AMD Ryzen 7 7800X3D: Windows 11 vs. Ubuntu 23.04 Linux Performance

[flower]

Perf difference is frequently 5-10% difference but as much as 20% in some cases and on some hardware. HVCI / VBS can significantly impact performance because they virtualize the entire operating system, and segment off a separate high-trust kernel, along with a lot of memory translation overhead.

Windows Virtualization has an up-to 20% perf penalty? That really sounds too much (or they need to optimize it)
I've played for years in a windows vm running on a linux host and never saw a performance impact that big.

only thing that really had an impact was heavy io

[mdedetrich]

Windows Virtualization has an up-to 20% perf penalty? That really sounds too much (or they need to optimize it)
I've played for years in a windows vm running on a linux host and never saw a performance impact that big.

only thing that really had an impact was heavy io

Yeah that makes zero sense. Windows Virtualization is heavily used in Azure, an area where Microsoft cannot afford a 20% perf penalty since that is their biggest sources of revenue.

[smitty3268]

The problem is that it is not an apples-to-apples.

Nothing is apples to apples, you are comparing entirely different OS's that do tons of things differently. The only halfway reasonable thing to test is the defaults, since that's what 99% of people use. That's as apples to apples as you are going to get.

Linux has plenty of non-optimal defaults too, and you see people in these threads all the time complaining that Michael didn't do stuff like change the scheduler to being performance mode, because it's so easy and everyone does it. But the truth is, most users actually don't do it. They use the defaults.

[Anux]

I agree with smitty3268 here. Those that play around with spectre and VBS are clearly a minority and it would also be a never ending search for the right settings. What if I deactivate defender and not VBS or the other way around? Just testing the standard is probably the only way to go as long as you don't want to compare specific functions of an OS.

Also up to 20% slowdown from virtualisation? Whoa, that is what MS shoves their users up the a** with a standard install? Than it's not to bad to test this so MS has a reason to go fix their VM. KVM/Qemu can do much better.

[ll1025]

I agree with smitty3268 here. Those that play around with spectre and VBS are clearly a minority and it would also be a never ending search for the right settings. What if I deactivate defender and not VBS or the other way around? Just testing the standard is probably the only way to go as long as you don't want to compare specific functions of an OS.

Also up to 20% slowdown from virtualisation? Whoa, that is what MS shoves their users up the a** with a standard install? Than it's not to bad to test this so MS has a reason to go fix their VM. KVM/Qemu can do much better.

As I said it depends heavily and is usually 5-10%. It's also not just virtualization. The hit is on par with other mitigations that Linux enables-- which Windows *also* enables, sometimes with a lower hit. And VBS is not a consistent default-- it can be, with the right hardware, on a fresh install of Windows 11. Or it might not be, if you have an older install, or performed an upgrade, or don't have virtualization enabled in the BIOS, or have machines from certain OEMs, or are virtualizing Windows, or have policies disabling VBS due to its various errata.....

Defender should not be factored in because you're again testing unlike to unlike. Corporate environments where this stuff matters often mandate endpoint detection products for linux, and home users who care most about performance are frequently disabling mitigations whether on Windows or on Linux. Running a performance test where one install has antivirus and another does not is straight biased.

If the goal is to compare the two operating systems performance then the goal should be to minimize other factors. If this is to be a general "what is it like running Windows vs Linux", then I'd expect to see some of those differences detailed (e.g. that Windows does a better job of protecting the kernel from compromise; that Linux has a better native hypervisor). But it isn't that, it's a performance comparative, and so it is reasonable to suggest that things like VBS that are not a default in all cases be turned off.

[ll1025]

flower mdedetrich
I said that it was 5-10% in benchmarks and *up to* 20%, in certain cases, and that is NOT just virtualization overhead. VBS is a lot more than just virtualization a la Azure, and involves a lot more context switching. Normal virtualization of an OS has that one OS operating in a single trust domain. VBS has the kernel and userland in different domains, with a second kernel to boot. That involves some penalties that normal VM usecases are never going to experience.

[smitty3268]

things like VBS that are not a default in all cases be turned off.

It's either a default or it isn't. Tests on more hardware are always welcome, and that can pick up cases where Windows is faster on some hardware than others, if that's really the case.

Frankly, this is a little silly. Windows typically has far more optimized default options for desktop usage than Linux does, which has all sorts of things optimized more for server use OOTB. If you are going to say the systems should be optimized, that's likely to end up hurting Windows more comparatively.

[Amarildo]

Linux fans claim that Windows is so slow that it is impossible to work on it.

Because it is. From boot, to opening programs, to working on those programs. I'll give you a few examples:

- Booting my Fedora takes 4 to 5 seconds. Windows (on the same NVMe drive) takes 10;
- Opening GIMP in Linux takes literally 1 second. On Windows it takes 4 every single time;
- Opening Autodesk Maya 2024 on Windows takes 17 seconds for me. On Linux? 5 to 7.

And not only is Linux more "snappy", it's has 0 bloat, 0 privacy issues, 0 licensing issues, 0 pricing issues, 0 updating issues....... and it runs the programs better too.

[ll1025]

Because it is. From boot, to opening programs, to working on those programs. I'll give you a few examples:

- Booting my Fedora takes 4 to 5 seconds. Windows (on the same NVMe drive) takes 10;
- Opening GIMP in Linux takes literally 1 second. On Windows it takes 4 every single time;
- Opening Autodesk Maya 2024 on Windows takes 17 seconds for me. On Linux? 5 to 7.

And not only is Linux more "snappy", it's has 0 bloat, 0 privacy issues, 0 licensing issues, 0 pricing issues, 0 updating issues....... and it runs the programs better too.

0 pricing issues? I'll put you in touch with our RHEL VAR. And on the bloat front, you haven't used Ubuntu recently have you?

Good and "0 issues" are not the same and saying something has "0 issues" suggests your usage of it is very shallow.

[Anux]

0 pricing issues? I'll put you in touch with our RHEL VAR.

And what does RedHat have to do with the pricing of random distros? You can easily give a RedHat Linux free of charge to whomever you want.

And on the bloat front, you haven't used Ubuntu recently have you?

The good thing with Linux, there is not only one that you are forced to use.

Good and "0 issues" are not the same and saying something has "0 issues" suggests your usage of it is very shallow.

Agreed.