There’s a new Microsoft video on Windows 11 making the rounds, but it doesn’t contain the performance proof that’s been claimed for it in various press writeups. The video in question is a discussion of various aspects of the new OS, but the actual performance demonstration doesn’t demonstrate anything.
This is a little odd, honestly, because we genuinely expect that Windows 11 will deliver better performance than Windows 10 in certain scenarios. When Apple launched the M1, reviewers noted that M1-equipped laptops felt much more responsive than their x86 counterparts. While the M1 is an excellent mobile CPU, lightweight OS tasks aren’t typically considered difficult enough to create meaningful desktop performance differences.
It turned out that the M1 delivers excellent performance by shifting background tasks explicitly to low-power cores while foreground tasks are executed almost exclusively on “big” cores. This improves system responsiveness and the overall user experience. It was a clever trick, and we know that both Intel and Microsoft are planning to take advantage of similar ideas in Windows 11.
Intel’s upcoming 12th Generation Alder Lake supports features like Intel Thread Director, an embedded microcontroller that monitors how workloads are executing across the CPU and communicates this information back to Windows 11. This allows the OS scheduler to make decisions about which threads should be executed on which CPU cores with more granularity than ever before. We’ve seen leaked benchmarks suggesting that hybrid CPUs like Lakefield, which does not support ITD, still pick up 5-10 percent performance under Windows 11. Alder Lake may benefit even more from these scheduler improvements compared with Windows 10.
Microsoft’s video discusses how these new features should work in the future and mentions that the OS is designed to give execution priority to foreground applications. To demonstrate this, Microsoft VP Steve Dispensa shows a system running at 90 percent CPU usage. Despite this relatively high level, he’s able to launch both Excel and Word in roughly a second.
We were curious how a modern Windows 10 system would behave in the same scenario, so we fired up the same “HeavyLoad” application you can see in the video above and ran it on a testbed. Microsoft says nothing about how it configured the app, but we can see from the video that it consumes 20 to 40 percent of the CPU’s performance and puts around a 30 percent load on the GPU.
We tested the application with 28 out of 32 CPU threads loaded on a Ryzen 9 5950X and its GPU testing mode engaged. Once it started, we opened both Excel and Word from a clean boot. It took roughly 1s to open both applications — about the same amount of time as shown in Microsoft’s test. We confirmed that HeavyLoad had stressed our system to 80-90 percent resource utilization before running the comparison. Changing the default priority from “Below Normal” to “Normal” did not seem to impact Word or Excel’s launch time.
Using real-world applications created a better test. Upscaling two separate videos in Topaz Video Enhance AI while simultaneously encoding two different videos using FFmpeg increased the launch time for Excel and Word to ~2s each. But this test scenario included multiple applications known to put a heavy load on the CPU and GPU, with total CPU usage pegged at 100 percent. Of course, it would be different if Microsoft was running these tests on a low-end Atom CPU or a similarly weak system, but the company didn’t disclose any testbed details.
In this case, we don’t think Microsoft is being dishonest. Its test just doesn’t seem to demonstrate the actual improvement of Windows 11 very effectively. It’s difficult to showcase UI performance improvements unless they’re large, so the best way to demo this kind of feature would be to show an identical Windows 10 and Windows 11 system performing a series of operations in a foreground application while already under heavy load from background tasks. If the Windows 11 system is allocating resources more effectively than its Windows 10 counterpart, we should see that reflected in faster screen updates and smoother UI transitions.
By all means, watch the video above if you want more general information on Microsoft’s upcoming OS, but don’t expect to see evidence of meaningful performance improvement. There’s nothing on display here that Windows 10 doesn’t match. This seems to say more about the method Microsoft chose for showcasing its improved performance than anything in particular about the OS itself.
Alienware Really Doesn’t Want You to Buy an AMD Ryzen PC
Just because Alienware sells Ryzen systems doesn't mean the company wants you to buy one. All objective evidence suggests it would rather you didn't.
PlayStation 5 Hits Game Data Compression Ratios the Xbox Doesn’t Touch
Sony's PlayStation 5 offers much smaller downloads for some games compared with the Xbox Series S|X. We dive into the compression technology behind the company's achievements.
No New Version of Windows Can Turn Back the Clock, and Microsoft Doesn’t Want To
Microsoft will unveil Windows 11 on Thursday, but the company isn't likely looking to reboot its operating system — and it doesn't need to.
New Study Suggests Dark Matter Doesn’t Exist
Most scientists currently believe the iron grip of gravity is augmented by dark matter, an invisible material that makes up about 85 percent of the universe. A new study makes the case for an alternative model, one in which dark matter doesn't exist and gravity works a little differently than we thought.