No announcement yet.

9-Card AMD Radeon Team Fortress 2 Linux Benchmarks

  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by DrYak View Post
    Huh... How did you "lock" the specific version of Team Fortress in place ?

    Note: I would be good if game maker and steam would come with a solution to keep a few older fixed version available, for benchmark-reproductibility version.
    Keep the data as 'delta' on the server to lower space usage.
    There's no even need to do the whole "LTS" route here. Just keep older version hanging around for benchmarking purpose.
    This isnt going to change. You're asking valve to change steam into something that it shouldnt be.

    Here is part of the problem, just my opinion of course, is that in order for Michael to get what he wants he'll need to distribute the tests through PTS much like how the OSS games are benched now. That isnt going to happen. He'll never get what he wants. Not unless he's willing to contact distributors and make deal with them.


    • #12
      Originally posted by Michael View Post
      Still doesn't do any good for when installing the game on another test system, when anyone else wants to go install the game to compare to my results, etc.
      Michael this is not the real problem because as opposed to FOSS games there are no major changes between different versions. Don't you have contacts inside the Steam staff to ask if/when they plan to release a benchmark tool for metro last light like the windows one?
      Another insteresting thing would be comparing graphic between DX11 and OpenGL. OpenGL doesn't seem at the same level of DX11, at least at first glance
      ## VGA ##
      AMD: X1950XTX, HD3870, HD5870
      Intel: GMA45, HD3000 (Core i5 2500K)


      • #13
        Originally posted by widardd View Post
        Although 63 fps on a HD5750 does looke nice, i'd be more interested in "minimum fps". I am using a HD5750, Athlon II X3, latest fglrx and performance is completely unstable and always has been for source-games. huge loading times, hicups all the time...

        Nice to see some benchmarks besides the usual OpenArena stuff.
        I agree also about the loading times, but the games are working fine for me.
        And I know Valve is working on the long loading times on Linux.

        Besides LFD2, the games run good on my hardware.
        HD5750 (1gig ram) AMD phenom II X4. 3.2ghz

        I use the latest Catalyst beta drivers.

        All Valve games do seem to suffer from some loading issue. It takes a bit before they run smooth.

        openSUSE 12.3 64 bit, kde with almost every eye candy option disabled. and Cool and Quiet, turned off in the bios.
        That last part, has to do with what Micheal found out, about the governor.
        This bios solution comes from croteam, the guys who make serious sam 3.

        That was months ago, but that article about the latest kernel being faster, reminded me about the governor.
        The difference in SS3 was nice, but the beta driver supporting openGL4, gave performance like on windows.
        Maybe even better but I am biased.
        ( pc is dual boot, openSUSE openlGL vs W7 DirectX )

        Why I am very interested in SS3 benchmarks.
        The developers of the game have stated the game should run a little bit faster on Linux.
        And told us, amd and nvidia were working on the drivers.


        • #14
          Benchmarking Minecraft?

          Now we just need a Minecraft benchmark as well, and the boys favorite games are covered

          Also looking forward to an Intel graphics Team Fortress 2 benchmark.

          BTW. Just upgraded to a premium account


          • #15
            Originally posted by Gps4l View Post
            And I know Valve is working on the long loading times on Linux.

            All Valve games do seem to suffer from some loading issue. It takes a bit before they run smooth.
            Do you know anything more about this? Because there is a loading issue in Expeditions: Conquistador as well, which results in multiple-minute loading times on Linux compared to nearly instantaneous loads on Windows. There is probably some function related to resource loading that works fast on Windows and slow on Linux...


            • #16
              Originally posted by tomato View Post
              yeah, because wanting to see benchmarks using something other that modified id tech engines and synthetic benchmarks is so unreasonable... doubly so for one of the most popular online games out there, completely crazy
              I'm not denying the usefulness of benching games like TF2, what I find annoying is how people act like it is the only test that matters, when as far as I'm aware Xonotic is more GPU demanding than TF2. I'm also tired of the chronic nagging at Michael regardless of how many times he has to tell people why he didn't use such tests.

              Synthetic benchmarks and older games might not be that useful, but the purpose of the benchmarks is to show progress between drivers. You don't need a popular game to do that.


              • #17
                Just a guess:


                • #18
                  Originally posted by schmidtbag View Post
                  Interesting - I'm surprised AMD overall performed better than nvidia on these tests.
                  Probably due to the fact that the latest Catalyst beta improves OpenGL performance massively as well as fixes the latency issues in Source games (especially and mainly TF2).


                  • #19
                    Interestingly enough, minecraft would make a good automated benchmark because anyone can have the client download x version of the game. They built in the ability to download old versions. Also, the game can be launched via command line. It is also one of the most popular games of all time and quickly heading for the number 1 spot, which means it is relevant to allot of gamers. It would be interesting to see how well it performs on the open-source drivers.


                    • #20
                      Yes, and Minecraft would also be a pretty good benchmark of both CPU, GPU and even Java performance. It could even be interesting to see some benchmarks showing the performance difference between different open source JVMs (OpenJDK, JamVM, Cacao) and Oracles VM.