8 Replies Latest reply on Dec 20, 2013 12:16 PM by Alex - DV411

    a simpler benchmark?

    Alex - DV411 Level 2

      How would one measure the effect of GPU acceleration without running a full PPBM benchmark? Is there a simpler benchmark perhaps?

       

      The question stems from the natural desire to quantify performance improvements in Adobe announcements. and in general to get a feeling how different cards stack up.

       

      Would manually measuring render times on PPBM project (with and w/o GPU acceleration enabled) do the trick?

       

      How does Adobe measure GPU acceleration improvements?

       

      (The PPBM benchmark is fantastic but rather complex: the two times I tried to install and run it, there were cryptic error messages and I gave up.)

        • 1. Re: a simpler benchmark?
          Peru Bob Adobe Community Professional & MVP

          Alex Gerulaitis wrote:

           

          (The PPBM benchmark is fantastic but rather complex: the two times I tried to install and run it, there were cryptic error messages and I gave up.)

          You must not have exactly followed the instructions.

          • 2. Re: a simpler benchmark?
            jasonvp Level 3

            Peru Bob wrote:

             

            You must not have exactly followed the instructions.

            I think it'd be nice to have a standard way to test the GPUs regardless of the platform.  PPBM has and probably will stay a Windows-only based testing solution.  The main reason for that is it's using Visual Basic (*GAG*!) though maybe that can change.

            1 person found this helpful
            • 3. Re: a simpler benchmark?
              ECBowen Most Valuable Participant

              The easy way is to create a project with a known good threading codec in Adobe that really pushes the system ie Red. Then scale that media, along with set GPU only acceleration plugins. That will insure you are truly pushing the GPU with your test. Then render the timeline for playback and then render to a set codec on export that threads well ie can be many options here. Just make sure you use the same codecs for this entire process. I would love to test the AMD versus Nvidia further but the only ATI cards we ever sell here are the 7750 cards on audio systems. No one ever buys an ATI on the video systems unless maybe on a Vegas system. So convincing the boss to buy a ATI card for testing only is not very easy

               

              Eric

              ADK

              1 person found this helpful
              • 4. Re: a simpler benchmark?
                Alex - DV411 Level 2

                So convincing the boss to buy a ATI card for testing only is not very easy

                I hear you. Perhaps finding someone on this forum with Premiere Pro 7.2 and a decent ATI card might be easier? I.e. if we standardize on a specific project (source files, project file), tell people where to download it, and ask them to time certain things and count dropped frames? Could this be done with PPBM source/project files since quite a few people already have it?

                • 5. Re: a simpler benchmark?
                  Bill Gehrke Adobe Community Professional & MVP

                  Jason Van Patten wrote:

                   

                  I think it'd be nice to have a standard way to test the GPUs regardless of the platform.  PPBM has and probably will stay a Windows-only based testing solution.  The main reason for that is it's using Visual Basic (*GAG*!) though maybe that can change

                   

                   

                  PPBM does not use vbs in any way to run the benchmark.  It does use a vbs script to get the results. About a year or so ago I built a Hackintosh and was able to use UNIX scripting to find the times of how long each generated file took to be produced.  But maintaining and distrbuting this UNIX script requires more time than I can handle.

                  • 6. Re: a simpler benchmark?
                    Alex - DV411 Level 2

                    Bill, I was hoping you'd chime in. What would be your recommendation re: simpler benchmark? Stick with PPBM?

                    • 7. Re: a simpler benchmark?
                      Bill Gehrke Adobe Community Professional & MVP

                      Alex,

                       

                      There is no reason why anyone would need anything other the two MPEG2-DVD steps of PPBM6 DE to find the gain of a GPU for a standardized test.  I have plenty of numbers of the results.  The only trick is after running steps 2 and 3 to get a GPU accelerated time and a CPU only time is the find a way to get the time from the creation of the file to the modification time.  This is part of the Statistics.vbs process for the a Windows PC.  I had to of course use the Terminal mode of the Apple UNIX machine to achieve the same results, which may be beyond some Mac users

                       

                      Here are some early PC GPU accelerated results.

                       

                      MPEG2-DVD-w-GPU-Encoding-2.jpg.

                      1 person found this helpful
                      • 8. Re: a simpler benchmark?
                        Alex - DV411 Level 2

                        There is no reason why anyone would need anything other the two MPEG2-DVD steps of PPBM6 DE to find the gain of a GPU for a standardized test.

                        Thanks Bill, I'll look into how simple this can be.