Install Steam
login
|
language
简体中文 (Simplified Chinese)
繁體中文 (Traditional Chinese)
日本語 (Japanese)
한국어 (Korean)
ไทย (Thai)
Български (Bulgarian)
Čeština (Czech)
Dansk (Danish)
Deutsch (German)
Español - España (Spanish - Spain)
Español - Latinoamérica (Spanish - Latin America)
Ελληνικά (Greek)
Français (French)
Italiano (Italian)
Bahasa Indonesia (Indonesian)
Magyar (Hungarian)
Nederlands (Dutch)
Norsk (Norwegian)
Polski (Polish)
Português (Portuguese - Portugal)
Português - Brasil (Portuguese - Brazil)
Română (Romanian)
Русский (Russian)
Suomi (Finnish)
Svenska (Swedish)
Türkçe (Turkish)
Tiếng Việt (Vietnamese)
Українська (Ukrainian)
Report a translation problem
There are very few benchmarks available online. Overal it seems like reads are about the same or slightly slower, writes are a lot slower as expected.
But there is much more to a filesystem than just raw read and write performance. ReFS is significantly more resiliant to corruption than NTFS.
This needs more testing, this FS seems to have all kinds of fancy features we can play with.
Data dedup is a bug feature also, it can save storage space in some scenarios.
Nothing, that I myself can report on.
Tested it on a Ryzen true 16-core, doing some heavy Terabyte robocopy file shovelling. Difference is probably in the sub-ms area?
This seems to be a very 'academic', if not obscure subject, unless you are working in Data Centers at Google, Amazon, Netflix, etc. or are joining the "Storage Networking Industry Association" https://www.snia.org/
My guess would be: does not matter in a 'game' environment. Dev or player.
In the end - as with all gamedev things - (r/w) cache is king? As is more and/or faster lanes & bandwidth (m2, pcie5, pcie6 ...)?
In itself, this is quite an interesting topic: how to ever improve performance AND reliability, but without the cost of 'data failure'. Contradictory. Doing RAID type of things ... on a single drive, via drivers and chip controllers.
Reminds me of the decades old UNIX zfs vs xfs discussions, long before btrfs was even an idea. Once again, Microsoft trying to copy UNIX tech and stay relevant. Not a bad idea, though. Slow to adopt, due to their equally 'resilient' customer base?
You're starting out with a flawed premise. ReFS is meant to expand the feature set for enterprise and server storage needs, not as a wholesale replacement for NTFS as a general-purpose filesystem. It was initially built from some portions of the NTFS code when it shipped as a trial feature on Windows Server 2012, by the time it was added to Windows Server 2016 it was pretty much a wholly different codebase.
https://learn.microsoft.com/en-us/windows-server/storage/refs/refs-overview
While long-term they hope to be able to expand its feature set to service the features of NTFS; that hasn't been the immediate intention (for the past decade of its development). It isn't a bootable filesystem currently, doesn't support compression or encryption, doesn't support extended file attributes or disk quotas; so it isn't being targeted at the consumer general purpose systems.
Back in March for the Windows Server Summit they did discuss adding boot support so ReFS could be used as a bootable volume for their new "Confidential VMs" as part of Windows Server 2025. However, I still don't see it replacing NTFS as the default filesystem for general purpose use on consumer Windows any time soon. Most of the focus is on supporting features to expand the functionality for enterprise features and virtualization such as Azure VMs, Azure local, Hyper-V, etc.