On Windows and Linux 3D Performance

May 22nd, 2000 by

We’ve finished our latest article on 3D game performance, and this time we’re doing something a little different. We’ve tested a Matrox G400Max, an NVIDIA GeForce 256 DDR, and a 3dfx Voodoo3-3000 under Linux and Windows. Now, with these results we have a clearer picture of where Linux stands relative to Windows, especially for playing the very latest in 3D games.

There are some interesting results, and even a few surprises, in our tests, so click here to start reading.

14 Responses to “On Windows and Linux 3D Performance”

  1. x11quaker Says:

    This was an excellent article. I’d really like to see more stuff like this on this site. These are the kind of benchmarks you’d seldom see on other sites, and those that are considering Linux for gaming will probably take an interest in articles like this.

    Those that are considering the purchase of an NVIDIA card for Q3 in linux should consider that the current drivers are very unstable. As mentioned in the article, it’s impossible to change video modes without the game crashing. It’s also impossible, in my experience, to get many of the mods like Quake 3 Fortress working, with my vanilla TNT2.

    Keep up the good work guys.

  2. Astro Says:

    I’ve been thinking of a card to replace this pci voodoo banshee, and it convinced me not to get a voodoo 3/4/5. I’ll get a geforceDDR thank very much.

    How unstable are the drivers?
    I’ve had the mod problem with my 3dfx also, so i doubt its the card.

  3. mallan123 Says:

    The following is a letter I wrote to the author of the article which addresses the open/closed driver issue. This is an important issue that I feel has not been given balanced treatment by the Linux community, so I wanted to post it in a public forum.


    Hi Matt –

    One thing that bothered me about your article was the obvious sidebar ‘Open and Closed’. This is a touchy issue, and I don’t see why it had to be brought up. It is a two-edged sword that cuts both ways. What happens if the closed NVidia drivers remain two steps ahead of the open source solutions? It would certainly damage the case for open source drivers.

    I’ve been playing various versions of Quake for ages with a Voodoo2, and unless I ran as root, it would always crash when I changed video settings.

    The latest 3dfx open source XFree4.0/DRI Voodoo3 drivers will crash Quake3 whether run as root or not when video settings are changed.

    I have an Elsa GLADIAC (GeForce2), using the 0.9-2 drivers. Quake3 can successfully do vid restarts if run as root, although it still crashes for me on a vid restart as a non-root user. In your article, you said that you suspected that there was not much of a difference between the latest drivers and the ones you used. In the changelog for the latest drivers is the entry:
    “Improved mode switching in quake3 (still see occasional crashes)”

    And what if it were actually a Quake3 bug? Q3 is closed source, too. I don’t think I’ve heard *anyone* complain about that. I wonder why?

    What were the ‘issues’ with Q3A that you were referring to with the 0.9-2 drivers? It runs just fine here. The only problem with it is the sluggish mouse response which is an XFree4.0 issue, not NVidia’s. Mouse performance with the V3/DRI driver is equally unacceptable.

    I am a strong believer in Open Source, but I also do not believe that closed source software is inherently inferior. There are plenty of examples of excellent closed source software (Quake3 is one), and there are plenty of examples of crappy open source software. I want to see as many open source drivers as possible, but I understand the NVidia’s reasons for not open sourcing their current drivers. Constantly bringing up this issue could end up damaging open source drivers. I do not want to see closed source drivers become the norm for companies with less engineering talent than the NVidia/SGI/VA Linux team responsible for the NVidia drivers.

    Your article was mostly fair in addressing the open/closed source issues, and I appreciate your attempts to keep it objective and balanced. But many open source zealots have a ‘filter’ that blinds them to weaknesses in open source software and prevents them from making objective comparisons.

    I am an OpenGL developer, and I have been closely following the progress of the alternatives under Linux. I have a GeForce2 in my main machine, a PCI V3 2000 in my old Linux box, and a G400 Marvel in my Windows box that I occasionally boot to Linux to check the progress of the Utah driver. Of the three, the NVidia drivers are by far the best. They are the most stable, they support all of the cards GL extensions, and they are the fastest. I do not believe it prudent to point out the closed/open aspects of driver development in light of this.

    Just my 2 cents,

    P.S. Next time you do a comparison, could you include the V3/DRI drivers in the mix?

  4. Temporal Says:

    Hi Matt. I noticed you are having trouble with your nVidia drivers. I suggest you upgrade to the latest version, which fixes the problem you mentioned. Oh, and there is no “issue” with Q3A. I am using the latest drivers with a GeForce 2, and Q3A works perfectly, even when I rapidly change resolutions. I assure you, nVidia’s drivers have no significant stability problems.

    Oh, and x11quaker, I have no problems running any mods. I have Q3F running just fine. It doesn’t show up on the mod list, though. You have to start it up manually. This is a bug in Q3A, not nVidia’s drivers. This gives us an opportunity to raise a question that will eventually have to be addressed: Would this bug necessarily be fixed any sooner if Q3A were open source? I mean, the game has been out for what, 5 or 6 month now?

    Those that are considering the purchase of a 3dfx card for Q3 in linux should consider that the current drivers perform very poorly.

    Don’t get me wrong. I spend most of my waking hours writing open source software. But the attitudes of some zealots really pisses me off.

  5. TMan Says:

    This is just great!! Now maybe its just my machine or I don’t have something setup quite right. But my g400max do not give even close performance in linux as they do with windows…and I don’t even use the TurboGL stuff. Maybe its time to re-compile from the CVS??

  6. linuXgod Says:

    Actually id like to see these benchmarks on X4.
    This V3 is hitting 100fps in X4. X3 is rather slow. The factor of DRI vs NON-DRI is being tested here.

  7. linuXgod Says:

    Its not linux vs winblows here also, its XFree86 vs winblows. GLQuake actually runs faster under the linux console than it does in win. about %30 faster. Linux is a faster operating platform and IS far supperior in GL, if more games where made for the console it would be faster, then we wouln’t have X to bloat the game down.

  8. jvmatthe Says:

    NVIDIA, and in fact all the driver writers involved, had an opportunity to read over the results of the article before it went to press. Their response indicated _no_ problems with my comments about my difficulties with their latest drivers. I’ve talked it over with them, and the feeling I’ve gotten is that I need to wait for an updated set of drivers because there is a problem. You’ll note that the release notes comment that there are still occasional crashes, so perhaps not all is fixed, no?

    Biased? As a matter of fact, our goal was to be as unbiased as possible on this piece, and contacting the driver writers was part of that. Also, we relegated the comment on Open vs. Closed to a sidebar, since it needed to be separated from the factual content of the article itself.

    As for XFree86 4.0 benchmarks…I’m working on it.

    Thanks for the comments, guys. It’s always good to hear what people think, even when I don’t agree completely.


  9. michael Says:

    … is not the OpenGL driver but the way to install it.
    What gamer knows about cvs to get the driver source? What gamer cares about file permissions of /dev/3dfx?

    And if you have solved all that problems…
    you will meet the DGA mouse problem.
    This one exists for a long time now in the X server.
    If you play Quake3 with “/in_dgamouse 1″ you will loose your mouse/keyboard control after some time. Logging in from another computer via telnet and entering perhaps “ls” or “ps” gives me back the control for some time…
    That is not exactly the way it should be :-)
    Of cause you can play without DGA mouse but your movement and aiming will be very bad (unplayable!!!).


  10. x11quaker Says:

    Does anyone have the older XFree86 4.0 drivers for NVIDIA cards, the ones that contained GPL’ed code?
    Those drivers worked better for me, and NVIDIA appears to have removed them from their site.

  11. greenhatman Says:

    In reply to the guy that wants the old nvidia drivers, I have them. Does anyone know how to get the mousewhell to work with XFree86 4.0?
    I already am using the imps/2 line in the XFree86.conf file. I have a Microsoft Intelimouse Explorer. Also anyone that says that the Nvidia drivers are unstable are wrong because the Q3 Demo runs fine on my Linux box and I haven’t had any problems with the proclaimed video mode switching in the Q3 Demo. I have a Guillemot 3D Prophet.

  12. Astro Says:

    option “zaxismapping” “4 5″

  13. lubricated Says:

    given two drivers with the same features and performance I think most of us would use the opensource drivers. Granted that this is not often the case but many of us want open source and there is absolutely no good reason why a company such as NVIDIA should keep their drivers closed.

  14. mallan123 Says:

    Re:NVidia – No problems??

    Try disabling AGP in your XF86Config by adding

    Option “NvAgp” “0”

    To the “Screen” section. That worked for me. I have a TNT and a GeForce2 and they have both been very, very stable.

    Re:closed source is inferior

    I don’t think anyone would argue that open source would be preferable, but they do have good reasons why they can’t open source the current drivers: They share the same codebase as their Windows drivers, and there is intellectual property in that code which they do not control. Please remember, however, that open source drivers for NVidia cards does exist in the Utah driver code base (although performance is nowhere near as good as the official drivers). Hopefully, work on these drivers will continue:

Leave a Reply

You must be logged in to post a comment.