with an earlier Windows Server OS media release that predates the aforementioned minimum builds, specific pre-installation steps are necessary to facilitate a successful deployment. These methods circumvent the initial core count limitations of older media, allowing the installation to proceed before a full system update can be performed. It is critical to follow one of these options diligently.

Option 1: Adjusting Core Complex Die (CCD) Settings in BIOS¶

This approach involves temporarily modifying the processor’s configuration within the system’s BIOS/UEFI settings to reduce the visible core count during the initial OS installation phase. AMD EPYC processors are built using Core Complex Dies (CCDs), which are modular units containing multiple CPU cores and shared caches. By limiting the number of enabled CCDs, the system presents a lower effective core count to the installation media.

Follow these steps:

1. Configure BIOS Setting: Access the system’s BIOS or UEFI settings during boot-up. Locate the setting related to “Number of CCDs enabled per processor” or a similar option. Set this value to a maximum of 8 CCDs per processor, which will significantly reduce the reported logical processor count.
2. Perform OS Installation: With the adjusted BIOS setting, proceed with the standard Windows Server operating system installation using your existing media. The installer should now recognize a compatible core count and complete the setup without issues related to excessive cores. The system will operate with fewer cores temporarily.
3. Update Operating System: Once the initial OS installation is successfully completed and the system boots into Windows Server, immediately perform a comprehensive Windows Update. This process will download and apply the latest Cumulative Updates, which include the necessary kernel and HAL improvements to fully support high core count processors. This step is crucial for long-term stability and performance.
4. Restore BIOS Setting: After the system has been fully updated and is running the latest Cumulative Update release, reboot the server and re-enter the BIOS/UEFI settings. Restore the “Number of CCDs enabled per processor” setting to its original, full value (e.g., the maximum supported by your processor model). This will enable all available cores for optimal performance.

Important Note: The availability and precise naming of BIOS settings, such as “Number of CCDs enabled per processor,” can vary significantly between different server manufacturers (OEMs). Always consult the specific BIOS documentation provided by your system’s OEM for accurate guidance on locating and modifying these settings. Misconfigurations in the BIOS can lead to system instability.

Option 2: Updating OS Installation Media Images using DISM¶

This advanced method involves integrating the necessary Windows Updates directly into your operating system installation media before beginning the deployment. This ensures that the installer itself is capable of supporting high core count processors from the outset, eliminating the need for BIOS adjustments. This process typically utilizes the Deployment Image Servicing and Management (DISM) tool.

Follow these steps:

1. Download Required Packages: Obtain the latest Windows Server Cumulative Update packages and any prerequisite packages. These are typically available from official Microsoft update catalogs. Ensure you download the correct architecture (e.g., x64) and version for your target Windows Server release. These packages contain the critical fixes needed.
2. Prepare Installation Media: Mount your Windows Server ISO image or copy its contents to a local directory. You will specifically need to access the boot.wim (Windows Preinstallation Environment) and install.wim (actual OS image) files within the sources folder. These files contain the core components of the operating system.
3. Integrate Updates with DISM: Use the DISM command-line tool to inject the downloaded update packages into both the boot.wim and install.wim images. This process involves mounting the WIM images, adding the prerequisite packages first, followed by the Latest Cumulative Update package, and then unmounting and saving the changes. This step makes the installation media “aware” of high core counts.
   • Conceptual DISM Workflow:
     • Mount boot.wim and install.wim to temporary directories.
     • Add prerequisite .msu packages to the mounted images.
     • Add the latest Cumulative Update .msu package.
     • Unmount the images, committing the changes.
       This updated media can then be used for installation, which will proceed smoothly without BIOS modifications for core count.

The following Mermaid diagram illustrates the conceptual flow for installing Windows Server on high-core-count AMD EPYC systems:

```mermaid
graph TD
A[Start Windows Server Installation] → B{System has > 64 Cores per Socket?};
B – Yes → C{Is OS Media Build Sufficient?};
C – Yes (e.g., WS 2019 Build 17763.3532+, WS 2022 Build 20348.859+) → D[Proceed with Standard Installation];
C – No (Older OS Media) → E{Choose Remediation Option};

E -- Option 1: BIOS CCD Adjustment --> F[Reduce CCDs to Max 8 in BIOS];
F --> D;

E -- Option 2: Update OS Media (DISM) --> G[Download Latest Cumulative Update Packages];
G --> H[Integrate Updates into boot.wim and install.wim using DISM];
H --> D;

D --> I[Initial OS Installation Complete];
I --> J[Run Comprehensive Windows Update];

J --> K{If Option 1 Used: Restore CCDs?};
K -- Yes --> L[Restore CCDs to Original Value in BIOS];
K -- No --> M[System Fully Configured and Ready];

L --> M;

```

Known Issues and Limitations¶

While Windows Server offers robust support for AMD EPYC 9004 and 9005 series processors, certain configurations, particularly those with exceptionally high logical processor counts, may encounter specific behaviors or limitations. These are typically related to the scaling capabilities of the operating system’s kernel and Hyper-V virtualization components in highly dense environments. Understanding these issues is vital for effective system management.

Windows Server 2019 with Hyper-V Minroot Configuration¶

On systems running Windows Server 2019 with a significant number of logical processors and the Hyper-V virtualization feature enabled, the operating system defaults to a specialized configuration known as “Hyper-V Minroot.” This configuration optimizes the host partition for its role as a hypervisor, but it can introduce some observable behaviors when dealing with extremely high logical processor counts (typically exceeding 320 logical processors on AMD EPYC 9004 or 9005 series-based systems). These behaviors are inherent to the design of the Minroot kernel.

The following issues and limitations may be observed:

• Task Manager CPU Utilization: The Task Manager within the root partition may not accurately display the CPU utilization attributed to virtual machines’ workloads. This can lead to an underestimation of resource consumption by the hypervisor and its guests. While this is primarily a reporting anomaly, it can complicate performance monitoring.
• Collaborative Processor Performance Control (CPPC): The operating system may not effectively utilize Collaborative Processor Performance Control (CPPC) for processor power management, even if CPPC is configured to be enabled in the system’s BIOS. CPPC is a modern power management interface that allows the OS and processor to collaboratively optimize performance and power efficiency. Its non-utilization could lead to less optimal power consumption or performance scaling.
• Root Partition Logical Processor Utilization: The root partition itself might not fully utilize all of the maximum 320 logical processors available when operating in a Minroot configuration. This implies that some processing threads might remain idle or underutilized, potentially limiting the overall performance of host-level operations.

Furthermore, an additional limitation applies specifically to AMD EPYC 9005 series-based systems when the total number of logical processors exceeds 512. In such extreme configurations, the root partition will recognize and utilize a maximum of only 512 logical processors. Any logical processors beyond this count will not be enumerated or leveraged by the host operating system. This is a hard architectural limit for that specific configuration.

These limitations on Windows Server 2019 in a Minroot configuration are important considerations for very large-scale virtualization deployments. They highlight the need for careful planning and potentially different monitoring strategies. While the system will still function, optimal resource allocation might require additional tuning.

Windows Server 2022 with Hyper-V Logical Processor Enumeration¶

On systems running Windows Server 2022 with the Hyper-V virtualization feature enabled, a specific limitation was identified concerning AMD EPYC 9005 series-based systems that feature a very high number of logical processors per Non-Uniform Memory Access (NUMA) node. If a NUMA node contains more than 256 logical processors, the root partition would only enumerate and utilize the first 256 logical processors within that specific NUMA node. This could restrict the effective utilization of very dense NUMA configurations.

This limitation was subsequently addressed and resolved in a specific Windows Update: KB5044281 (OS Build 20348.2762). Ensuring that Windows Server 2022 installations are updated to this build or a later version is critical for systems with such high logical processor counts per NUMA node. Applying this update ensures that the root partition can properly enumerate and leverage all available logical processors, thereby maximizing system performance and resource allocation.

Windows Server 2019 Recovery Environment (WinRE) Issues¶

Attempting to boot into the Windows Server 2019 Recovery Environment (WinRE) on systems with greater than 64 cores per socket may result in a critical system error, commonly manifesting as a blue screen with the error code 0x5C HAL_INITIALIZATION_FAILED. This error indicates that the Hardware Abstraction Layer (HAL) within the WinRE image is unable to properly initialize or interact with the high core count processor configuration during the boot process. The original WinRE image, being an older environment, was not designed to anticipate such extreme core densities.

To enable proper support for these high core count configurations within the Windows Server 2019 Recovery Environment, it is essential to update the WinRE image itself. This involves applying the latest cumulative update for Windows Server 2019 directly to the WinRE image. Updating the WinRE image ensures that it contains the necessary HAL and kernel improvements to recognize and operate correctly on systems with more than 64 cores per socket. This preventative measure is vital for maintaining system recoverability.

Best Practices for High Core Count AMD EPYC Systems¶

Optimizing the performance and stability of Windows Server on high core count AMD EPYC 9004/9005 systems extends beyond initial installation. Implementing a set of best practices can significantly enhance operational efficiency and prevent potential issues. These practices cover areas from firmware to ongoing system maintenance.

BIOS/UEFI Configuration¶

Always ensure your system’s BIOS/UEFI firmware is updated to the latest version provided by your server OEM. Firmware updates often include critical microcode updates for the processor, performance enhancements, and bug fixes that directly impact stability and hardware compatibility. Review power management profiles within the BIOS. For demanding workloads, setting the power profile to “High Performance” or “OS Controlled” is often recommended, allowing the operating system to dynamically manage processor frequencies. Carefully configure NUMA (Non-Uniform Memory Access) settings based on your workload’s memory access patterns. Proper NUMA configuration can significantly reduce memory latency and improve application performance by ensuring processes primarily access local memory.

Driver Management¶

While Windows Server typically includes inbox drivers for most hardware, it is paramount to install the latest chipset drivers and any specific device drivers provided by AMD or your server OEM. These drivers are optimized for the specific architecture of EPYC processors and system components, ensuring maximum performance and stability. Regularly check for updated drivers, especially after major Windows Server updates or hardware changes. Using outdated or generic drivers can lead to performance bottlenecks or intermittent issues.

Monitoring and Performance Tuning¶

High core count systems generate a vast amount of performance data. Utilize Windows Server’s built-in Performance Monitor, Resource Monitor, and Task Manager to gain insights into CPU utilization, memory consumption, and I/O operations. For virtualized environments with Hyper-V, pay close attention to the host’s logical processor utilization and how it distributes workloads across virtual machines. Consider using specialized tools for in-depth analysis, especially for complex or highly parallel applications. Proper monitoring helps identify bottlenecks and inform tuning decisions.

Patch Management Strategy¶

Maintaining a rigorous patch management strategy is non-negotiable for high core count systems. Regular application of Windows Updates, especially Cumulative Updates, is crucial. These updates frequently include performance optimizations, security fixes, and critical bug resolutions related to processor support and scalability. Automate or schedule updates during off-peak hours to minimize disruption. Verify system stability after updates, especially in production environments, to ensure no regressions occur. Staying current with patches directly impacts system security and operational integrity.

Future Outlook for High Core Count Computing¶

The relentless progression of processor technology towards higher core counts, exemplified by AMD EPYC 9004 and 9005 series, continues to reshape the landscape of data centers and cloud computing. Both operating system vendors like Microsoft and chip manufacturers like AMD are in a continuous cycle of innovation, striving to ensure seamless integration and optimal performance. This ongoing collaboration addresses current limitations and paves the way for even more powerful and efficient server platforms. As applications become more parallelized, the importance of these architectural advancements and OS-level optimizations will only grow.

---

We hope this in-depth guide has provided valuable insights into the installation and support of Windows Server on AMD EPYC 9004 and 9005 series processors. Do you have experience deploying Windows Server on high core count systems? Share your tips, challenges, or specific optimizations that have worked well for your environment in the comments section below. Your feedback helps the entire community build more robust and efficient server infrastructures.