Announcement

Collapse
No announcement yet.

SteamOS vs. Windows 8.1 NVIDIA Performance

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by ua=42 View Post
    Phoronix does not benchmark <insert game here> because it does not have a benchmark mode for linux. Phoronix does benchmark unigine, which is a modern game engine. So improvements in unigine's performance would mean improvements in Metro: last light, TF2, etc.

    The older open source games use more basic graphical features, so improvements in them will also translate to performance improvements. Until the open source drivers catch up performance wise with these older games, there is plenty of value benchmarking them. There are plenty of indie games that are on linux or a coming that won't be much more complicated graphically than these open source engines, so knowing how well the hardware can handle them is important.
    If you've done any benchmarking yourself you know this is complete bullshit. Performance improvements/regressions had in one title on an updated driver/kernel needn't translate into performance improvements/regressions on another title. This is a FACT.

    Another fact is that people don't care too much about precise reproducibility. They care about some rough idea of how a game THEY ACTUALLY PLAY performs on hardware roughly comparable to what they have. Add a disclaimer to the benchmarks saying that they were not automated, though most probably wouldn't even care! But just benchmark games people care about!

    Comment


    • #32
      Originally posted by tomtomme View Post
      and yes - it was all tested in opengl 4.4, as clearly stated in the box on the first page of the benchmark. And yes, a directx comparison would be interesting!
      Open Arena OpenGL ES
      Unigine Heaven OpenGL 4.0
      Unigine Sanctuary Opengl 3 (<- I'm not sure but i thought that Nvidia can run it with Opengl 2.x but Ati needs Opengl 3.x)
      Unigine Tropics Opengl 2.x
      Unigine Valley Opengl 4.0
      Xonotic Opengl 2.0 (<- The las time i check you even can disable 2.0 and go back to 1.3)

      So, no Opengl 4.4 yet, The box stated that the Nvidia driver 331.20 with a Gtx 760 does support 4.4 not that the test are going to run 4.4.

      The list above can contains some errors but i'm 100% sure none of the test are 4.4.

      Comment


      • #33
        Originally posted by tmpdir View Post
        Didn't expect this performance on an RT kernel, although never tested an rt kernel myself. Wil install steamos this weekend, curious how the game play 'feels'.
        Well, the RT Kernel would struggle a lot more performance wise if you had a lot of other tasks running in the background. With only a single app though, with a limited number of threads, the type of kernel and scheduler aren't going to have a major impact.

        Now, bring up two separate instances, run an encoding app in the background, and bench against Windows, and you'll get significantly different FPS/Latency based results.

        Comment


        • #34
          Originally posted by curaga View Post
          Playing back X input is not reproducible, because games have randomness. The enemies might behave differently, and so on, causing a different graphics load.
          So true, but so many people don't know it.

          Comment


          • #35
            Originally posted by molecule-eye View Post
            If you've done any benchmarking yourself you know this is complete bullshit. Performance improvements/regressions had in one title on an updated driver/kernel needn't translate into performance improvements/regressions on another title. This is a FACT.

            Another fact is that people don't care too much about precise reproducibility. They care about some rough idea of how a game THEY ACTUALLY PLAY performs on hardware roughly comparable to what they have. Add a disclaimer to the benchmarks saying that they were not automated, though most probably wouldn't even care! But just benchmark games people care about!
            Um wow. Chill. You are getting way too emotional over automated benchmarking and forum posts. Also, you might want to avoid using insulting language when posting.

            If you've done any benchmarking yourself you know this is complete bullshit. Performance improvements/regressions had in one title on an updated driver/kernel needn't translate into performance improvements/regressions on another title. This is a FACT.
            I guess I wasn't clear enough in my post if that is what you thought I meant. I will strive to improve my writing in the future. What I meant was the current selection of games Phoronix uses for benchmarking covers a decent selection of openGL commands/features/modes. Seeing performance improvements in these titles usually means performance improvements in other game titles as they use a wide swath of game related openGL commands. There are of course corner cases, where certain games will be bottle-necked on some command not found (or used very little) in these test samples. But as it is impractical to benchmark every game in existence, and these benchmarks are easy to automate and open source, using them is logical. I never said that performance improvements in one title was a perfect indicator in another.

            Another fact is that people don't care too much about precise reproducibility. They care about some rough idea of how a game THEY ACTUALLY PLAY performs on hardware roughly comparable to what they have. Add a disclaimer to the benchmarks saying that they were not automated, though most probably wouldn't even care! But just benchmark games people care about!
            If you haven't read the forum for previous benchmarks, you would know that Micheal doesn't benchmark anything he can't automate because the entire website is run just by him. And it isn't that profitable. Spending the time necessary to run each test manually is time he can't afford to spend. If you want benchmarks run manually, there are some of linux running minecraft on the minecraft wiki. http://minecraft.gamepedia.com/Hardware_performance

            With the upcoming steam box, there will likely be more of these manual tests that you so desire on other websites. They will most likely only use the binary driver, so if you are curious about the open source performance I don't know what to suggest other than paying Phoronix to do some tests manually.

            Comment


            • #36
              Originally posted by justmy2cents View Post
              don't know if it is just conspiracy theorist in me, but that paper was contradicting it self on all fronts like he wrote it half drunk and forgot what he wrote in previous sentence. just to name a few
              - directx coders know how to code, opengl... NOT (translation, if it is a car and red it is a sports car, if it is blue it is a truck)
              - all coders more optimize directx implementation, then says valve obviously didn't (how so?)
              - claims valve coded for dx9 and then graphs show almost same results for 9 and 11
              - coders optimize for dx11 rather than 9 (leaving me wondering if everything till 11 was a pile of crap)

              but, the claim where my alarm went ballistic was this
              "If we really want better OpenGL drivers, what we need is programmers that write better OpenGL programs."
              huh????? based that they wrote their benchmarks for OpenGL.... it leads to
              - benchmarks they provide are complete crap as there are no coders that would know how to code, god forbid optimize OpenGL
              - they are sole coders on this planet that know how to use OpenGL efficiently
              - wondering as last time i checked ps3 and ps4 for example don't have DirectX and since there is no coder who'd optimize all games for those run at 70%

              and i'm no linux'be'all evangelist. if it doesn't suit somewhere, it doesn't, if directx is better, why not. why would i care? i don't use windows from personal belief and i wouldn't even if they strapped 3 hookers on my win box for the same price as OEM.

              still, reading that was triggering same bullshit alarm as reading MS published reasoning why Server 2003 over linux. and then you look at hardware they used... absolute worst from NEVER, EVER use that in linux, where winner was Raid controller which worked at best 5MB/sec in linux. was really funny looking at results they got there/ i wonder how much network traffic was. maybe, 5MB?
              The writer of the article may be biased and speaking some BS, but the numbers can't be interpreted any other way (I actually didn't read the article itself, just read the numbers). There are also other articles comparing performance on Windows with OSX and Windows is also faster on that end. So, as long as developers continue to use DirectX on Windows then we need comparison against that.

              Comment


              • #37
                Originally posted by sarmad View Post
                The writer of the article may be biased and speaking some BS, but the numbers can't be interpreted any other way (I actually didn't read the article itself, just read the numbers). There are also other articles comparing performance on Windows with OSX and Windows is also faster on that end. So, as long as developers continue to use DirectX on Windows then we need comparison against that.
                Sure, it is of same value as comparing FPS between PS3 and Nintendo Wii. Dx is not platform agnostic, same with consoles. The only thing that matters is that developers create titles that perform well under target platform.

                If the users like target platform (and not its FPS), then developers will write and use this techology.
                Per your logic, users should lock-up into platform with more FPS, which is only dependent upon developers optimisation, which again depend upon market share (users preference).
                Your logic is either flawed or absent.
                Last edited by brosis; 17 December 2013, 02:53 PM.

                Comment


                • #38
                  DirectX vs OpenGL performance.

                  DirectX with games on windows usually perform better because the game engine usually (not always) has more optimizations for DirectX than openGL. If properly optimized, they will perform similarly (see valve's lfd port).

                  There are plenty of examples of unoptimized ports, I've personally dealt with NS2. In NS2 I can get 200fps with DirectX9, but with the OpenGL renderer I will only get 80-100 fps with several options turned down. I also get about 100fps with several options turned down with DirectX11 for the game. The reason is that these are new render modes for the game and they are currently both buggy and unoptimized, while the directX9 renderer has been continuously optimized over the last 4 years.

                  It is not the fault of the graphics drivers if the developer does a bad DirectX to OpenGL port.

                  Comment


                  • #39
                    Originally posted by iniudan View Post
                    Sadly good benchmark require reproducible test and commercial game on Linux side are lacking benchmark tool currently.
                    Half-Life 2: Lost Coast Benchmark http://store.steampowered.com/app/340

                    The reason it isn't tested is Larabel is overly dependent on automation. "reproduceability" is stupid easy to do by right clicking the game, selecting "Properties", hitting the "Updates" tab and picking "Do Not Automatically Update This Game" from the drop down menu.

                    Al he has to do is download the game once then back it up. Even if he left auto updates on it wouldn't matter since the valve games outside of DOTA2 and Team Fortress 2 are rarely updated. Those games are only because of the microtransaction wares that are constantly being added as both games are a cash cow for Valve.

                    Comment


                    • #40
                      Originally posted by schmidtbag View Post
                      if you already own the game, why do you care? If you don't own the game, a simple google search will tell you how it plays.

                      The problem at hand is suppose Michael supports something like Metro LL, the requests will never end. There will NEVER be a point where someone doesn't complain about a game not on his list.
                      Hardware upgrades.

                      Wrong, as new hardware and games come out you cycle out the older/less demanding games just as every other hardwre review site does. What this results in is maybe 5 new games a year replacing less demanding titles.

                      Said games can probably be had for free as well since the devs would want the free advertising of constantly being posted on a site that ranks fairly high in search engine results.

                      Comment

                      Working...
                      X