Clean Windows 11 23H2 benchmarked against Windows 11 22H2

Sayan Sen contributed to this feature, and also provided the benchmark graphics.

As you may be aware, we have been benchmarking different scenarios with the Intel Core i9-14900K sent to us. It provides an ultra-modern test bed to test Microsoft's flagship Windows 11 OS, but also against the older, but more trusted, Windows 10 counterpart.

However, today's benchmarks pitches Windows 11 against itself! To put it plainly, which is better, Windows 11 22H2, or the newest 23H2 with Copilot and all the changes that it brought with it?

First up, here is a run-down of our test system, you may recognize it from the 14th gen Intel Core i5 and i9 CPUs review we did in October:

Test Hardware

• Intel Core i9-14900K (Amazon|Newegg)
• ASUS ROG STRIX Z790-H GAMING WIFI (BIOS v1402) (Amazon|Newegg)
• ASUS ROG STRIX 850W Gold (Amazon|Newegg)
• Thermalright Peerless Assassin 120 SE ARGB CPU Air Cooler (Amazon|Newegg)
• Carbonaut thermal grizzly 0.2mm (Amazon|Newegg)
• 2x 16GB Corsair Dominator Platinum RGB DDR5 6000MT/s in XMP (Amazon|Newegg)
• Sabrent 1TB Rocket 4 Plus-G (PCIe 4.0) (Amazon|Newegg)
• MSI GeForce RTX 4070 Ti Ventus 3X (Amazon|Newegg)
• Samsung Odyssey G9 Neo at 5120 x 1440 @ 240Hz on DisplayPort (Amazon|Newegg)

Software test scenarios
On both systems, ASUS Armoury Crate was allowed to install all the drivers that our system needed.

October (Windows 11 22H2)
Data used below is from when Windows 11 Pro was clean installed for the Intel 14th gen CPU review, 22H2 was updated to Build 22621.2468 (Moment 4) and Nvidia driver 538.58 was used, which was the latest at that time.

November (Windows 11 23H2)
The same hardware was clean installed and updated to build 22631.2715 (KB5032190) and Nvidia driver 545.92 was used, which at the time of testing, was the latest.

The ambient room temperature while testing was measured to be between 19C (66.2 °F) and 20C (68 °F), and our test rig was open plan.

Benchmarks

In 3DMark, supplied to us by UL Benchmarks, it's clear to see that 23H2 does a better job overall in the Time Spy benchmark (DirectX 12), that was allowed to run three times, with the best score used. For CPU score, 23H2 has a 958 point lead, and an impressive 427 point gap in the overall score, which combines GPU as well as the CPU score.

In terms of percentages, those are 4.4% and 2% better respectively.

In Fire Strike Extreme, which is DirectX 11, the gap narrows in Physics to just 82 points, and 134 in the overall score with 23H2 again proving to be better than its predecessor (best score after three runs). Hence the gap is not really as significant as in Time Spy.

In the CPU Profile test, which tries to gauge how good the scaling is across CPU threads, 23H2 once again takes the crown, besting 22H2 at every stage except for the single thread; however, it is completely within the margin of error with a difference of only 3 points. The highest scores were used after running the benchmark three times.

The PCMark 10 benchmark suite, performance is a mixed bag with 22H2 coming out on top overall in six of the eleven tests, but specifically in the productivity benchmark. And the difference is quite astounding, to put it mildly. Going from 22H2 to 23H2, we see a 15.24% drop in the spreadsheet processing test and a 7.57% drop in the case of writing.

In the Digital Content Creation, they trade blows with 23H2 taking the win in Rendering and Visualisation while 22H2 emerges victorious in Photo Editing.

Finally, in the Essentials benchmark which considers app startup, web browsing and video conferencing, the two versions of Windows are quite close to one another with 23H2 winning two out of the three tests, though by very small margins.

23H2 beat out 22H2, though, by only 14 points in the Cinebench 2024 CPU benchmark. This falls in line with PCMark 10's Digital suite, which means you wouldn't really feel the difference in a blind test.

We also ran the default 32MB 7-Zip benchmark which measures in GIPS (billion instructions per second). Here 23H2 edges out 22H2 in both Compression and Decompression tests, but the differences are once again within the margin of error.

So what have we proven here? Well, we can see that 23H2 really is a worthwhile update, as it bests 22H2 in 17 of the 24 data points tested versus just seven victories for 22H2. Microsoft could claim some performance gains with this update and they would not be lying.

You might be wondering why we didn't mention the PCMark 10 productivity suite and that's because we are considering this benchmark as an outlier similar to what we found in the previous Windows 11 vs 10 comparison where the Rendering and Visualisation numbers heavily favored one over the other.

Outside of PCMark productivity, we also did not notice much of a dip going from 22H2 to 23H2 and hence, the recent reports about performance loss are most likely limited to buggy Defender.

Officially, our advice is to update.

Check out our other benchmark coverage of Clean installed Windows 10 22H2 vs Windows 11 23H2 benchmarked for performance, Windows 10 clean install vs Windows 11 in-place upgrade and our Intel Core i5-14600K and i9-14900K review.

As an Amazon Associate I earn from qualifying purchases.