Announcement

Collapse
No announcement yet.

VMware's Linux 3D Guest Driver Is Ready

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

  • #16
    Originally posted by allquixotic View Post
    -We're using completely different host operating systems
    -We're using completely different guest operating systems
    -You're using Player, and I'm using Workstation
    As already specified by others, compare hardware. It could well be a problem with a peripheral, or the CPU's virtualization features, or something.

    Also, compare guest machine configurations. There's a lot of options that could potentially have a very meaningful impact on some hardware. Number of cores, choice of emulated hardware, CPU virtualization configuration, etc.

    Comment


    • #17
      Originally posted by elanthis View Post
      As already specified by others, compare hardware. It could well be a problem with a peripheral, or the CPU's virtualization features, or something.

      Also, compare guest machine configurations. There's a lot of options that could potentially have a very meaningful impact on some hardware. Number of cores, choice of emulated hardware, CPU virtualization configuration, etc.
      OK.

      Physical:
      CPU: Core i7 920 (original Nehalem, quad core, HT)
      Mobo: Asus P6T Deluxe
      RAM: 16GB dual channel DDR3 1600
      GPU: Radeon HD7970
      Audio: Sennheiser USB 2.0 headset
      Physical monitor config:
      2011 model Panasonic 1920x1080 HDTV connected by native HDMI 1.4a (no cable adaptors between monitor and GPU)
      2008 model commodity Dell 1680x1050 LCD connected by DVI converted to Mini-DisplayPort (uses DVI on the monitor end, plugs into Mini-DP on the video card)
      2004 model commodity Dell 1920x1200 LCD connected by DVI converted to Mini-DisplayPort (same connectors as other LCD)

      Guest config:
      OS: Linux Mint 12 with Cinnamon desktop
      Memory: 4GB (even with this I still have 4 to 8GB of completely unused RAM on the host)
      CPU: 1 x 4-core CPU
      "Virtualization Engine": Default setting of "Automatic" with the "Virtualize Intel VT-x/EPT" setting unchecked
      HDD: 200 GB SCSI virtual disk residing on a separate physical HDD than the host operating system HDD, but mapped as a file on an NTFS filesystem
      Monitor config: "Dedicated" full screen on a 1680x1050 LCD (Host OS programs configured not to start on second monitor so it's always just VMware on that monitor)
      Graphics: "Accelerate 3d Graphics" checked
      Kernel: 3.3 rc7-precise from Ubuntu Mainline Kernel Builds
      DDX: Latest VMware stable driver release, hand-compiled
      Mesa: From Xorg Edgers repository; it's "pretty recent" from git
      OpenGL renderer string: Gallium 0.4 on SVGA3D; build: RELEASE;
      OpenGL version string: 2.1 Mesa 8.0.1

      3d graphics performance in the guest is actually quite impressive -- for a virtualization solution. I really can't complain because it's light-years more stable and more usable than anything we've ever had before, as far as VMs go.

      Only problem that remains is the seemingly random micro-lags within the guest, where the entire guest will hang for extremely short periods of time whenever the host does something "interesting", which can be triggered by something as trivial as playing a youtube video on the host. These events are unpredictable to the point that they sometimes happen when I'm doing absolutely nothing except playing an audio stream on the guest; I'm guessing some background service on the host is causing disk I/O when this occurs.

      But... but... if RealNC gets the same problem with a *Linux* host, it can't be the Windows host! He says he even gets this problem with Linux guest Linux host, so Windows is totally out of the picture.

      So I'm confused.

      Comment


      • #18
        I use:

        CPU: Intel i5 2500K
        Mobo: MSI MSI P67A-C43 B3
        RAM: 12GB (8GB 1666 + 4GB 1333, both running at 1333)
        HD: Western Digital WD5000AAKS-00V1A0
        FS: EXT4 (and disabling barriers doesn't help)
        Audio: Asus Xonar D1 (PCI)
        GPU: GTX 560 Ti
        Only one monitor connected through DVI-D at 1920x1080

        To the guests I give 2GB RAM and an 8GB HD. 3D disabled/enabled doesn't matter.

        Might be CPU related. A while ago when I was using a Core 2 Duo, I/O blocking wasn't that bad. But I was using older software too then, so I can't be sure.

        Comment


        • #19
          Originally posted by RealNC View Post
          I use:

          CPU: Intel i5 2500K
          Mobo: MSI MSI P67A-C43 B3
          RAM: 12GB (8GB 1666 + 4GB 1333, both running at 1333)
          HD: Western Digital WD5000AAKS-00V1A0
          FS: EXT4 (and disabling barriers doesn't help)
          Audio: Asus Xonar D1 (PCI)
          GPU: GTX 560 Ti
          Only one monitor connected through DVI-D at 1920x1080

          To the guests I give 2GB RAM and an 8GB HD. 3D disabled/enabled doesn't matter.

          Might be CPU related. A while ago when I was using a Core 2 Duo, I/O blocking wasn't that bad. But I was using older software too then, so I can't be sure.
          Our configuration is very dissimilar. Except: how many cores are you assigning to the guest?

          I'm gonna try something later on, and start with 1 core, test, and increment the cores to 2, then 4 then 8 and see if the behavior changes. It may have something to do with the virtual IOAPIC, which is required if you have greater than one core.

          Comment


          • #20
            Originally posted by RealNC View Post
            You have a RAID setup, I only have one disk. You use an old kernel, I use 3.2.
            He's using Linux 3.1

            The version reported in Fedora 15 is 2.6.41 so that they don't have to make patches for everything that expects that kind of versioning.

            Comment


            • #21
              Originally posted by allquixotic View Post
              Our configuration is very dissimilar. Except: how many cores are you assigning to the guest?

              I'm gonna try something later on, and start with 1 core, test, and increment the cores to 2, then 4 then 8 and see if the behavior changes. It may have something to do with the virtual IOAPIC, which is required if you have greater than one core.
              I assign just one, but I also tried two, three and four. It makes no difference.

              Comment


              • #22
                Confirming the issue. Guest performance suffers when the host is doing disk IO. This is most noticable in the form of audio stuttering. On a windows7 host (workstation 8.02), all of my guests have the same issue. Guest OS's range from windows 98 through windows 7, Redhat 7.3 through Fedora 16, and OSX 10.5 through OSX 10.7.3. The issue is worse when IO is occurring on a disk that is connected to my Jmicron controller than it is on my intel controller. Updating the host Jmicron driver does not resolve the issue.

                I generally use guests for non-audio-related automated/scripted software testing, and as sandboxes. The issue doesn't cost me any productivity.

                Bench system is an old i5 750 on an Asus P7P55 Deluxe.

                Comment


                • #23
                  Originally posted by russofris View Post
                  Confirming the issue. Guest performance suffers when the host is doing disk IO. This is most noticable in the form of audio stuttering. On a windows7 host (workstation 8.02), all of my guests have the same issue. Guest OS's range from windows 98 through windows 7, Redhat 7.3 through Fedora 16, and OSX 10.5 through OSX 10.7.3. The issue is worse when IO is occurring on a disk that is connected to my Jmicron controller than it is on my intel controller. Updating the host Jmicron driver does not resolve the issue.

                  I generally use guests for non-audio-related automated/scripted software testing, and as sandboxes. The issue doesn't cost me any productivity.

                  Bench system is an old i5 750 on an Asus P7P55 Deluxe.
                  Hi,

                  Do you have a support entitlement with VMware? Mine expired. If you have an entitlement, please submit a ticket. They WILL get back to you -- they're very good about that.

                  I will buy the $49.00 single-incident support and submit a ticket if you don't have a support entitlement. This needs to get fixed.

                  Comment


                  • #24
                    To anyone following this thread for the VMware guest lag problem, take a look at this thread on VMware Communities: http://communities.vmware.com/thread/394250 (the threads are now cross-linked so hopefully we can combine brain power with people who've been using Workstation for years, and VMware employees too).

                    Comment


                    • #25
                      Do you have a support entitlement with VMware?
                      Yes and no. I was entitled to VMWare support up until the moment I unlocked the OSX guests. In order to file a support ticket, I would have to reinstall VMware, and I honestly do not have the initiative to undertake something that may results in tonight's batch failing.

                      Side Note: The app I am working on is a context-sensitive aspect-ratio convertor. 4:3 video is converted to 16:9 using every trick in I can find, including the context gained from panning/zooming. The current issue is figuring out which techniques to use (in combination), and when, and what can be done in real time. The neat thing is that the same techniques can be used in reverse (16:9 -> 4:3) to fill in the top and bottom bars.

                      Comment


                      • #26
                        Hey, a fourth person on the VMware forums confirmed he is experiencing similar symptoms: http://communities.vmware.com/message/2005634#2005634

                        This isn't just us imagining it, folks. We need to make sure VMware knows about this and is taking steps to rectify it.

                        Thanks everyone for participating. I feel like we are making progress on the issue. It's up to VMware to fix it, but if they don't know about it, it'll never get fixed.

                        Comment


                        • #27
                          Are you guys sure it's just the vm freezing? From my experience with Intel hardware the entire machine will freeze while doing intense I/O even without running any virtualization software. The most obvious case is while unpacking a large archive in the background, like the linux kernel. Usually the audio does not skip, but all GUI applications become unresponsive until unpacking finishes. I'm getting this with a Core 2 Quad CPU and I noticed the same with the previous Core 2 Duo machine the company I work for provided.

                          Comment


                          • #28
                            Originally posted by Ansla View Post
                            Are you guys sure it's just the vm freezing? From my experience with Intel hardware the entire machine will freeze while doing intense I/O even without running any virtualization software. The most obvious case is while unpacking a large archive in the background, like the linux kernel. Usually the audio does not skip, but all GUI applications become unresponsive until unpacking finishes. I'm getting this with a Core 2 Quad CPU and I noticed the same with the previous Core 2 Duo machine the company I work for provided.
                            I don't see something like that. And I don't need intense I/O to see the problem in VMware either. All I need to do is enter "sync" in a terminal and the VM freezes. And it does it *every* time, even if "sync" completes in 0.2 seconds or such.

                            Comment


                            • #29
                              Originally posted by Ansla View Post
                              Are you guys sure it's just the vm freezing? From my experience with Intel hardware the entire machine will freeze while doing intense I/O even without running any virtualization software.
                              While I cannot directly answer the question, I can state that the symptoms (audio stuttering, performance degradation under moderate IO) do not exist on the host. We've all tried a number of quick fixes to resolve the issue, including driver updates, bios changes (enable/disable VT-d), etc. The only clue that I can give is that high pci-latency exacerbates the issue (IO on the JMicron controller produces more audio stuttering than the onboard ICH9).

                              Reproduction is simple. Open a VM, open youtube, play a video that contains audio, kick off a file transfer on the host, Liss..ss...ss..ten to the audio.o..o.(pop).o.

                              F

                              Comment


                              • #30
                                Originally posted by russofris View Post
                                While I cannot directly answer the question, I can state that the symptoms (audio stuttering, performance degradation under moderate IO) do not exist on the host. We've all tried a number of quick fixes to resolve the issue, including driver updates, bios changes (enable/disable VT-d), etc. The only clue that I can give is that high pci-latency exacerbates the issue (IO on the JMicron controller produces more audio stuttering than the onboard ICH9).

                                Reproduction is simple. Open a VM, open youtube, play a video that contains audio, kick off a file transfer on the host, Liss..ss...ss..ten to the audio.o..o.(pop).o.

                                F
                                You hit the nail on the head. Nice description of the problem and test steps.

                                Comment

                                Working...
                                X