Multi-monitor survey v2 – The results

(The data and analysis presented in this post are fully released under the Creative Commons CC0 license. Basically, you can do whatever you want with it, just don’t blame me for your personal conclusions and don’t sue me if something wrong happens due to this data :))

Like for the previous survey, there aren’t much answers anymore after about one week, so I think it’s time to thank everyone and close the survey, in order to publish the results along with some interpretation.

Here is a PDF copy of the survey. Branching rules were that only people who have a multi-monitor main display go through pages 2 to 4, people who don’t use it with seamless mode skip page 3, and only people who have a “slave” display go through page 5. For results, see below…

Here you get raw data and an overview of the results. A few words of warning about this overview, though…

  • On questions with checkboxes (multiple choices allowed, tick whatever feels right), PollDaddy counts people who ticked nothing as people who skipped the question. Also, the percentages displayed are a percentage of answers, not of users who ticked the box. All in all, percentages for these questions are off and not to be trusted.
  • As mentioned above, I’ve made the survey so that it automatically skips some questions, related to what people don’t own or don’t use. PollDaddy counts people who didn’t answered those questions as skipping them, too.
  • For the “Likert” questions 8 and 11, there are several issues, which are all related to the fact that I could not find a proper way to introduce the “I don’t use this OS” choice, and so introduced it as a regular choice of numeric value -1 (other choices have a value between 1 and 5). Result is that for Mac OS X, which is not much used among the sample of respondents which have multi-monitor setups, the “I don’t use this OS” choice totally crushes the other answers in the visualization, and that the “Average” column is meaningless.

So now, let’s analyze the answers to this survey :

  • There are 683 respondents in total, which means that integer percentages are to be considered within a margin of error of 1%. The sample is big enough that rounding effects should not bring significant uncertainty to the results. Respondents come from OSnews, the OSdev forums, and this blog, so I effectively study the behaviors of computer literates (OS geeks) : do not consider that these numbers are representative of the average Joe, but you can probably assume that people knew the answer to the relatively technical questions I asked and could most of the time figure out what I’m talking about.
  • On question 1, I was highly surprised to discover that only less than half of the respondents (45%) claim to have used an external display, which is in fact significantly less than the amount of respondents who have used a main display made of multiple monitors (65%). I thought that this practice was significantly more widespread. Similarly, the 8% of the “dumb terminal” choice sounds quite low, considering the kind of audience we’re talking about.
  • Question 2 reveals some problems (thankfully, not so many) with the wording of question 1, as people frequently add as “other setups” some stuff which should fit in question 1’s categories (in particular, the first choice should have made it more clear that although the monitors, together, are the main way to display information, they can be managed separately in software, e.g. with one virtual desktop per screen). Interesting exceptions are :
    1. Setups made of multiple independent computer, like those provided by the Synergy software
    2. “Extremely high-res single displays that are addressed as multiple independent screens”
    3. Multi-seat setup, which is indeed slightly different from a server with dumb terminals, as it is a local construct
    4. Using the optional monitor as a main display when it is plugged in (particularly for laptops)
    5. Touch screen monitors used as programmable input devices.
    6. A software setup which I didn’t know of, where the desktop is extended along multiple screens, but full-screen applications, like games, only fill one screen.
    7. Braille main displays for blind people (definitely goes beyond the scope of this survey, though, in my opinion)
  • Question 3 reveals that when combining multiple displays, a line shape is by far the most common choice (which means that configuration software should optimize it), although a solid 16% of people use something more complex.
  • Question 4 shows that combining multiple displays is mostly an unplanned decision. 80% of respondents have put together unrelated displays at least once, 44% have always done it this way. Among the 20% of people who buy screens with the explicit intent of putting them together, and can as such make sure that they go well together, a significant portion (15% ± 1%) will replace them in an isolated fashion anyway. In this situation, although they can still make sure that the new screen fits *best* their multi-monitor setup, they can’t control the wild decisions of manufacturers, which jump from one trend to another and ditch whole aspect ratios and panel designs in the process.
  • Question 5 tells us that the “seamless” software setup, where content and windows can spread across the boundaries of screens, is extremely popular, as I suspected. 85% of respondents regularly use it, and 66% only use that. Questions 6 and 7, which solely study this particular software setup, tell us that the majority of setups are made of totally unrelated screens, with neither DPI not physical size as matching characteristics, which means that system software will have to abstract away a lot of differences for a perfectly smooth user experience. Considering the previous results about the origin of those setups, this is not exactly surprising.
  • Now, question 8, which reflects the satisfaction of users of each OS as far as combining multiple monitors is concerned, is an interesting one. Playing a few minutes with OpenOffice Calc allowed me to make an usable version of PollDaddy’s results plot, which does not include the “do not use” choice, uses normalized data (not number of users, but percentage of each OS’ user base), and is structured in a way that allow comparison of OSs.
    Overall, lots of people like the way their current OSs  does the job. “Good” is the most represented answer in all cases, and more than half of the users always rate their OS as “Good” or “Excellent”. “Okay” always gets roughly a quarter of the answers. As far as per-OS differences go, Unices are strongly viewed as an inferior platform, as they only get about 50% of their user base rating them above “Good” while Windows and Mac OS X get a bit more than 65% to do so. They also get much more “Bad” and lower ratings. Graphics driver problems are mentioned as a major part of the issue in comments. Mac users tend to be much more enthusiastic about their platform, be it to classify it as excellent or terrible. For the “Terrible” ratings of OS X, the global menu bar and new multiscreen-hostile features of Lion are mentioned as reasons, but no explanation is given that would explain the unusually high rate of “Excellent” evaluations.
  • Question 9 was designed for people who wanted to rate other OSs, but as a whole it has mostly been used to comment on question 8. People frequently point out that the question does not discriminate between different versions and distributions of the OS, which is especially a problem of Linux, an OS which has a relatively heterogeneous ecosystem (though most distros just bundle some version of kde/gnome/xfce and xinerama). For others, I meant current releases (Windows Vista and 7, Mac OS 10.6 and 10.7) when there’s a difference, apologies for not mentioning that. Also, it is regularly mentioned that installing and configuring extra software such as Matrox PowerDesk on Windows and AwesomeVM and Xmonad on Linux can seriously improve the user experience. Don’t know how much all of this affects Q8’s results, but they should probably be used with some caution. Also, it seems that Solaris is good at multiple monitors.
  • Question 10 was about how people would set up “slave” displays, which are infrequently plugged in (large monitors, video projectors), in software. Here, “seamless” mode (although it is often not so seamless in this case), remains the most popular choice, but not as extremely as before : it has 47% of respondents going for it, with 35% for treating both displays independently and an honorable 19% for the least popular choice of having the secondary screen mirror whatever happens on the primary one.
  • For question 11, which reflects the satisfaction of users of each OS as far as setting up “slave” displays is concerned, here is what the same calculation as before gives as a result :
    While at first sight there seems to be strong similarities with the previous results in global rating distributions, there are some differences. To start with, there are much less enthusiastic people in either direction, as “Excellent” ratings all lose an absolute 5%, while OSX’s “Terrible” ratings also lose 5%, making them more on par with Windows’ and Linux’s scores. There are more differences between OSs in the higher scores, as OSX now gets more than 70% “Good” or above, Windows gets 60%, and Linux sinks at 45%. The “Okay” ratings of Windows and Linux strongly rise, to the point where Linux has now more “Okay” ratings than “Good” ratings. As for how these differences can be interpreted, perhaps the wider range of software setups illustrated by question 10 makes OS defects in one specific setup stand out more. Also, perhaps the inability of some OSs to remember the settings of unplugged monitors, highlighted in the previous survey’s comments, is more penalizing here.
  • Question 12 about other OSs got very little following and there are no substantial news in there.
  • Question 13 and above discussed how people would create a main display made of several physical monitors if they had full control on it, and question 13 discussed the geometry of that display in a fashion that’s identical to question 3. Although a slightly higher number of people (18%) would go into more exotic geometries, there is nothing newsworthy here.
  • Question 14 shows that “seamless” software handling of the display is likely to remain just as popular in the future as it is today, as 78% of people would still use it, facing only 22% of people who would want to treat displays in an independent fashion.
  • Question 15’s results are, as mentioned above, a bit garbled. True numbers are that for a perfectly seamless and comfortable experience, 41% of respondents think that it would be fair of the OS to impose matching DPI between two screens, and another 41% think that it would be fair to impose that screens touching each other have a common edge of the same length. All in all, a weak majority thinks that OSs should deal with the most heterogeneous multi-screen setups that can be built without a single issue. This is coherent with question 4’s result that multi-monitor setups are essentially built from an heterogeneous set of screen that weren’t bought to go well together.

The last question was an open call for comments on anything that respondents feel is important. This is the open question which got the most following, with 70 answers. Apart from what has been mention above, here are, in my opinion, the juicy bits :

  • This whole comment : “I want to choose what to do with the taskbar. Stretch it? Put it on screen A. Put it on screen B.
    If I loose one screen I need to be able to access any windows previously on that screen. Never should a window “disappear” and be put outside a screen (some applications remember their position and will not be relocated when restarted). Why should the OS even allow a window to be placed so you can’t move it?
    Further, everything should be just plug-n-play. I love Win+P in Win7 for switching between modes. Use something similar for multiple screen setup.
    Also allow for separate wallpapers. Oh, and Aero snap-like features should work on multiple screens as well. Make it work even on the border between two screens.”
  • Someone mentions the importance of portrait mode displays for real work. In practice, however, I think that with widescreen-optimized software (vertical toolbars and the like), this can be mitigated quite a bit. But still, being able to rotate a display doesn’t hurt, especially when one targets those devices with bundled accelerometers…
  • “As cheap as large LCD (and the like) TVs are now, I wonder if multi-monitors will be a thing of the past.” > Good question indeed. I think that at least the “temporary slave display” form still has a future, because there will always be a need for projecting the image of a tiny device on a huge screen (watching a movie from a cellphone, doing an oral presentation from a tablet or netbook, switching to a bigger screen for work, etc…). For permanent multi-monitor setups, they probably still have a future in the form of giant “screen walls”, but their future on small-scale personal computers can indeed be questioned…
  • “While I want tasks to be separate on separate monitors, I want to be able to move tasks between
    monitors, (by clicking and dragging).” > Pretty much what I currently have in mind… Seems at least like the best option when associating unrelated screens. But as another comment says it best : “Seamless/independent use usually depends on task at hand: If you have for example 2×2 monitors you might want to use it as a gaming/moviewall at one time and keep independent tasks limited to a single screen at for instance, office work. A nice observation here is that if content is stretched over multiple displays it usually occupies the majority of screen space on each of those displays.”
  • “Xinerama is very slow. The only feasible solution to having more than two screens on Linux without slow
    Xinerama is to use AMD EyeFinity with the binary driver.” > This could explain the low ratings of Linux a bit further…
  • Synergy and similar other software solutions unifying the controls of multiple distinct computers (like Distributed Multihead X or KVM switches) are obviously very popular, and there’s even a request for native Synergy support. No answer at the moment on my side, graphics are still too far away…
  • OS limitations are definitely not popular. I know that already with the votes, folks :)
  • “next time give us the possibility to go back and change our answers” > I’m not enough of a web developer to hack into PollDaddy’s system, and I don’t plan to ever be, but if you realized what a question meant after clicking next a few time it’s a survey design problem on my side. Apologies.
  • “This survey still ask useless questions… (And one that do not relate to the OS, like in Q.15, how the length of the screen is a limitation of the OS ?) What is important is the management of your windows, the workflow… not if the text show the same height, or with the same number of pixels to be displayed at the same size…” > I think that both are important. If you want windows that span across multiple screens (e.g. for games and movies), you want content that smoothly flows from one screen to another, as much as possible. Also, if your multiscreen setup forces you to move your window in a maze in order to get it from screen A to screen B, it is a workflow problem (which is why I’m all for letterboxing when physical screen size only mismatches slightly).
  • “Think you should have had some questions related to graphics cards.
    Do you use one quad head graphics cards, are all your graphic cards the same or are they a mixture
    The first survey had a lot of questions about monitors but that’s the last link in the chain. If your OS and hardware is setup then monitors become more of a preference issue then a performance issue. I had a Quad head card that did not work well in a asus mother board but was fine in a gigabyte. I now have two different cards – I had to set manual refresh rates under linux for a time but later updates seemed to resolve the issue but then I guess it’s what type of information you are after with your survey.” > In my opinion, it shouldn’t matter on which graphic card screens are. There are a number of screen, the user tells the OS how they are arranged relative to each other, and we know which graphic card they are connected to when we need to draw stuff. It should be all that matters.
  • “Apps should remember the display they were last on but should NEVER display completely outside the currently available desktop-area (Windows apps tend to save geometry and restore even if the second display is absent!).”
  • “I’m bonking your girlfriend.” > Hmmm… No, you’re not :P
  • “this survey is way too small and forget something very inportant: the window manager which dictate how we use the multi screen configuration. and what about virtual screen ? ( I have 32 of them and can’t miss a single)” > Guess this user didn’t tick the first two boxes in Q1, which significantly shortens the question list ^^ Anyway, for me the window manager is a part of the OS. I acknowledge that in the context of Linux, it is a controversial statement.

And that’s all for today, I guess… Thanks again everyone !


7 thoughts on “Multi-monitor survey v2 – The results

  1. aaa September 11, 2011 / 9:39 am

    I don’t feel like using two monitors in Ubuntu 11.04 is easy (almost all of the problems are caused by Unity) but it’s out of the box, and doesn’t require any magic. So I don’t exactly understand the comment about Xinerama? Maybe it’s a problem of AMD graphic cards? I have Nvidia+Intel (Optimus technology).

  2. Splinter September 11, 2011 / 2:31 pm

    Good job. Thanks for the opportunity to participate and comment on your surveys. :-)

  3. Portablenuke September 11, 2011 / 4:30 pm


    Fedora is the only distro that I’ve found that correctly sets up my two 24″ displays without extra work. Everything else has required me to mess with the binary Nvidia drivers.

    “Also allow for separate wallpapers.”

    That’s a DE problem not a multi-monitor problem. For instance, Xfce 4.8 has this ability.

  4. aaa September 12, 2011 / 9:01 am

    @Portablenuke – Ubuntu does that too (I had no problems with dual monitor setup with both Nvidia and Nvidia+Intel combo – there is an indicator where you can set all up), but some recent update to Unity in 11.04 broke it. I hope it’s fixed in 11.10.

  5. max0x7ba September 14, 2011 / 2:43 pm


    The problem with Xinerama, AFAIK, is that it renders the the entire desktop on every underlying X server. That is, if there are 3 displays each handled by a separate X server, the entire desktop gets rendered on each of the 3 X servers. This yields a very visible and annoying delay when redrawing, especially with Qt applications which apparently use a lot of small widgets.

    I was using NVIDIA Quadro FX with 2 GPUs and 4 DVI outputs (2 displays per GPU). Open-source NVIDIA drivers can only drive the first GPU of this board, so I had no option but to use NVIDIA binary driver in order to connect 3 displays. Moreover, because it had 2 GPUs there had to be a separate X server for each GPU (NVIDIA TwinView can only merge two displays into one X screen on the same GPU). Hence, I had to use Xinerama to merge several screens on different X servers into one big desktop. And Xinerama performance, as I mentioned, was rather poor, forget about full screen video on a 4-core Xeon or desktop effects with Compiz.

    Now I am using ATI Radeon with AMD Eyefinity Technology with 3 displays configured as one big screen on one X server using AMD Catalyst driver. Because this videocard has only one GPU and AMD driver stitches all displays together into one display underneath X server there is no need to use slow Xinerama. This setup works beautifully, plays full screen video easily, desktop effects with Compiz work great.

  6. Tom Novelli September 16, 2011 / 4:51 pm

    I unexpectedly got dual monitors just a few days ago when we were given a beautiful old 21″ Trinitron CRT. My wife’s got that (and loves it), and I’ve got both 20″ LCDs (cheap 4-year-old Samsungs), driven off a midrange NVidia card (one DVI, one analog VGA), using the open-src Nouveau driver, XRandR, Fedora Linux 15, XFCE 4.8 desktop. Love it.

    Your survey questions don’t quite fit. This setup allows seamless mode, which is occasionally useful, but maximize/fullscreen mode fills just one monitor, which is perfect. I like that the taskbar only uses one monitor, leaving the other for fullscreen apps. I mainly use the extra monitor for reference docs while I’m programming.

    XFCE doesn’t have much of a multi-monitor config GUI. You kinda need to use ‘xrandr’ in Terminal, or edit xorg.cfg. GNOME 2 (under Ubuntu 10.10) has a much better user experience for the terminal-phobic. I imagine GNOME 3 works well too, but it has other issues.

    I hadn’t used dual monitors since 1997, on Intergraph Unix workstations and then Windows NT 3.51 and 4.0. The Intergraph system worked beautifully. Windows did some stupid things, like putting modal dialogs (i.e. login) right in the middle of the split. I think Microsoft fixed those issues by 2000 or XP.

    My other computer is a 10″ netbook :)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s