Back when Windows 11 launched in 2021, Microsoft, as with any other company, wanted to hype people up about its new product. One aspect which the company focused on was suggesting that users would get better performance on 11 as the OS was designed to get the best out of their hardware. This was a bit of an ironic statement considering Microsoft introduced upgraded system requirements for its new operating system and, to this day, as it recently clarified, has stuck to it.
After the promise of better performance fell mostly flat, a couple of months later in November 2021, Microsoft devs promised that improving the overall snappiness of the various UI elements in the OS was its main focus in 2022.
A year and a half later, in May of 2023, Microsoft published a blog post highlighting all the improvements that it had made till then, although third-party testing from a few months prior suggested no notable improvements.
We have tried testing Microsoft's claim in the past in both scenarios, an in-place upgrade and clean installation. Interestingly, we did find Windows 11 to be faster in some scenarios, although nowhere near the margin of what Microsoft wants you to believe by making up ridiculous performance claims using paid studies, for example.
Besides Windows 10 vs 11, the performance differences between the versions of Windows 11 too are something worth looking at. As such, we decided to see what a user going from feature update 23H2 to 24H2 would experience.
For a clear progression, Windows 11 24H2 has to perform at least somewhat better, and some users do claim that is indeed the case even when running unsupported systems.
Unlike last time, our test would not just be straight-up apples-to-apples comparisons here. We are trying to replicate the usage of a general user and as such, all settings would be kept at default. This is how we will be evaluating the performance differentials for an in-place upgrade to Windows 11 24H2 from 23H2.
Thus, some of the settings like Core Isolation or Virtualization-based Security (VBS) have been kept enabled. Since we are looking at an in-place upgrade experience here, we updated our system to the first 24H2 Patch, ie, October 2024 Patch Tuesday (KB5044284).
In case you are wondering about the hardware we tested this on, Steven and I worked on this together (remotely), on his test bench comprising Intel's Core i7-14700K CPU and AMD's 7800 XT GPU (provided by AMD for review). The latest GPU drivers were used at the time (Adrenalin version 24.10.1).
We start with synthetic benchmarks first.
We measured the CPU performance in synthetic gaming tests using Physics / CPU tests available in 3DMark's Fire Strike Ultra and Time Spy Extreme. The former is based on DirectX 11 while the latter is based on DirectX 12.
Interestingly, as you can see, 24H2 does significantly worse in Fire Strike. We wonder if this is because of the single-threaded nature of DX 11 which is something the 2024 Windows 11 feature update is not all that optimized for.
Following that we have 3DMark CPU Profile which is essentially a scaling test, and bizarrely, Windows 11 24H2 shows poor scaling for some reason.
After looking at the Fire Strike Ultra results above, we were honestly expecting the opposite, as we figured 24H2 may excel in multi-threading. Clearly, we were wrong.
After that, we ran 3DMark GPU tests. We chose Speed Way and Steel Nomad since these are more modern than Time Spy Extreme.
Although both Time Spy Extreme and Steel Nomad are running on DX 12, the latter has more modern workloads which is more representative of typical modern rasterization. Meanwhile, Speed Way is a DirectX Ray Tracing benchmark. [Please note here that we used the white background in our charts to indicate a CPU-heavy test and a radiant gray background was used when it's a GPU-heavy load. Thus the 3DMark chart below is radiant gray.]
The 7800 XT performed similarly on both 24H2 and 23H2 and the scores were within the margin of error.
Following synthetic gaming tests, we move on to real games.
We already mentioned that we were sticking to default settings since we are trying to evaluate the standard experience of a user, hence, we did not tweak any game (graphics) settings. As such, we were not GPU bound.
We start with the most interesting outcome as we are greeted with enormous performance improvement in Black Myth: Wukong on Windows 11 24H2 when Frame Generation (FG) option was enabled.
In the average framerate, we saw a 17.3% bump, while in the 95%iles, it was 15.5% better. This surprised us a lot as we were not expecting to see such a massive difference in any title.
With frame generation off, the performance was nearly identical.
Another interesting result was in the case of Far Cry 6 where we see that Windows 11 24H2 lags behind version 23H2. Far Cry games are known to be single-thread bound often, so this may perhaps be explained by the initial poor result we saw in 3DMark Fire Strike Physics.
In Assassin's Creed Odyssey, which is yet another Ubisoft title, the two Windows versions run equally great.
We also included Intel's APO results in supported games. In the case of Marvel's Guardians of the Galaxy, we saw no differences.
On Shadow of the Tomb Raider, the APO did not work as it kept crashing on 24H2, but worked fine on 23H2. But besides those, the scores were within the margin of error.
Finally, we have Final Fantasy XV and The Callisto Protocol.
As you can see in these two titles, neither Windows version was truly better.
After games, we move on to productivity tests, and we start with Cinebench 2024 CPU (multi-threaded) and GPU rendering tests, and there is nothing too exciting going on here.
Next up, we have 7-Zip, and while the decompression performance of both systems was pretty even, in compression, Windows 11 24H2 was around 6% ahead.
Finally, we test PCMark 10 to get a general idea of how the two systems feel.
The first thing we notice on Windows 11 24H2 (below) is that the app start-up performance for some reason is significantly lower than on 23H2 (above). Video Editing is another area where 23H2 comes on top.
Overall, 23H2 definitely has the edge with 10,734 points compared to 10,460 on 24H2.
Finally, we tested memory allocation on Windows 11 24H2 and 23H2. This shows how much system RAM each OS requests for caching page files, prefetch data, game/app data, among other things. Speaking of RAM, we were using TeamGroup's 32 GB DDR5-7600 kit that we reviewed last year.
Again, interesting results here as we find Windows 11 version 24H2 consistently allocates much more memory in apps and games than 23H2. The only exception was in 7-Zip.
So should you upgrade?
given that Microsoft has now started pushing the feature update out to more users.The answer depends on the situation. 24H2, for example, showed a clear victory in the case of frame generation, though, in single-threaded instances, Windows 11 24H2 started to fall behind. And also, if snappiness is your priority, then you may find 23H2 better since the app start-up performance on 24H2 was significantly behind.
RAM allocation is also something to keep in mind since 24H2 may warrant a minimum of 32 GB.
Those were our results. What has your experience on 23H2 or 24H2 been like?
20 Comments - Add comment