Early Windows 11 25H2 benchmarks confirm the update provides no performance improvements over 24H2
Makes sense, considering 25H2 is based on the same source code as 24H2.

Windows 11 25H2 is not out yet, but that hasn't stopped media outlets from benchmarking the upcoming Windows 11 version. Phoronix tested Windows 11 25H2 against Canonical's upcoming Ubuntu 25.10, Ubuntu 24.0.3 LTS, and Windows 11 24H2 in a head-to-head benchmarking battle to see if 25H2 delivers any new performance improvements. Sadly, though, Phoronix's testing revealed that 25H2 was unable to outperform 24H2, even technically losing to its predecessor in many tests.
The four operating systems were tested using a Ryzen 9 9950X paired with 32GB of DDR5 memory. Phoronix benchmarked a variety of applications (41 benchmarks in total), including LuxCoreRender, Embree, Intel Open Image Denoise, OSPRay, and IndigoBench.
If you are familiar with previous Phoronix Windows vs Linux testing, it should come as no surprise that Ubuntu came out on top. Across Phoronix's 41 benchmark geomean, both versions of Ubuntu managed to outperform Windows 11 25H2 by around 15% respectively. Looking at the Windows operating systems exclusively, 25H2 provided precisely 0% more performance compared to 24H2 across Phoronix's entire benchmarking suite, on average.
Phoronix's individual numbers for each benchmark further elaborate on 25H2's underwhelming performance behavior. Many of the individual benchmarks show 25H2 and 24H2 competing for third and fourth place, with 24H2 trading blows with 25H2. For example, in LuxCoreRender, Windows 11 24H2 was 2% more performant than 25H2. But in ASTC Encoder 5.0, 25H2 was almost 2% faster than 24H2.
Windows 11 25H2's underwhelming benchmarking performance is not surprising to see. Microsoft's upcoming version of Windows 11 is based on the same servicing branch as 24H2, meaning that both versions are largely identical under the hood. In fact, Windows 11 24H2 already has many of 25H2's features incorporated in a disabled state. 25H2 merely turns these disabled features on and guarantees their availability, while 24H2 users will have to wait for these features to be turned on incrementally over time. This is a far cry from 24H2, which was a major overhaul over 23H2, including major parts of the OS being rewritten in Rust.
As a result, 25H2 is also one of the smallest "major" updates Microsoft has created for Windows 11 thus far. 25H2 includes just a handful of new features, while also removing some existing features, such as PowerShell 2.0 and the Windows Management Instrumentation command-line. 25H2 will also be delivered as an enablement package, requiring just one restart to update from 24H2 to 25H2 (just like a standard cumulative update).
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.

Aaron Klotz is a contributing writer for Tom’s Hardware, covering news related to computer hardware such as CPUs, and graphics cards.
-
ezst036 Microsoft is going to need to remove the advertisements and all of the logic required to snoop on user data and make advertising possible.Reply
Recall can't be good for CPU cycles either. -
RodroX Lets not forget to look on the bright side of life, like the movie said...Reply
It doesn't bring any performance improvemente, but it also doesn't bring any performance drops (for now).
And yes Microsoft, remove all the crap you are forcing to users if you want a bigger audience for your horrible Win11 -
EzzyB
Makes me wonder just how these tests are run. I've gone to using Enterprise versions of Windows that allow you to yank most of that stuff out.ezst036 said:Microsoft is going to need to remove the advertisements and all of the logic required to snoop on user data and make advertising possible.
Recall can't be good for CPU cycles either.
I wouldn't say it's unfair to run the benchmarks against out-of-the-box Windows, but it may not apply to all of us. -
Blacksad999
Right. Most updates are bug fixes or implementing new things. Not "does it peform faster?"JRStern said:Did someone expect it would? -
rluker5
I can only hope it is properly made to run as a background task and will be relegated to e cores/low power e cores.ezst036 said:
Recall can't be good for CPU cycles either. -
_Shatta_AD_ Yes Microsoft will spend more dev time and resources on how to embed more advertisement codes and know how to facilitate such purpose rather than streamlining and improving performance. Their company model is operating more like some influencer on YT maximizing viewership and followers by any means necessary while producing cheap contents using content rehashes to monetize on a massive scale.Reply
Zero creativity and innovation just pure pressure and memes. -
Stomx When some tech bloggers finally test the speed of all Windows versions for the last 25 years and show that all these "speed improvements" were within the error bars?Reply -
Mr Majestyk I wish Tom's would test a third party debloated Win 11 distro and see if that affects speed. I have removed as much as possible via autounattend.xml file on install, but there are people offering stripped naked versions. I wonder if these super-lean versions are quicker?Reply