No announcement yet.

Unity 6.8 Doesn't Change Much For Performance

  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    nouveau is certainly more often problematic than radeon in current days. informer i had to fix myself a bug within mesa and some oem cards which are less popular (like x700se which showed a checkerboard first). nouveau however likes to crash the system often badly. also when used with llvmpipe the system likes to crash as well, no idea why (especially with pci-e 3.0 and kelper, did not happen that often with pci-e 2.0 and the same card). nouveau.modeset=0 is often needed, one card that is pretty popular and needed that hack with the default 3.2 kernel (precise) was btw. the nv gtx 550 ti.

    i do not understand why there is no fallback possible to work without compiz on ubuntu, i do not like the whole unity thing as well. marc hold a presentation a few years back at linuxtag where he showed slides how he wanted to have got the next ubuntu interface, the first thought was definitely that it was optimized for small netbooks or tablets. i still do not see the huge benefit of that ui for standard desktops. of couse the name implies that it should be used everywhere - a similar approach is done by windows 8 - both changes are not seen positive by everybody.

    btw. kde allows the use of ";" in desktop icons gnome 2 did not, no idea about gnome 3. i really like to use more than 1 command without writing a script.

    unity is a special case because no other distro ships it by default, so ubuntu has to fix all bugs basically on their own. i am no fan of gnome - maybe the first version was ok, but as soon as kde 2/3 was out i never used gnome. i tested gnome shell very shortly and i disliked the fact that i could not create desktop icons anymore. well there are dconf hacks but i speak of the out of the box experience. every DE has got its bugs, so basically have to select one that has got the less important ones for your typical workflow. if raw speed for games is important unity does not seem to be that good - but a real gamer usually has got a faster gfx card and it should not be important if a game runs at 200 or 60 fps - most tft are displaying not more than 60 anyway.
    Last edited by Kano; 10-07-2012, 08:40 PM.


    • #22
      oh sure, back in the day ubuntu was the best gnome distro and everyone loved it. now, they have serious issues. you can't make the excuse that at least it's not complete crap. if you went to someone's home and they offered you soggy old bread, would you say- hey, at least it's not scooped from the trashcan? I want Linux to be as good as the competition- for that they would have to fix up x/wayland and then pulseaudio.
      bitching like me helps Linux more by bringing problems into attention. By constantly ignoring the problems and showering them with praise does not make the issues go away.
      if you want to know I did submit a bug report and helped to fix a bug in nouveau. so i'm not just a passive complainer.


      • #23
        Saying that each version implements a regression based on DEVELOPMENT versions is completely rediculous. Currently, I have almost the exact framerate between Unity and LXDE when using Undirect fullscreen, which is planned to be default.


        • #24
          there was a time where i tested several nvidia cards with nouveau, for old cards with 1 performance level the speed was not that bad. but as dynamic reclocking does not work and the manual way often crashed the system i usually do not use it anymore. the geforce fx 5 series also does not support 1080p via dvi and therefore i have got no huge interest in those cards (there still exists the 173 driver series but they are not optimal for composite) and for everything newer there are much better binary drivers.


          • #25
            Originally posted by Kano View Post
            if raw speed for games is important unity does not seem to be that good - but a real gamer usually has got a faster gfx card and it should not be important if a game runs at 200 or 60 fps - most tft are displaying not more than 60 anyway.
            I dunno, man. I have a Core i7 and a Geforce GTX 570 and XBMC is a slideshow and tearing in games (and video) is so bad that I don't even play them to check performance. In fairness, tearing in GNOME Shell is a problem for me too. Not to mention the random desktop hiccups and bugs that generally make using Ubuntu incredibly frustrating and (to be blunt) amateurish. Canonical had no business releasing this thing in its current state. No professional company can release software that is this messed up.

            Valve is going to be in for a world of surprise if this stuff isn't fixed by then.

            I guess my frustration is that none of these problems existed in GNOME2 + compiz. Why we're moving backwards, I don't know.


            • #26
              The six month release cycle is a joke. Linux has to fix the abi, shared library hell so that devs can write apps for old releases. Right now only rhel, Slackware and Debian cut it in the stability. The rest are basically alpha products in terms of regressions. I think using bundles like chakra Linux or PBI like pcbsd is the way forward.


              • #27
                Originally posted by mayankleoboy1 View Post
                I would say that the 6 months release is a rabid release. The QA dont have enough time to bug fix a whole OS.
                QA is not tied in any way to the six months release schedule. 6 months release schedule only means you release whatever is ready every 6 months. The features themselves may have been under development or QA for years.


                • #28
                  As mentioned before, it's useful for gamers to enable unredirecting full screen windows by installing compizconfig-settings-manager and launching CompizConfig Settings Manager -> Composite -> Unredirect Fullscreen Windows.

                  I think the setting is aimed to be the default eventually, but as indicated by the last comment in bug 980663, there are currently especially driver bugs preventing it. Hopefully for 13.04 enough of the issues could be resolved.

                  When it comes to other compiz performance problems and the actual regressions in 12.10, the bug 1024304 is the most interesting one including the links it has in comments. It claims that actual compiz performance is better in 12.10 but benchmarks show regressions, and offers another workaround in addition to unredirecting fullscreen windows.

                  Finally, nvidia binary driver users are heavily affected by bug 1049214.


                  • #29
                    Unity is great and I wouldn't want to go back to old Gnome. With unredirect fullscreen windows enabled the performance is on par with other DE's.

                    The screen tearing existed on Gnome 2 and it's present in Gnome Classic with or without efffects, just like in every other user interface on Linux. It's here because it's the problem of underlying technology (X.Org X Window System).

                    Valve is perfectly aware of how the games will work on Ubuntu. That's the distro they will support and you obviously forgot about Faster Zombies! blog entry from Valve Linux Team.

                    John Knight
                    What desktop environment were your tests utilizing on the Ubuntu 12.04 32-bit test?
                    Valve Linux team
                    We are developing and testing on a variety of desktop environments with Unity being the most common. In addition, we are working with others to make sure that the Steam client and L4D2 run well in all environments. So far, we haven?t noticed any specific issues.


                    • #30
                      Originally posted by jrrr View Post
                      Right, so this is about graphics performance of applications running within the Unity environment - what about the performance of Unity itself? Does opening the dash, moving/resizing windows, etc. happen noticeably faster?
                      Yeah, I agree that that's what should have been tested. Even then, I don't know why Michael expected to find any performance improvements other than those for LLVMpipe. Looking at the list shows only one: Spread - Scaling all the windows is too slow (LP: #1055643). If he hadn't uncovered that Xonotic regression this article would be a waste of space.