Booting Windows 24H2 with ReFS 3.14 as Native FS

[Mike]

The question comes about if it is currently possible and stable to install Windows 11 Pro (or Enterprise) on ReFS. Apparently this appears possible based on a Guru3D post: testing instaling windows 11 24h2 build 26100 on reFS

I currently have converted my D drive to ReFS by copying all files to C and running the ViveTool:

vivetool /enable /id:42866187

Then going into "For Developers" / "Developer Settings":

C:\Windows\System32>fsutil fsinfo refsinfo d:
REFS Volume Serial Number :                0xc81430fd1430f054
REFS Volume Version :                      3.14
REFS Driver Maximum Supported Version :    3.14
Number Sectors :                           0x00000000e8e00000
Total Clusters :                           0x000000001d1c0000
Free Clusters  :                           0x000000001673fbdc

Am I the only one trying to do this for C?

The reason I ask is due to the block cloning/storage enhancements in ReFS:

However I do not want to reinstall Windows (of course). I am close to seeing if this is possible on a VM.

It should be noted that Windows 11 24H2 appears to include an update of ReFS to 3.14 which may support native booting of Windows 11 (although I cannot fully verify if it was supported before this).

 

[ChatGPT]

Booting Windows 11 Pro or Enterprise on ReFS (Resilient File System) natively is a highly experimental and advanced topic. While it appears Microsoft is making strides with ReFS in newer builds, and developers have shown various methods to enable ReFS, it is crucial to understand the current limitations and potential pitfalls. Key Points:

Compatibility and Stability:​

• Official Support: ReFS is mainly designed for data storage rather than as a boot volume. Windows traditionally boots from a volume formatted with NTFS.
• Experimental Features: By using tools like ViveTool, you can enable experimental features, but these may not be fully stable or supported for a production environment.

Steps Taken:​

From your description, you've successfully converted your D: drive to ReFS:

C:\Windows\System32>fsutil fsinfo refsinfo d: REFS Volume Serial Number : 0xc81430fd1430f054 REFS Volume Version : 3.14 REFS Driver Maximum Supported Version : 3.14 Number Sectors : 0x00000000e8e00000 Total Clusters : 0x000000001d1c0000 Free Clusters : 0x000000001673fbdc

And controlled the drive through "For Developers" settings, emphasizing block cloning/storage enhancements.

Performance:​

The comparative table you've provided indicates significant performance improvements with ReFS using block cloning:

File(s) Copied	NTFS	Dev Drive w/ Block Cloning	% Improvement
10GB file	7s 964ms	641ms	92%
1GB file	681ms	38ms	94%
1MB file	11ms	9ms	18%
18GB folder (5815 files)	30s 867ms	6s 306ms	80%

Next Steps and Considerations:​

• Testing in VM:
• It is prudent to experiment with installing Windows 11 on ReFS within a virtual machine (VM). This provides a safer environment free from the risk of data loss or system failure on your primary system. [*Potential Roadblocks:
• Bootloader compatibility: The Windows Boot Manager (bootloader) traditionally expects an NTFS formatted system partition.
• System Recovery Tools: Windows recovery environments and certain system tools may not support ReFS natively. [*Backup and Recovery:
• Ensure you have a robust backup and recovery plan before experimenting with booting from ReFS. This is especially important if ReFS is not fully supported for your boot drive.
• Community and Official Documentation:
• Engage with the developer community and monitor official Microsoft documentation and forums for updates on ReFS support and relevant experiments, similar to the one you referenced (https://forums.guru3d.com/threads/testing-instaling-windows-11-24h2-build-26100-on-refs.451918/).

Conclusion:​

While the feasibility of booting Windows 11 on ReFS shows promise, it remains a complex and experimental setup. Utilizing a VM for testing phases and ensuring a robust backup strategy is crucial. Continued monitoring of community findings and official updates will provide insights into the evolving support for ReFS as a native file system for boot volumes. Should you decide to proceed further, documenting your process and findings can greatly benefit the community exploring similar capabilities.

 

[Mike]

@ChatGPT Based on some tests in CrystalDiskMark:

• Sequential Read (1MiB, Q=8, T=1):
  • NTFS is 1.81% faster than ReFS.
• Random Read (4KiB, Q=32, T=16):
  • NTFS is 2.80% faster than ReFS.
• Sequential Write (1MiB, Q=8, T=1):
  • ReFS is 0.80% faster than NTFS.
• Random Write (4KiB, Q=32, T=16):
  • ReFS is 13.28% faster than NTFS.

This would require hundreds of additional tests though to verify and may not be indicative of real world speeds.

 

[Mike]

@ChatGPT I was able to accomplish this using Fn+F10 at Windows setup when doing a clean install:

format c: /fs:refs /q

Most notably, the new BitLocker encryption on install will fail. OneDrive will also fail. But this will update the OS filesystem:

Clearly, not supported:

 

[Mike]

@ChatGPT Analyze this image:

What performance gains does this show for ReFS in terms of copying large files and folders on the same drive/partition? What percentage difference is this between NTFS (shown left). Where can I expect to see other improvements in a single drive environment on a normal user use case scenario on a laptop for instance? Please go into great details and elaborate on this further.

 

[ChatGPT]

It looks like you've conducted some performance tests using CrystalDiskMark comparing the file systems NTFS and ReFS. Here’s a summary of your findings:

Performance Tests Summary:​

1. Sequential Read (1MiB, Q=8, T=1):
   • NTFS is 1.81% faster than ReFS.
   []Random Read (4KiB, Q=32, T=16):
   • NTFS is 2.80% faster than ReFS.
   []Sequential Write (1MiB, Q=8, T=1):
   • ReFS is 0.80% faster than NTFS.
2. Random Write (4KiB, Q=32, T=16):
   • ReFS is 13.28% faster than NTFS.

     Considerations:​

   • Further Testing: You noted that these results would require hundreds of additional tests to verify and may not fully reflect real-world performance. It's important to consider various factors like workload, drive conditions, and specific use cases when evaluating file system performance. This analysis would be quite useful for users deciding between NTFS and ReFS based on their specific needs. If you'd like to delve deeper into your findings or need additional insights, feel free to ask!