5 Replies Latest reply on Oct 26, 2011 2:24 PM by Jim_Simon

    Question about benchmarking with Premiere Pro

    RobWilliamsTG Level 1

      Hi all:


      I am a writer for a website that focuses a lot on benchmarking PC hardware and am in the process of refreshing our CPU test suite. I'd like to implement Premiere Pro into our testing, but as a new architecture launch is due in a couple of weeks, I am working on this with some haste. I'm a novice with this tool, so it's my hope that I can get the "a-OK" from experienced users who know whether we're testing with the tool in a reasonable way. This is important because we implement real-world benchmarks for the sake of showing users of such tools how one CPU can compare to another in terms of raw performance. Other tools we benchmark with include 3ds Max, Maya, Handbrake and Lightroom (which I desperately wish supported more than a couple of threads).


      That all said, the project I've created is not conventional, as it's more of a worst-case scenario. I recorded about 18 minutes from a new PC game using Fraps, resulting in 35GB worth of source files. These are not "lossless", but are close to it (300~370Mbit/s). I take these files, crop most of them, add in cross dissolves and then encode it to MPEG 2 Blu-ray using almost default settings. In order to stress the CPU better, I enable the "Use Maximum Render Quality" (is that often used in production encodes?) option, and also change to a VBR 2-pass.


      Here are some quick results from the benchmarking I did this afternoon:


      Intel Core i7-975 (4 core, 8 thread, 3.33GHz) - 46m 40s

      Intel Core i7-975 (4 core, 8 thread, 4.00GHz) - 38m 45s

      Intel Core i7-980 (6 core, 12 thread, 3.33GHz) - 30m 36s


      As seen there, in an equal-clock 4 core vs 6 core battle, we saw the latter encode finish about 50% faster (the scaling is rather amazing). But the question remains - can I feel confident in benchmarking this as a project? Would it ultimately prove useful to you guys?


      Thanks much for any help!


      PS: As an added question, does Premiere Pro take advantage of SSE4 (or beyond) instructions, and will it support AVX in the future?

        • 2. Re: Question about benchmarking with Premiere Pro
          RobWilliamsTG Level 1

          Hi Jon:


          Thanks for the quick response!


          Generally speaking, I don't like to use pre-packaged benchmarks, but in the case of this one, I would consider using it as an addition but not a replacement of the project I mentioned. That benchmark executes rather fast, while I'd prefer a lengthier project that would do well to separate the men from the boys (with a benchmark that finishes in 1m on a six-core CPU, it doesn't leave much room for noticeable scaling).


          I am checking this out nonetheless, since it claims to use some of the GPU, which my project does not. Waiting for VB to install so I can run the script to generate the results...



          • 3. Re: Question about benchmarking with Premiere Pro
            Jim_Simon Level 9

            Personally I think the benchmark would be more telling with standardized media - specifically AVCHD.  Set up a 30 minute sequence using 1080p/24 media, and export out to MPEG2-DVD at the same frame rate.  Leave "Maximum Render Quality" on.  Have the project file on one drive, the media on a second, and send the export to a third.  (None of these should be the System drive with Windows on it.)


            I think this is still a pretty common workflow - shooting HD and delivering DVD - and it'll stress the CPU enough to show how well it performs.

            • 4. Re: Question about benchmarking with Premiere Pro
              RobWilliamsTG Level 1

              Hi Jim:


              Thanks for the response!


              I don't own a video camera (boy, do I ever feel like a reject, now). Would it be an inaccurate way of doing things if I took the same project I have as mentioned above, export it to AVCHD 1080p/24 and then use that as the source file to encode to MPEG-DVD? It'd still be game footage, but essentially the codec is what really matters (I could be wrong; perhaps the codec would work better with real images).


              Sorry if that comes off as a ridiculous question. If that's not a good solution I will see about getting such a source from somewhere.


              Thanks a ton!


              Edit: Would it be sufficient to use a Blu-ray .m2ts file specced at 1080p/24 AVC MPEG-4 as a source? I assume it's AVCHD given it's 1080p.

              • 5. Re: Question about benchmarking with Premiere Pro
                Jim_Simon Level 9

                Personally I'd rather not make the assumption.  If you start with raw camera footage, as most people use, then you have eliminated one possible variable.  And that's the key to good benchmarking.