Some performance discussion

[RaSystemlord]

Here are some perhaps interesting performance comparisons. The actual system and data descriptions are in the end. The data is with Handbrake, 4.2 GB video from Rings of Power and its typical conversion. Times to finish varied between 10 minutes and 105 minutes.

Faster here means that the time required to run is that much slower with the slower system (2 times faster means that the time for the slower system would be 2 times as much).

It is thought that Win 11 is slow to run virtualization, if you don't disable VMS&stuff (=all the extra security). According to these tests, it is also very slow as a virtual machine. Win 11 VM is much slower than Win Server 2025 VM - and it is really slow compared to Kubuntu VM.

Host performances are: SCAR, Z590, STRIX (Turbo) = 10.6 - 11.4 - 16.5 (minutes to complete the task). Please observe that SCAR gives about the same performance as computer than Z590.

Results from that: 3.5 year old workstation is equal to top of the line brand new laptop model. The new SCAR top of the line laptop running Win 11 is 60% faster than 3.5 years old AMD Ryzen 9, not top of the line, running Win 10 - not a good result for a very expensive, new laptop.

As a summary from the below, you can say that Linux should always be used, if possible. WIN 2022 Server should be used instead of Win 11 as a VM.

These tests are not ideal, nor complete, and thus not all the possible conclusions can be made.

---

Kubuntu is always faster in running VMs.

Z590 is 1.1 times faster than SCAR to run Kubuntu VM.
Z590 is 1.2 times faster than SCAR to run Win 11 VM.

---
Kubuntu Host is really much faster to run Kubuntu VM than Win11 to run Win 11 VM.

Z590 Kubuntu/Kubuntu is 2.2 times faster than SCAR Win11/Win11.

---

It is not only about Win 11 Host fine tuning for virtualization (=getting rid of all its extra security features),  because here we have Win 10 Host. Win 10 is also very slow in running Win 11 VMs compared to Kubuntu VMs. Running Kubuntu VM on Win 10 is over 2 times faster than running Win 11 VM.

Z590 Kubuntu/Kubuntu is 1.4 faster than Strix Win10/Kubuntu 10.
Z590 Kubuntu/Kubuntu is 3.7 faster than Strix Win10/Win11.

---

Running Win S 2022 is much faster as a VM than Windows 11.

Win 10 Host is 2.6 times faster to run Kubuntu VM than running Win 11 VM. It is not faster to run Kubuntu VM than Win S 2022 VM.

---

Systems:
Z590 (Workstation, Z590 series Asus ROG Strix Motherboard, i7-11700K 3.6 kHz, 8 cores, 32 GB)
SCAR (Laptop, Asus ROG Strix Scar, Intel Ultra 9 HX 275, 24 cores, 64 GB)
STRIX (Laptop, Asus ROG Strix, AMD Ryzen 9 5900 HX, 8 cores, 32 GB)

VMs had 4 processors and from 4-6 GB RAM (more on Windows).
Handbrake was 1.7.2 or 1.9.2 (newer on Windows).

There are slight differences, but they didn't do any favors for Win 11 nor Win S 2022.

[RaSystemlord]

VMware version was missing from the above, it is 17.5.2.

Comparison with the same system might be more interesting - I will lay them out later with Z590 system between Win 11 Pro 24H2 and Kubuntu 24.04.2. However, the Host systems, SCAR and Z590 give about the same performance as "standalone" computers and thus you CAN make deductions even now, about VM performance. The same system comparison will not give out any surprises: Win 11 Host is slower than Kubuntu Host - especially it is slow when running a Win 11 VM. With Strix Host, Win 10, you can see that, too.

Comparisons were made with Kubuntu, an Ubuntu derivate. I expect all major Linux systems give the same result. Kernel might make some difference, now it is 6.11 . Don't know if some different kernel models give a difference - like Ubuntu Studio no-latency kernel. I would expect them not to mean much - Linux is always faster or drastically faster than Win 11 Pro. Mind you, hardware virtualization is ON on those VMs.

Whether getting rid of Hyper-V and VBS will make Win 11 Pro 24H2 computer much faster - I don't know. Nobody has said it here - just talked about Nested Virtualization.

To have Intel hardware support for virtualization is believed here to have a big effect. Not really, with Kubuntu only a minimum effect. In many of the Windows tests, it WAS enabled - even though nothing of the Win 11 added security, like VBS, was disabled.

[Herb Fickes]

To get good performance with Windows 10 and 11, you must disable Hyper-V and Core Isolation.  By default, Win 10 and 11 run with these on and place the host Windows OS into a virtualized environment.

If you want VMWare to run fast, it has to have access to the hardware virtualization which means Hyper-V and Core Isolation must be disabled.  You can only run one or the other fast.

------------------------------
-Herb
------------------------------

[RaSystemlord]

Herb: Thanks for the comment.

I have not wanted to take them out for the reason mentioned elsewhere. Furthermore - it does NOT look like that:
- I AM able to have hardware support on in the VM (not tried Performance Counters but the main selection is ON). Are you saying that VMware selection doesn't mean anything? This is the same in Win10 and Win11
- Hardware Acceleration ON, meant almost nothing in Kubuntu ... the job times where about the same if it was ON or OFF

Is there a logical explanation for these?

[Herb Fickes]

I turned off Hyper-V and the Memory Integrity protection.  I wish I could keep the later on, but it uses virtualization and thus it has to be off as well.

Once these were off, all of my VMWare instances ran much faster.

Hyper-V runs at the hardware level and you can only have one virtualization engine doing that at a time.  If its running, then VMWare Workstation will still run, but it will be slower.  There are lots of articles about this, and it applies to software other than VMWare and Hyper-V.

Here are a few links.

https://learn.microsoft.com/en-us/troubleshoot/windows-client/application-management/virtualization-apps-not-work-with-hyper-v

https://www.switchfirewall.com/2025/03/vmware-performance-issues-on-windows11.html

[RaSystemlord]

I haven't read the articled, but I'm referring to posts here.

If only that would be that simple - either it is seriously incomplete information or wrong for Win 11. (only talking about Pro versions).

It's true that in my tests Kubuntu is always faster - there is no surprise there. It is a faster distro than Win 11 Pro. However, Kubuntu is NOT THAT MUCH faster to run a Kubuntu VM than Win 11 Pro is. (20% or so).

However, Win 11 Host is really slow to run Win 11 VMs. The same seems to apply Win 10 Host where those anti-virtualization things are not even completely ON.

OR are you saying that the HOST needs to run without VBS-thingies and also EVERY Win 11 Pro VM needs to be stripped from VBS & Hyper-V thingies? This is a serious reason not to use Win 11 VM but to use WIN S (referring to my test results) when Linux VM is not possible.

That is supported by the result that also Kubuntu runs Win 11 Pro VM slow - not that slow than Win 11 Pro Host, but seriously slow nevertheless. Kubuntu does not have - on the host -  things that slow down VMware.

There has also been a discussion why I'm using a CPU intensive test tool - the simple answer to that is that CPU power affects every heavy application and every heavy operation - save the 3D graphics in games and elsewhere, which I wasn't testing at all. What is there nowadays than heavy operations to be a problem - notepad you can use everywhere. There are other factors with may or may not favour one particular OS, but if that is so, these are the state of the art Windows computers that I'm using. I'm not testing hardware changes either, because I do have balanced systems to start with, I dare say. I'm using NTFS, which isn't ideal for Linux - I gather, but tests are not particularly disk intensive.

[Herb Fickes]

The host needs Hyper-V and those settings off to run VMWare Workstation at full speed.

[kasper]

I am not clear as to why you are using an esoteric benchmark test (handbrake, etc) vs a more generic application like GeekBench.   To make matters more confusing each OS setup has its own dependencies.   The biggest dependency is malware protection on the host and proper hardware driver configuration.   Performance measures for storage can vary greatly depending on caching.   Lets skip that one.   At the end of the day after you strip out all the extraneous OS thingys you may want to go with memory transfer bandwidth,  Whetstones and Drystones to measure performance.   However, if you are application bound, go with what you have.

[RaSystemlord]

Well, I'm not sure what confusion you mean - there has been talks about performance on this Forum, but no data. Now you have it. That shouldn't be confusing. You need to come up with your own data, if this is not interesting.

Generic performance tests can be difficult to evaluate in terms of any application performance. I chose an application where the performance really counts - whether your 20 piece season takes 4 hours or 16 hours to process probably means something, when your CPU is running 90% all the time.

Fine tuning hardware is a very different subject. I'm using existing hardware, also laptops, which are very balanced as they are - they are not food store PCs. I'm not testing PC's now - there is no need to test if ASUS ROG Strix (SCAR) computers are good - they are.

As such, I would like a discussion about OTHER results or the results that I just gave. To summarize: Win 11 Pro is lousy, also as a VM, which is perhaps something new information. Win Server 2022 is on bar with Kubuntu as a VM, which has been talked, here is now relevant data about it - WIN S isn't expensive, if you know where to buy it. Especially running Win 11 VM on Windows platform is very bad. If possible, use Linux on both ends, host or VM, preferably both - with Handbrake you are able to.

Handbrake HAS been used in many benchmarks. The benefit is that you CAN run it on Linux, not just Windows.

If not obvious, different applications demand different things. For instance a decent OpenGL can be a hard requirement even for user interaction, not just rendering the screen. Some of the benchmarks do something like "scrolling Excel back and forth" ... that might have been an issue 20 years ago, but nowadays it is just ridiculous to test Excel scrolling as if that would make a difference to anybody with a properly balanced computer. As I said, many benchmarks don't reflect real-world-requirements. Game performance benchmarks are realistic, because they use games, read applications, to test those. However, I'm not interested in testing those.

I hope this explains further what I was testing and why. Moving to Linux would be a way to go - wherever possible. Word and Excel can be used anywhere - there is no performance problem with that.

[kasper]

Have you removed malware protection as this can affect the results?   On Windows that would include the msmpengine (windows defender) and a bunch of other stuff.  Not sure what it would be for KUbuntu.

Benchmarking on frames per second can vary widely between OS's on the same hardware.   Best of luck getting the right drivers and other updates and tweaks to get the most out of your system.

[RaSystemlord]

Kasper, thanks for the tips!

However, I'm not really trying to tweak the systems, but checking what they come as they are. Malware protection - not useful for Linux, necessary for Windows and thus I don't want to take that out. Also VBS - well, don't know if that is included in Win 11 for a reason or just for trying to force people using Windows virtualization - be that as it may, I have it ON.

The drivers, the same thing. Microsoft loads what it loads. Many of them are in Linux kernel and no reason to tweak anything. However, I'm talking about Asus Rog Strix (Scar) computers and the expectation is that everything important comes from Asus (through ASUS Armory Crate software), for gaming, and thus they are the best there is for Windows. For Linux, they are what Kubuntu offers as standard.

I added the "same computer", Z590, i1 11nd gen, comparisons to my matrix. There were no big surprises. Kubuntu runs both Win 11 and Kubuntu VMs 1.1 - 1.2 times faster compared to Win11 Host. Kubuntu/Kubuntu is 1.6 faster than Win11/Win11 - which is inline with STRIX results. (The details of these systems are in previous posts).

Frame rates & resolution - yes, important in gaming, but I'm not testing any of that.

It seems that Win Workstations are especially bad in hosting Win 11 workstations. Hosting WIN Server 2022 is much better and so is hosting Kubuntu VM. Kubuntu is in all cases better or much better than Win 11/10 Pro versions - in hosting or as a VM. WIN Server 2022 Host was not tested at this time, only the VM was.

[kasper]

If you are Handbrake app centric (I use this on occasion, nice app) and want to run it on a guest then for best performance go with Linux host/guest.