When X-Plane checks X-Plane.com for updates, it calls the server with an identifier for itself that contains a little bit of information about the host machine it is running on: operating system type and version, whether it is the 64-bit or 32-bit version, and whether it is running as a demo.* (X-Plane does not send any personally identifying information about you, but the server can see your IP address because all servers can see the IP addresses of all incoming network traffic.)
I sometimes get asked by third party developers: what percentage of users are using 64-bit, or what percentage of users are on Mac or Linux. So I wrote a script to analyze the incoming data and break it down by platform, etc. Here are the results. (I have done this kind of analysis before, but this is the first time I wrote a good script to remove possibly confounding results.)
Platform Breakdown
The data set consists of 24,917 unique IP addresses that ran a non-demo, global X-Plane 10.25r1 in the last week of 2013. This excludes users who have the regional version, a demo, don’t have their DVD in the drive, are running an old version, or who don’t have net connectivity. So that’s a big enough sample to get good data, even though it’s only a fraction of the total X-Plane 10 copies sold. Here’s the platform break-down:
IBM: 65.7% APL: 32.2% LIN: 2.0%
This matches the number we’ve seen ever since FSX development was halted: growing market share for Windows. (We used to be 60-40 Windows/Mac back in the day). Since X-Plane 10 is selling better than X-Plane 9, I believe that what we are seeing is growth, and the growth is disproportionately among Windows users. The Linux share appears to have shrunk, but it’s hard to tell since past data wasn’t as carefully analyzed. (The highest percent I have ever seen for Linux was 5%, but from this data I estimate the error bars on old data might be +/- a few percent, so who knows.)
64-Bit Adoption
Here’s the rate of 64-bit adoption for each OS.
All OSes: 82.0% APL: 85.1% IBM: 80.7% LIN: 72.7%
It doesn’t surprise me that OS X has the highest 64-bit adoption – every Mac is running a 64-bit operating system and OS X has the least available address space. What does surprise me is that Linux has the lowest 64-bit adoption rate, since Linux users have had the strongest desire for 64-bit. (This desire is, I think, driven by the difficulty of setting up cross-64/32-bit operation on modern distributions.)
Operating System Breakdown
We don’t have a break-down of Linux distros or Kernels – the Linux version of the sim doesn’t report that, but we do have operating system versions for Windows and Mac. On Windows the _32 and _64 bit suffixes tell whether the user is running the 32 or 64 bit “edition” of the OS.
The numbers include the 64-bit adoption rate for that particular OS; naturally the 64-bit adoption rate on the 32-bit editions of Windows is 0% because those OSes can’t run 64-bit X-Plane. Fortunately 64-bit editions of Windows are becoming the norm – of the users running Windows 8, over 98% have the 64-bit edition. On OS X, every version of the OS is 64-bit capable.
Windows: 5.1_32: 1.8% (64-bit: 0.0%) 5.1_64: 0.2% (64-bit: 0.0%) 5.2_64: 0.1% (64-bit: 100.0%) 6.0_32: 0.9% (64-bit: 0.0%) 6.0_64: 0.8% (64-bit: 77.8%) 6.1_32: 4.1% (64-bit: 0.0%) 6.1_64: 66.0% (64-bit: 87.2%) 6.2_32: 0.4% (64-bit: 0.0%) 6.2_64: 25.7% (64-bit: 87.3%) OS X: 10.6.5: 0.1% (64-bit: 42.9%) 10.6.6: 0.0% (64-bit: 100.0%) 10.6.8: 8.8% (64-bit: 78.9%) 10.7.0: 0.1% (64-bit: 28.6%) 10.7.2: 0.1% (64-bit: 100.0%) 10.7.3: 0.0% (64-bit: 100.0%) 10.7.4: 0.1% (64-bit: 81.8%) 10.7.5: 11.3% (64-bit: 78.7%) 10.8.0: 0.0% (64-bit: 100.0%) 10.8.1: 0.0% (64-bit: 0.0%) 10.8.2: 0.4% (64-bit: 88.2%) 10.8.3: 0.3% (64-bit: 88.0%) 10.8.4: 0.7% (64-bit: 84.7%) 10.8.5: 15.0% (64-bit: 82.0%) 10.9.0: 8.0% (64-bit: 84.3%) 10.9.1: 54.7% (64-bit: 88.5%) 10.9.2: 0.2% (64-bit: 100.0%)
Hopefully this is useful for third parties in deciding what operating systems and platforms to support.
* This is a standard practice – the update check runs over HTTP, just like your web browser.
Does the USB dongle count as “DVD in” or “DVD out”?
I _think_ USB out for these stats…but the number of DVD copies is a lot higher than USB copies. It’s also a little bit moot — the pro market is its own market with its own profit margins and up-front costs. These stats are meant to give a picture of the consumer/home pilot/non-commercial/entertainment market.
You can get a “USB key” (it doesn’t contain the software) of X-Plane for the general users. Which means you don’t need to use a DVD.
I’m guessing that’s what the original poster was wondering about, not home vs professional/commercial usage
I am supprised to see so low usage amongs the Linux users. Curious the know what the reasons are, but I bet it is related to software compabilities. Personally I would love to change completely from Windows to Linux, but software gigants like Autodesk and Adobe prevent me due to studies and work.
Is it fear to say that X-Plane 10 is fully compatible with Linux? If so any prefered distros?
For what it’s worth, 2% is higher than the general Linux desktop percentage, at least according to Wikipedia.
http://en.wikipedia.org/wiki/Usage_share_of_operating_systems#Desktop_and_laptop_computers
(But then Mac usage is much higher than OSX vs Windows ‘in the world’.)
I’d say this: installing the sim, getting sound, graphics drivers, multiple monitors and joysticks to work on Linux is harder than on any other OS, and those problems are almost entirely out of our control. Desktop Linux distributions simply don’t consistently provide a complete set of services for games. Some distros do a good job, but the very fact that there are multiple distros is indicative of the fragmentation.
As I’m a long term Linux user now (comming from Win) I can tell that X-Plane runs best on Ubuntu.
Problems mostly occur with proprietary GFX drivers on some graphics cards. I have the feeling nvidia drivers are still a bit better than ATIs on Linux.
The low Linux share however must not prevent you from keeping up the Linux port of X-Plane. You would loose third party developers this way as geeks love linux, righto?
Cheers
Hi Ondrej,
Personally I use Ubuntu, but that doesn’t make the graphics driver situation easy. Sometimes the drivers bundled with the proprietary manager are old, and I have found manually installing the nvidia driver on Ubuntu to be nightmarish. The AMD driver sometimes requires a particular Xorg version, which may not be the one in your Ubuntu distro, which is also fun. The situation is not as simple as Mac (the driver is just there) or Windows (get installer from NV/AMD, double click, go home).
We are definitely maintaining the Linux port to the end of the version run – we’re not going to remove supported systems mid-version. We’ll re-evaluate Linux for the next major version just like we have for every major version for the last few years.
The point of this post is -not- “we want to drop Linux” – but it is “if you make an add-on, here are the percentages of the total X-Plane customers you reach by supporting this sub-set of operating systems, versions, etc. The decision isn’t just about Linux, it’s about supporting old versions of OS X, older versions of Windows, 32-bit vs 64-bit, etc.
Cheers
Ben
I don’t think the proprietary nvidia graphics (using the NCurses based installer package from NVidia’s site) are that much or a nightmare to install.
One only needs to do three things:
* Make sure you have the compiler tools installed (ie: apt-get install build-essentials)
* Make sure the development headers for your kernel are installed (most of the time they will be anyway).
* Stop “X” (you can do this by stopping via your WM’s init script): eg:
/etc/init.d/gdm stop
/etc/init.d/kdm stop
service lightdm stop
Log in as root (CTRL-ALT-Fx to get at tty), run the installer. I have rarely had a case where the installer fails, NVidia keep up well with Xorg and Kernel versions (rare I have had a kernel blob fail to build).
You get the latest drivers that way. Installing from the repo’s are not hard either, but the drivers are older.
I have had far more drama with AMD drivers than NVidia. If you are on Linux, NVidia is totally the way to go.
I can see why it might be nightmarish for a Linux beginner installing the proprietary drivers, but there is a learning curve with any operating system. Once upon a time, the frustrated Linux newcomer had to learn Windows too, or MacOS and they were frustrated then.
Well, I think you have described the ‘gap’ between Linux and the other OSes pretty well here: installing the latest drivers requires having compiler tools, killing the window manager, and using the command line.
I have never managed to get the kill-X-and-install process to work right. In the end I made a calculation that it wasn’t worth the time to change driver sets; AlpilotX runs on NV on Linux, so I get good reports from him when something goes wrong; I stick with AMD, which I have been able to install.
Ubuntu should be the best choice regarding easy installing, drivers etc.
For nvidia or AMD graphics, you simply have to activate the official drivers.
Just give it a try, doesnt cost money 😉
See my previous comments to Ondrej — unfortunately we sometimes need a particular driver version to get a fix on the Windows side, and the versions packed into Ubuntu are often not that new.
Wonder if anybody has taken this issue up with the Linux kernal developer community?
The percentage may have decreased, but it’s possible that the absolute number may have gone up. XP10 by itself is linux friendly but there are various 3rd party devs who don’t support it, most likely because the small market share isn’t worth the effort Ben is describing (e.g. skymaxx pro). And it can be tedius. Getting my Saitek rudder pedals working took some digging that the average Windows user would never be able to do.
It will be interesting to see how the linux gaming landscape changes w/ the cash injection that will be SteamOS… And XP10 does work on the SteamOS beta. Quite well in fact…
I certainly hope that SteamOS will set a new standard for being ‘game ready’ (meaning having the problems of audio, graphics and input hardware be solved by the OS distro completely) since it is being built by gamers for gamers. That might help the other “friendly” distros (e.g. mint, Ubuntu) to figure out what they need to do.
Re: pedals, the thing that X-Plane _users_ can do to make this situation better is to file bugs and submit upstream patches to their distros re: hardware. Simply put, it is screwed up that you can get direct access to some joystick hardware via dev/jsXXXX but not via dev/eventXXXX. The whole point of the event API was to provide a modern hardware interface to overcome the limits of the old js interface (whose limits I could rant about at length), but if eventXXX is only available as root, that’s a failure.
Indeed, I had to create a udev event rule specifically for that device. The odd bit is that the joystick comes up without a problem.
Right – the existing udev rules don’t help make the event interface available for games, which is not so good. These things need to get pushed “upstream” for Linux to get more friendly, but this isn’t something I can do myself.
Looking fwd to your blog on 10.30 !!!
My take on the lo 64bit usage on Linux is that to install X-plane on Linux you need to use the 32-bit binaries. Thus, when installed, why go through the hassle all over again? (The hassle for me comes down to running 64-bit OpenSUSE and then needing to install heaps of 32-bit binaries to get X-plane of the ground, and then being able to install and get 64-bit X-Plane running).
If only X-plane 10 had been 64-bit from the start! There is so much hassle I would been spared from. Still, I am really glad X-plane runs and is supported on Linux 🙂 My enthusiasm for this have converted two people form MSFS to X-Plane!
Now, my next wish is for Laminar and others to focus on and contribute to, the use of free graphics drivers on Linux 🙂
Come to think of it, will the X-plane on Linux share rise with Steam-machines out in the wild?
Hi Jo,
True – you do have to get multi-arch to install. I think with the next complete issue of apps (e.g. 10.30 for apps, 3.30 for the installer) we’ll issue 64-bit binaries on Linux; that should eliminate the need for multi-arch entirely, which will make life easier.
Re: open source drivers, this is unfortunately simply never going to happen. It’s simple business.
– As you can see from the numbers, Linux represents a tiny fraction of our user base.
– We have many things we can do with graphics developer resources (that is, my coding time).
– We cannot sacrifice improvements to the entire rendering engine to work on just a Linux-specific feature.
In particular, one of the reasons why Linux is possible is because the proprietary NV and AMD OGL stack on Linux share code with the Windows drivers. This means we see the same issues on both, and thus we don’t introduce an additional OpenGL stack by porting to Linux. Our dev costs go up with each stack we support, so introducing another stack isn’t something we can afford.
For what it’s worth, a very knowledgable Linux user thoroughly investigated Mesa/Gallium compatibility with X-Plane and found:
– Some issues are with X-Plane, and are violations of the OpenGL spec (that happen to work on Mac and Windows). I _am_ fixing these issues for 10.30. The user did such a good job of organizing the report that I can’t help but respect his work and fix our code (which was wrong, albeit in a way harmless to Mac/Windows).
– Some issues were with Mesa/Gallium itself and he reported the issue directly to the driver team, which is great.
I don’t have time to chase down open source driver issues on Linux myself, but if someone comes forward with that kind of effort, I can do what I can to cooperate with it.
cheers
Ben
The reason I don’t use the 64 bit version of X-plane for mac is due to this one little plugin, Screen Waster. It is a simple plugin that prevents the screen saver from starting when running X-Plane. Screen Waster, works great in the 32 bit version of X-plane 10. However, it has not been updated to work with the 64 bit version. Maybe, future versions of X-Plane could include this feature.
If there is a standard Linux way to repress the screen saver that works on most distros and isn’t too weird, someone should send me a code snippet.
Nice statistics, but as normal user I don’t care much about it….I want to see good clouds, volumetric fog, nice scenery…all these things because of which I purchased X-Plane in the first hand….keep (not) working on this things and You will soon see how total numbers of users decreases. Strangely all these things were included in the 10.0x versions of the X-Plane and for some reasons removed in later versions…I almoust have feeling that I want my money back 🙁
We _are_ working on these things. I am not going to deprive third parties of information that they need to run their businesses because 10.30 hasn’t gone beta yet.
Given the current 64-bit adoption rate, and considering that X-Plane 10 has still many months to “live”, could it be reasonable to think of a future X-Plane 11 being 64-bit only?
It might be — we’ll make that decision when we get to v11 though, not now, and we’ll look at the numbers at that time.
Well, given that v11 will not be ready until late 2015 or even 2016 (my guess because you need to put some new stuff to appeal v10 users to make the shift and that takes time) we can easily assume that 95% of gaming OS will run 64 bit by that time.
Just so we’re all clear, these are Ricardo’s wild guesses on sim ship times – we (LR) have no comment and are not confirming or denying anything!!
Arghh…you easily unmask my laughable try to disclose future plan 🙂
I know many more clever veterans failed as well…
Interesting thought would on the numbers upgrading to the next version after a beta run?
This would give you a snapshot of how many actual users are using X-plane rather than having cull numbers and the various percentages on the adoption rates. A second run at the next version update is to see how those numbers have changed in users + or – over the period.
Would it help if us danglers (USB) put in the no1 disk when updating to add our data help in adding to the concensus?
The data includes donglers – I just didn’t include it for simplicity. My guess is that it won’t change the statistical picture.
A sitrep would be nice about where the v10.30 is at and what to expect in it, please. 😀
There’s nothing new to sitrep. X-Plane 10.30 is in internal development, and we are not pre-announcing any dates, schedules, or complete feature lists for it.
I don’t need/want to know when a 10.30 beta will ship and what it will ship with. It’ll ship when it ships with whatever is included. It’d be nice to get an idea of what is being worked on though. You already have a canned answer for questions of when and with-what.
what i find interesting is that of these 24917 users there are 18% more using mac 10.9.x (5047) than using win 8.x (4273.)
I have long believed that the unwillingness of many windows folks to upgrade their os and software was part of the reason microsoft pulled the plug on msfs–a huge number are still using 2004. It’s going to be interesting to see what happens to the x-plane windows numbers when XP11 does hit.
What I find surprising is that given that enough wide base of users, 3PD software houses do not start converting some amazing scenery for FSX into XP.
Once you have 3D textured builings developed from scratch (as most of new airports) and photoscenery already licensed, the step to convert into sdf should be quite fast. You could thus increase sell with a limited investment of time and money. But maybe my estimates are way overdone because I probably underrate the pirated market.
If even Aerosoft doesn’t pursue that path, it must be that the fraction of users willing to invest in quality payware addons is much smaller that what I can figure out (i.e. 30% ot 25000 xp users = 7500 X $30.00 [average scenery cost] = $ 225.000 which for me should be good enough to attract new or conversion projects). What is true for aircraft developers should turn true also for scenery devs, still I only read about a couple of announced XP projects among the many started for FSX.
Maybe the reason lies in that veteran FSX developers know little of XP engine and the time spent to train for a new system can deliver more money when spent in new FSX scenery.
We’ve sold a lot more than 25k copies! V10 has at this point (I think) outsold v9. A more interesting number might be update checks over a long time period to try to capture ‘active’ users. Over a single week the active user slice is kind of narrow.
Yes, I bet that. I took your 25k figure just to calculate a potential market. The more people actively flying XP the more I’m surprised of 3PD blindness.
Since XP addons market is way less inflated than competitors I guess you could expect the same profitability, even better in the future since the more I read forums the more I see people shifting to XP (as I did recently). Hence my disappointment to read few quality airport announced for XP.
Well, there are so many variables that are hard to estimate. I know total copies sold, but then there is the question of what percent will buy add-ons, at what price, and (as you point out) the saturation of the add-on market.
I think the primary issue is simple: third party developers have a complex tool chain that takes time to develop and targets a specific simulator. Redoing that tool chain is a huge up-front cost, so the easiest thing is always to keep doing more of the same thing you’ve been doing. (I think you don’t see a lot of XP devs go to the MSFS market either – for similar reasons.)
My thinking is that we’ll never lower that tool chain cost enough to make it unimportant … the only way to do so would be to precisely clone the MSFS third party scenery SDK – besides the risk of getting sued and the necessity of rewriting our entire engine, we would in the end end up with a visual product identical to MSFS…so why would anyone then use X-Plane.
Instead I think what we must do is increase the gradient from MSFS to X-Plane by continuing to add rendering quality, features, performance and SDK options to X-Plane that MSFS doesn’t have. This was one reason to do the global HDR lights in 10.0….it is a feature no one can add to MSFS, it makes a big visual difference, and it adds lots of opportunities for third parties to quickly make their add-ons look better.*
(I say quickly because I believe the time it takes to ‘light up’ an airplane should be significantly less than the time it would take to try to fake the same effect using render-baking.)
Could you please clarify your statement “It doesn’t surprise me that OS X has the highest 64-bit adoption – every Mac is running a 64-bit operating system and OS X has the least available address space. ”
1) how/why is the address space on Mac smaller than on 64bit windows/linux?
2) how is the size of address space connected to adoption, i.e. why are you not surprised?
Hi,
1. A 64-bit address space may have chunks reserved by the OS. In Mac OS X’s case, 512 MB are reserved for IO by the OS, and then another big chunk (I think it’s > 800 MB) is reserved for system DLLs.* I think Linux provides more total address space to the app but I am not 100% sure.
2. The lower the address space available to X-Plane in 32-bit mode, the more likely the app is to crash due to running out of address space with add-ons. So Mac users “hit the 32-bit wall” first. (Part of this is due to the Mac OS X OpenGL drivers allocating textures in the app’s address space, which eats address space fast.)
Just not using X-Plane on Linux because of the poor performance of ATI in my new laptop with the very recent Radeon 8000HD.
XP10(_64) runs pretty fine with my Radeon 6850 and the open source drivers. The new kernel 3.13 has a bunch of changes for recent Radeon GPUs which seems to improve performance significantly, maybe you can give it a try.