Announcement

Collapse
No announcement yet.

Ubuntu 21.10 Performance Still Pushing Ahead Of Windows 10, Latest Windows 11 Build

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Ubuntu 21.10 Performance Still Pushing Ahead Of Windows 10, Latest Windows 11 Build

    Phoronix: Ubuntu 21.10 Performance Still Pushing Ahead Of Windows 10, Latest Windows 11 Build

    With less than one month out from the official release of Microsoft Windows 11, I was curious to run some fresh benchmarks of the latest Windows 11 Insider Preview build against Windows 10 21H1 to see how the performance is looking. Of course, also to see how Windows 11 is shaping up against Ubuntu 21.10 also due for release in October.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Originally posted by Setif
    Windows 10/11 Pro is not optimized for more than quadcore.
    Compare with Windows 10/11 Pro for Workstations.
    How do you know this? Links?

    Comment


    • #3
      Originally posted by Setif
      Windows 10/11 Pro is not optimized for more than quadcore.
      Compare with Windows 10/11 Pro for Workstations.
      Pro supports up to 2 physical CPUs and 2TB of ram whereas Pro for Workstations supports up to 4 physical CPUs, 6TB of ram, and has some extra features like SMB Direct and ReFS.

      Comment


      • #4
        And where exactly does it say so?
        The first paragraph states not quadcore but 64 threads, which is plenty:
        Most home users that have Windows typically have versions of Windows 10 Home or Windows 10 Pro, which are both fairly ubiquitous even among workstation users. The problem that these operating systems have rears its ugly head when we go above 64 threads. Now to be clear, Microsoft never expected home (or even most workstations) systems to go above this amount, and to a certain extent they are correct.

        Comment


        • #5
          How is this relevant to these tests when 5900x has 24 threads? Also, where did you hear that the pro edition is not optimized for more than quadcore?
          Last edited by user1; 10 September 2021, 10:36 AM.

          Comment


          • #6
            Originally posted by Setif
            Windows 10/11 Pro is not optimized for more than quadcore.
            Compare with Windows 10/11 Pro for Workstations.
            Oh, no. Not this discussion again. I remember when Michael published the first benchmarks for AMD Threadripper a couple of years ago comparing Windows vs Linux. Linux completely destroyed Windows, it was not even close, it was humiliation, embarrassing. Some people could just not believe that Windows was so bad at NUMA compared to Linux, they suggested that it must be because of the Windows version used, or some configuration missing or some "driver package" not yet available for Windows or some other nonsense. But as we all know, that was not the case. Windows did really, really suck compared to Linux, at least initially for the first generation Threadripper. Later benchmarks by Michael has shown Windows to be better now for Threadripper, but it still trails Linux.

            EDIT:

            Found the latest benchmark article for Threadripper:

            https://www.phoronix.com/forums/foru...er-3990x/page2

            Linux is still faster of course. But the funniest part is this comment from Setif
            ​​​​​https://www.phoronix.com/forums/foru...05#post1258705

            ​​​​​I think you should have installed Windows 10 Pro for Workstations or Windows Server 2019 which have a better support for such CPUs.
            😂
            Last edited by tomas; 10 September 2021, 01:05 PM.

            Comment


            • #7
              understand that some newer code benchmarks understand that windows programs run with multiple processor groups because they know how to allocate like winspd.

              older dumber windows code can be run with (say you have two processor groups) :
              start /NODE 0 /AFFINITY 0xFFFFFFFFFFFFFFFF dumb_windows_program.exe --run--threads--for-example===64
              start /NODE 1 /AFFINITY 0xFFFFFFFFFFFFFFFF dumb_windows_program.exe --run--threads--for-example===64

              ___^^^___theoretically__^^^___






              Comment


              • #8
                Guys, seriously, if you want Michael to run benchmarks on a different proprietary OS, then you really should pony up and buy him a license to dis/prove your point. Otherwise, shut up about it.

                Comment


                • #9
                  Originally posted by stormcrow View Post
                  Guys, seriously, if you want Michael to run benchmarks on a different proprietary OS, then you really should pony up and buy him a license to dis/prove your point. Otherwise, shut up about it.
                  Nope, sorry, that is not how it works. These benchmarks show that Linux is performing better than Windows for this specific configuration. If you or anyone else think the results would be any different with a different version of Windows, then it is you that has the burden of proof and you that should "pony up" and buy him a license to prove your point. Otherwise you can stay silent.
                  Last edited by tomas; 10 September 2021, 01:25 PM.

                  Comment


                  • #10
                    Originally posted by tomas View Post

                    Nope, sorry, that is not how it works. These benchmarks show that Linux is performing better than Windows for this specific configuration. If you or anyone else think the results would be different with a different version of Windows, then it is you that has the burden of proof and you that should "pony up" and buy him a license to prove your point. Otherwise you can stay silent.
                    If you have a problem with the benchmark, YOU need to prove your point. That's exactly how it works! Burden of proof is on the person raising the objection. You don't get to say "well this is wrong because I have my reasons". The article you're pointing at doesn't support your conclusions. If you want Michael to check your assertions for you, then you need to pay for the license to do so. Pro or Pro Workstation shouldn't make a huge difference on 5900X performance according to Microsoft's own support documentation (being well within Pro's support matrices) unless there's something there that Microsoft hasn't well communicated (which is possible given Microsoft's gawd awful customer communication).

                    Historically, Windows has been awful - and I mean horribly awful - at HPC & highly multi-threaded loads. There's a reason why Windows Server isn't in the top 10 super computers and it's not because of Unix bias. The current benchmarks are unsurprising.

                    Edit: I wrote Threadripper rather than 5900X. Oops!

                    Additionally I might add that Windows has undergone three feature updates since then. That article is a year and a half old, and it only measures 64+ threads.
                    Last edited by stormcrow; 10 September 2021, 01:54 PM.

                    Comment

                    Working...
                    X