9 Replies Latest reply on Mar 11, 2011 12:34 PM by Jim_Simon

    Any benchmarks using Cineform?

    qseep Level 1

      This may not sound like a hardware question, but it affects what hardware I buy. Cineform claims vast speed gains with their codec without the need for expensive hardware. Yet they do not present any hard numbers to back up the claim.

       

      Does anyone have any hard numbers comparing performance when using the Cineform codec vs. other codecs? Ideally someone with a well-ranked system on the PPBM5 list could run the PPBM5 test but with all the source files pre-transcoded to Cineform. I have a dual-core laptop with no CUDA capability so I can't make a fair comparison. In fact I've tried using NeoScene and don't see any significant difference, but I haven't tried a rendering as complex as PPBM5.

       

      I realize there are some issues with this:

      • There are multiple Cineform products, such as NeoScene and NeoHD, which may vary in performance. I suspect that NeoScene simply provides the codec without accelerating rendering, while NeoHD accelerates rendering. But it's not clear from their marketing-speak.
      • You (as a potential tester) don't have a Cineform product. However, I believe it could be done with a trial version.
      • It involves some work transcoding and re-linking the source files in the benchmark suite.

       

      Does anyone have a clear idea of what kind of performance improvements I would see using Cineform on a newer system? Will it really help with rendering time? Is it good enough so that I don't need to bother with a fancy Nvidia card? Will I get better color reproduction due to the increased color depth in intermediate files?

       

      Thanks,

      Lyle

        • 1. Re: Any benchmarks using Cineform?
          Jim_Simon Level 9

          Without having read the actual Cineform literature, my guess would be that their claimed speed improvements come from the fact that their codecs are intra-frame only, and so are far easier to decode and play back than newer inter-frame codecs like AVCHD, which require some decent CPU horsepower.

           

          So basically, if you're system isn't up to snuff, Cineform can still make the media workable.  (Of course, you do lose whatever time it takes to transcode the originals to Cineform.  But on an underpowered system, it's a trade off you may have to live with.)

          • 2. Re: Any benchmarks using Cineform?
            qseep Level 1

            Well, it's not only that, but that they use wavelet compression, which they claim is easier on the CPU. On the Mac, they just use ProRes, so I suppose it's substantially the same thing. Which essentially gives you the same workflow as Final Cut.

             

            Thanks for your response, but I'm looking for some more nuts-and-bolts numbers. The question is whether I'd be wasting hundreds of dollars on a fancy video card or CPU when I could get substantially the same performance by using a $129 Cineform product. Provided I don't mind pre-transcoding.

             

            I think there would also be benefits to using Cineform as an intermediate output format from AE, when building complex animations for a timeline. Then there would be less generational compression loss.

             

            Which parts of the PPBM5 would be affected by using Cineform? By how much? Does it negate the usefulness of MPE? Would I get the same rendering performance by using Cineform with a cheap graphics card than I would using a fast CUDA-enabled card with AVCHD?

            • 3. Re: Any benchmarks using Cineform?
              Harm Millaard Level 7

              What you could do is convert all the clips in the ZIP file to Cineform format, but leaving all the properties intact, same resolution, same fps, etc.

               

              Then run the benchmark in its original form and repeat the tests with the Cineform clips replacing the original clips. That gives you the difference.

              • 4. Re: Any benchmarks using Cineform?
                Ann Bens Adobe Community Professional & MVP

                Not all formats in the zip are supported by CF.

                On converting the .mpeg NeoScene gives an error. It converts 1 second of what looks like a previous clip, the second clip is 5 seconds.

                 

                ppbm5.png

                1 person found this helpful
                • 5. Re: Any benchmarks using Cineform?
                  Harm Millaard Level 7

                  That looks like a typical case of EOS, End of Story, if CF does not support all formats. That means there is no way to get comparable results. Both these clips are HDV 1080i/25.

                  1 person found this helpful
                  • 6. Re: Any benchmarks using Cineform?
                    qseep Level 1

                    Thanks Ann, for your valiant attempt at testing Cineform. I agree with Harm; it is seriously concerning that Cineform can't handle this clip. I wouldn't expect it to handle any format you could throw at it, but HDV 1080i 25 seems particularly benign. They don't claim to handle PAL on their site, though they do list supported cameras: http://techblog.cineform.com/?p=2841. Furthermore, the fact that the transcoder would corrupt the file and return incorrect results, rather than simply reporting that it couldn't handle the file, is unacceptable.

                     

                    I'm quite skeptical that the Cineform codec would give me dramatic improvement in render times comparable to what CUDA acceleration will do. My own experiments have shown no noticable improvement in render time using Cineform, though I do have a rather underpowered machine in terms of CPU. (It does meet Cineform's minimum requirements.)

                     

                    It may be that the main benefit of Cineform comes in handling editing where multiple generations of effects need to be applied. It also seems suited toward cinema work where the 4:2:2 or 4:4:4 color space preserves more accurate color in the intermediate files. And it allows RED raw files to be compressed without significant loss of quality.

                    • 7. Re: Any benchmarks using Cineform?
                      jkosmicki Level 2

                      I use CineForm exclusively for rendering and down-converting Red 2K & 4K. The quality of the codec is the best in the business. Having said that, I would not suggest buying it in lieu of buying better hardware. It definitely does not give any speed advantages over any other format I use in Premiere, but it is absolutely real-time for me on my two systems and 10 bit video at roughly 12MB/s sure beats 120MB/s for uncompressed (my other option).

                      CineForm is for having a low data-rate, full-raster 10 & 12 bit, real-time solution with quality that is pretty much indistinguishable from uncompressed.

                      • 8. Re: Any benchmarks using Cineform?
                        qseep Level 1

                        That's great info, thanks!

                         

                        I always get skeptical when people say software is a "realtime" solution because it always depends on the hardware. And it depends whether you mean realtime playback of a single source, or realtime rendering, or playback of x number of overlaid tracks with any effects chosen from a certain list. No timeline will playback in full quality in realtime if you add enough tracks and effects to it. There is always a limit. OTOH if you couldn't play back uncompressed footage in realtime on your hardware, but you could when it was encoded in Cineform, without compromising on the image quality, that is a big win.

                         

                        I'll be editing HDV and AVCHD, and I can already play them back in realtime, on my meager system. What I want to do is accelerate rendering, and have the ability to play back more tracks and effects in realtime than I can now. For that, it sounds like Cineform won't help, but faster hardware will.

                        • 9. Re: Any benchmarks using Cineform?
                          Jim_Simon Level 9

                          I'd agree with that assessment.