Announcement

Collapse
No announcement yet.

NVIDIA/AMD OpenGL Benchmarks Of Unigine Valley

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

  • NVIDIA/AMD OpenGL Benchmarks Of Unigine Valley

    Phoronix: NVIDIA/AMD OpenGL Benchmarks Of Unigine Valley

    Continuing in the exclusive coverage of the yet-to-be-released Unigine Valley, here are some initial performance results for this visually-amazing multi-platform tech demo / benchmarks when using the OpenGL 3.2 Core renderer on Ubuntu Linux. A range of NVIDIA GeForce and AMD Radeon graphics cards were used for this initial testing of Unigine Valley. There's also benchmarks in this article of Unigine Heaven 4.0, which was just released yesterday.

    http://www.phoronix.com/vr.php?view=18476

  • #2
    Yeah Unigine Heaven 4.0 seems to be a lot slower than 3.0. Why?

    Comment


    • #3
      Interesting indeed. Here is my result: http://openbenchmarking.org/result/1...RA-GEFORCE6619

      I noticed that it uses a boatload of DOF. I don't remember it using it, at least to that extent, before, but then I haven't run the benchmark for quite a while now, so I might not remember correctly...

      Comment


      • #4
        Originally posted by d2kx View Post
        Yeah Unigine Heaven 4.0 seems to be a lot slower than 3.0. Why?
        I don't really think it's slower, they just added more post-process effects into the demo, like DOF. What's odd (and is going to be a source of confusion) is that the DOF [at least in the free version] is barely noticeable, and [to my knowledge] uncontrollable. There's not even warning telling you it's new or anything.. so to the unsuspecting benchmarker, it just appears that Unigine Heaven 4.0 runs half as fast as 3.0 for no apparent reason.

        Comment


        • #5
          Smaller resolution maybe?

          Michel can you please include smaller resolutions also in the test? I know higher resolutions stress GPU more but like me many people around the world don't have 1920 x 1080 monitors. If you see the Steam Hardware survey 2nd most used resolution is 1366 x 768 with 20.68%. Including smaller resolution helps to compare the test with our own PC. In your big comparisons, if you add a 800 x 600 or 1024 x 768, it would be awesome

          http://store.steampowered.com/hwsurvey

          Comment


          • #6
            Originally posted by Tiger_Coder View Post
            Michel can you please include smaller resolutions also in the test? I know higher resolutions stress GPU more but like me many people around the world don't have 1920 x 1080 monitors. If you see the Steam Hardware survey 2nd most used resolution is 1366 x 768 with 20.68%. Including smaller resolution helps to compare the test with our own PC. In your big comparisons, if you add a 800 x 600 or 1024 x 768, it would be awesome

            http://store.steampowered.com/hwsurvey
            768p is so big because it is a big de facto laptop resolution in the 13 - 15" space. I didn't even know they made desktop monitors like that. Unless you mean benchmarks for the laptop, but that would require him to benchmark mobile hardware too.

            Comment


            • #7
              Originally posted by zanny View Post
              768p is so big because it is a big de facto laptop resolution in the 13 - 15" space. I didn't even know they made desktop monitors like that. Unless you mean benchmarks for the laptop, but that would require him to benchmark mobile hardware too.
              Well at least in my country, monitors with 1366 x 768 resolution max is kinda common(18.5 inch monitors). Didn't know about the laptops though. And I am suggesting including some benchmark of 800x600 or 1024x768. That way all can compare those results.

              Comment


              • #8
                Originally posted by Tiger_Coder View Post
                Well at least in my country, monitors with 1366 x 768 resolution max is kinda common(18.5 inch monitors). Didn't know about the laptops though. And I am suggesting including some benchmark of 800x600 or 1024x768. That way all can compare those results.
                800x600 and 1024x768 is useless. We aren't living in the '90s anymore. If there was any testing with 1366x768 on Intel graphics I could see the usefulness.

                Comment


                • #9
                  Originally posted by LinuxID10T View Post
                  800x600 and 1024x768 is useless. We aren't living in the '90s anymore. If there was any testing with 1366x768 on Intel graphics I could see the usefulness.
                  Kind of dismissive, but in truth, if you have a 768p display, you have something else to invest in before a gpu upgrade.

                  Comment


                  • #10
                    A large number of laptops have 1366x768 displays,so upgrading isn't an option.

                    Comment


                    • #11
                      Originally posted by archibald View Post
                      A large number of laptops have 1366x768 displays,so upgrading isn't an option.
                      But neither is a GPU upgrade usually an option. You're stuck any way you slice it with mobile hardware.

                      Comment


                      • #12
                        Sometimes i think it is no good engine when it always runs with every option enabled. Sure for benchmarks you have to force it, but real games like Serious Sam 3 and most likely others have got a database and disable features for slower cards. Of course when you would test DX10/OpenGL 3 or lower hardware then you can not use Tesselation but thats not all. Games have to run fast out of the box without setting everything to minimum manually - for Unigine benchmarks i would say they should at least provide an option to use best suited defaults, the results are a joke - nobody could use that engine for a game with that preset.

                        Comment


                        • #13
                          Looks like AMD GPU performance on Linux isn't that bad after all (a 6770 is equal to a 550Ti on Windows at least, it seems to be equal too on Linux, but the 6770 was about 20-30 euros cheaper than the 550Ti when it was released).

                          I'd still wait for Haswell then use an Intel IGP for my next "upgrade" though, since Intel has good, official open source drivers and integrated GPUs make less noise/use less power than dedicated ones.

                          Comment


                          • #14
                            Originally posted by Calinou View Post
                            Looks like AMD GPU performance on Linux isn't that bad after all (a 6770 is equal to a 550Ti on Windows at least, it seems to be equal too on Linux, but the 6770 was about 20-30 euros cheaper than the 550Ti when it was released).

                            I'd still wait for Haswell then use an Intel IGP for my next "upgrade" though, since Intel has good, official open source drivers and integrated GPUs make less noise/use less power than dedicated ones.
                            In defense of dedicated GPUs, you can get a passively cooled 7750 that runs off PCI power for $100 that won't overheat under any workload at stock. Can't think of one that isn't 2 lanes wide. Though I can't argue that AMD has better FOSS drivers, because they don't. And I won't be surprised if Haswell has a good enough top end GPU that even thinking of a dedicated low end card isn't even worth it.
                            Last edited by zanny; 02-14-2013, 11:17 PM.

                            Comment


                            • #15
                              Intel is shooting themselves to the foot with that though: the fastest haswell editions, those with the embedded 512-bit ram on die (Crystalwell), will be mobile only and priced to high heavens. (source: semiaccurate)

                              If those were available on the desktop, and cost < 200$, they'd kick dedicated gpus hard. But as is it's like Intel doesn't want to dominate in that area on purpose...

                              Comment

                              Working...
                              X