Windows 11: First measurement results | heise online

The most prominent innovations of Windows 11 are, for once, no changes to the user interface or new functions, but the massively increased hardware requirements. But does that mean Windows 11 is slower? Or is everything no longer running? To find out, we put Windows 10 and 11 against each other.

More from c't magazine

More from c't magazine

More from c't magazine

More from c't magazine

Admittedly, we did not have high expectations during the preparations, because in the past such comparisons were usually rather boring. Since Windows Vista, the speed differences have been drowned out in the statistical noise. This did not change the fact that the perceived speed of Windows Vista was remarkably poor, especially before the release of Service Pack 2, and that Windows 8 felt unusable for many due to the lack of a start button and thanks to the start page instead of the menu. The measuring programs do not capture such subtleties, even if they drive some people crazy.

Only one thing had actually changed not only noticeably, but also measurably since Windows 8: Windows boots significantly faster thanks to a trick. When you click on “Shutdown”, Windows closes all running applications, but just puts itself to sleep (hibernation, S4). Waking up from it works much faster than starting up.

Our compatibility tests used to end rather boringly: Apart from the expected exceptions (software running close to the system, such as virus scanners), everything also ran under the newcomer.

But what about Windows 11 now? Microsoft justifies the increased hardware requirements primarily with improved security measures. Unlike its predecessors, Windows 11 requires a “compatible” processor, UEFI Secure Boot and TPM 2.0 as standard. All of this is necessary, among other things, for “virtualization-based security” (VBS, auch Virtualization-based Security). The details are complicated, but ultimately VBS creates virtual enclaves in memory that are isolated from the rest of the system. This should protect the data contained therein from attacks, such as certificates and password hashes. You can check whether VBS is active with the Windows program “System Information” (msinfo32.exe): On the first page it is relatively far down in the line “Virtualization-based security”. If not, you can activate it, although so far only one is known to us as a reliable way via the registry: Create under HKLMSYSTEMCurrentControlSetControlDeviceGuard a DWORD called EnableVirtualizationBasedSecurity with the value 1 and restart Windows.

VBS also includes “Hypervisor-Protected Code Integrity (HVCI)“, with which security-critical code can be protected. In the future, for example, banking apps should be better protected and the use of cheats in online games prevented. Here, too, msinfo32.exe reveals the status, also on the first page in the lines” Virtualization-based Security – configured services “and” Virtualization-based security – executed services “. If there is nothing behind this, HVCI is inactive. Activation is possible with a click of the mouse: Start the” Windows Security “app and click on the link” Device Security / Core Isolation ” Details on core isolation “. Slide the switch to” On “. If all drivers are playing, Windows requests a restart, after which HCVI is active.

However, these security functions are not really new, because they are already in Windows 10. The difference is elsewhere: Windows 11 tries to activate both VBS and HVCI by default. However, not on every supposedly Windows 11 compatible hardware. Microsoft states, among other things, a compatible CPU and HVCI-compatible drivers as prerequisites.

For most of the measurements for this article, we used a test computer with an Intel Core i3-8100, 8 GB DDR4-2133 RAM and a 256 GB Samsung SSD of the “850 Pro SATA” type. The processor meets the hardware requirements of Windows 11, but only barely on purpose. Because the slower a test computer is, the more likely it is that measurement differences are noticeable.

The decision for an Intel processor was made because AMD processors were causing problems under Windows 11 at the time of going to press: The L3 cache is sometimes dramatically slower than under Windows 10. Although Microsoft and AMD released the first patches and new chipset drivers shortly before going to press, they did they were able to alleviate the symptoms in our tests, but not yet eliminate them. More patches are likely to follow.

We use the “Bapco Sysmark 25” benchmark suite for our measurements. It brings various retail versions of known and widespread applications and measures how long typical work processes take with them. Adobe’s applications include Acrobat Pro, Lightroom Classic, Photoshop and Premiere. From Microsoft, Excel, Outlook, PowerPoint and Word are each included as version 2019. In addition, there is the audio editor Audacity, the scripting language AutoIt, the compression program WinZip 24.0 Enterprise from Corel and Google’s Chrome browser 81.0.

Bapco Sysmark expresses the results with points in different categories: Productivity, Creativity and Responsiveness, and there is also an overall rating.

Operating mode

Windows 10 Version 21H1

Windows 11 Version 21H2

UEFI without VBS / HCVI (4 cores)

1031 (1061/1082/817)¹

986 (1017/1043/760)¹

UEFI without VBS / HCVI (2 cores)

814 (883/791/696)¹

779 (850/763/636)¹

Legacy (4 Kerne)

1030 (1058/1086/813)¹

988 (1026/1040/755)¹

UEFI with VBS (4 cores)

998 (1018/1060/782)¹

962 (989/1024/731)¹

UEFI with VBS and HVCI (4 cores)

997 (1020/1056/781)¹

954 (974/1019/737)¹

¹Bapco-Punkte: Overall Rating (Productivity/Creativity/Responsiveness)

We tested Windows 10 and 11 with active and inactive VBS as well as with active VBS and HVCI. We wanted to prevent that we might consider Windows 11 to be slower just because a security function is active that is deactivated in Windows 10.

The results show that the safety functions actually brake, but, as expected, both systems do the same. We did not expect that Windows 11 would run slower with the same security settings: Windows 11 only achieves a good 95 percent of the performance of Windows 10. Also striking: Active HVCI does not slow down Windows 10 at all, but Windows 11 does, albeit very little – we do not yet have an explanation for this. The number of cores does not seem to have any influence. Test runs with the same CPU, for which we switched off two cores manually, showed the same differences between Windows 10 and 11.

Active security functions slow down Windows 11, but this also applies to Windows 10. But with identical settings, Windows 11 is still not as fast as its predecessor.

For a control measurement, we reinstalled both Windows versions on the same computer so that they start not via UEFI, but traditionally via legacy BIOS. The measurement results correspond to those of the UEFI installations without VBS and HVCI. So here too, Windows 11 was a bit slower.

To put it into perspective: The speed disadvantage of Windows 11 is indeed measurable, but practically never noticeable. Nevertheless, it is the first time in a long time that a tempo difference can be found at all, and that in the “wrong” direction.

To home page