10 Replies Latest reply on Jan 9, 2011 6:51 AM by Jeff Bellune

    About rendering differences CS5 vs CS4

    HarryPutnam Level 1

      I just did get CS5 installed, and from the excessive and high claims of  the `mercury' engine I expected to see something noticable in render times.

       

      I have not seen it so far and in fact may actually be slower.

       

      I am rendering on the timeline an AE project that is a semi basic introduction to some commentary taken at a wedding shoot to be included in the produced video.  Just some hopefully nice graphic introduction to interviews of people before the camea.

       

      The AE project that does it involves 2 layers, text and solid with some really basic text animation, and really primitive use of the `ramp' effect tool.

      Pretty basic stuff I'd think.

       

      Yet Premier CS5 seems not to handle it any better than CS4 did, and trudges along like atlas with the weight of the world on its back.  The AE part was done with AE CS5 so there is no verioning excuses.

       

      This is on a semi-substantial machine:  i7 CPU Q820 Quad core at 1.73 Khz, 8 GB ram.

       

      Premiere seems to take what appears to me to be more time to render this basic AE project, which I set up for rendering by setting the workspace ears and pressing enter, than does CS4... but nothing dramatically one way or the other.

       

      I was disappointed to see no change for the better, and now suspect if one really did some research on the alleged scientific breakthru phenomena of the  `mercury engine' one would find it isn't so very different or better than before.

       

      I guess I've grown as complacent as most in overlooking the lying claims that are allowed to advertisers and just figure it's par for the course, which is really too bad.  We shouldn't allow that kind of stuff to go on.  It is really kind of disgraceful that we put up with it although I guess it would be really difficult to police if a company were hell bent on exagerating.

       

      I don't mean to say that Adobe if the worst offender or such like but still...

       

      To tell the truth:

      I did absolutely no scientific comparision and in fact did'nt even try it side by side.   I already know it will be a very similar result and slower is likely I think, just from personal observation.

       

      All that said can others say they seem something remarkable in render times?

        • 1. Re: About rendering differences CS5 vs CS4
          Bill Gehrke Most Valuable Participant

          The "break through" is called the Mercury Playback Engine,  Its primary purpose is the assist in not having to render many Premiere effects in the timeline so that you can view them.  It is not a new encoding engine.  Hopefully my definition above is correct.  We should expect our faithful and very helpful Adobe employees to jump in here.

          • 2. Re: About rendering differences CS5 vs CS4
            Todd_Kopriva Level 8

            Show me the claims that were made that you have issue with.

             

            There are several performance improvements in Premiere Pro CS5.

             

            One of them is that some things are sped up using CUDA processing if you have a specific kind of graphics card. You didn't mention having one of these.

             

            Another improvement is the port to a 64-bit application, which helps with several things, including improving stability and making it feasible to work with very large frame sizes for footage (e.g., RED and other digital cinema sizes). You didn't say what kind of footage you're working with. I'm pretty sure that if it were RED or similar size you'd notice a huge difference.

             

            Another improvement is multithreading, which helps speed some things up on multicore computers, especially if you have a lot of RAM. You don't. People tend to report that the real advantages kick in around 12GB for Premiere Pro. For After Effects, we recommend 3-4GB of RAM per processor core, which is 12-16GB for a quad-core computer.

             

            Also, After Effects doesn't use CUDA processing, so if you're using Dynamic Link to After Effects, that's going to limit speed even if you have a CUDA card.

            • 3. Re: About rendering differences CS5 vs CS4
              Todd_Kopriva Level 8

              The "break through" is called the Mercury Playback Engine,  Its primary purpose is the assist in not having to render many Premiere effects in the timeline so that you can view them.  It is not a new encoding engine.  Hopefully my definition above is correct.  We should expect our faithful and very helpful Adobe employees to jump in here.

               

               

               

              The CUDA processing features do benefit rendering for output, too. They don't affect the encoding stage; but they do affect the stage right before encoding.

               

              Creating an image from a set of instructions and inputs is what "rendering" is. To oversimplify a little bit, the process of creating an encoded file from a sequence/timeline basically involves rendering followed by encoding followed by packaging. Encoding often involves compression, and that is a time-consuming and computationally intensive task.

               

              The use of the word 'playback' in the squishy name 'Mercury Playback Engine' is confusing and annoying.

              • 4. Re: About rendering differences CS5 vs CS4
                Harm Millaard Level 7

                If you look at the old benchmark results of http://ppbm4.com/Benchmark.html and look at the performance improvements of CS5 over CS4, you will notice around 30+ % increase in performance on the same system, just by moving from CS4 to CS5. That is a big improvement and that was measured with 5.0.0. Now further updates to 5.02 and 5.03 have given even more improvements, so the general conclusion can only be:

                 

                CS5 gives a much better performance than CS4.

                 

                In my own case the move from CS4 to CS5 dropped my results from 38.0 to 8.5 seconds, which is around 4.5 times faster, most notably with encoding to MPEG. DV export suffered in performance and still is not up to the old CS4 performance. It even appears to get slower and slower with each new point-release.

                • 5. Re: About rendering differences CS5 vs CS4
                  HarryPutnam Level 1

                  Todd_Kopriva wrote:

                   

                  Show me the claims that were made that you have issue with.

                   

                  There are several performance improvements in Premiere Pro CS5.

                   

                  One of them is that some things are sped up using CUDA processing if you have a specific kind of graphics card. You didn't mention having one of these.

                   

                  Another improvement is the port to a 64-bit application, which helps with several things, including improving stability and making it feasible to work with very large frame sizes for footage (e.g., RED and other digital cinema sizes). You didn't say what kind of footage you're working with. I'm pretty sure that if it were RED or similar size you'd notice a huge difference.

                   

                  Another improvement is multithreading, which helps speed some things up on multicore computers, especially if you have a lot of RAM. You don't. People tend to report that the real advantages kick in around 12GB for Premiere Pro. For After Effects, we recommend 3-4GB of RAM per processor core, which is 12-16GB for a quad-core computer.

                   

                  Also, After Effects doesn't use CUDA processing, so if you're using Dynamic Link to After Effects, that's going to limit speed even if you have a CUDA card.

                   

                  First let me thank all reply posters for the generous and excellent input.

                   

                  Next, let me say that you have answered some of my doubts that I will mention further on.

                   

                  Regarding claims, of course they are all over the internet but the last thing I happened to read is HERE

                   

                  Both in the writeup and the video a keyword used there is `Blazingly fast' , which even given a very solid and good product can only be a tad bit exaggeraged eh?  I mean that kind of language should be reserved for something several orders of magnitude faster than what went before or than the competition.

                   

                  I would be surprised if the Hardware (and in combination with the software) that promo was done on was not expensive enough to rule all but a small fraction of non professional videographers.  And yet it is made to look easy squeezy.

                   

                  Regarding the graphic card.  No not `cuda' but ATI Mobility  Radeon 5800 series (its a laptop)

                   

                  And on that subject.  One of the major marks of really well written and robust software is its ability to run on common hardware and with averagely talented people at the helm and still do a creditable job.

                   

                  The list of special stuff needed for Adobe tools is growing at a quickened pace.  Now a special video card is required to expect good results and you mentioned multi-core and  12-16 GB of ram as the point were benefits start to show.   That is begining to borders on ridiculous really.  I know that kind of stuff is available but it is by no means the norm just yet and especially amongst us poor folks.  But if Adobe only intends to appeal to pros and rich folks then maybe that is a non-issue.

                   

                  Now about being answered.  You mention that comparing the `playback' (rendering) of an AE project in Premier is not a fair test.  I did not realize that AE was the playback or render engine in that case and that was what I was using for a basis of some of my comments.  Also I'm not really sure about whether or not adobe link would have come into play.  I imported it as one does an footage and did nothing special regarding adobe link.

                   

                  However I will add, and I'm not sure how this sets with your comments and again I use some confusion of terms here... I did happen to render that AE piece out as an ntsc DV-AVI from Premiere, and my wife (who is QC)  looked at the result and noticed a small spelling error and so had to go to AE to fix it... so since I was there I just used AE to render it out to ntsc DV-AVI and I'm pretty sure (I did not time it since I don't have time right now to experiment) that the AE render was the faster one.  It was only a 45 second bit too so not much to compare with.  Both tools use a good bit over 45 seconds.

                  • 6. Re: About rendering differences CS5 vs CS4
                    HarryPutnam Level 1

                    Harm Millaard wrote:

                     

                    If you look at the old benchmark results of http://ppbm4.com/Benchmark.html and look at the performance improvements of CS5 over CS4, you will notice around 30+ % increase in performance on the same system, just by moving from CS4 to CS5. That is a big improvement and that was measured with 5.0.0. Now further updates to 5.02 and 5.03 have given even more improvements, so the general conclusion can only be:

                     

                    CS5 gives a much better performance than CS4.

                     

                    In my own case the move from CS4 to CS5 dropped my results from 38.0 to 8.5 seconds, which is around 4.5 times faster, most notably with encoding to MPEG. DV export suffered in performance and still is not up to the old CS4 performance. It even appears to get slower and slower with each new point-release.

                     

                    Thanks for the link... that looks like something I'd like to try.  I don't have time right now but in about 2 days, after delivering my current project I will be loose and will give a try myself.

                     

                    That last part of you  second bit above looks pretty dreadful.  I'm seeing something right now with a list of 6 renders in queue. 3 are to f4v and 3 to DV-avi

                    But it seems to be moving pretty slow... I wish I could try the same set in CS4 and will later on.

                     

                    What do you think might explain the growing disparity, as you mentioned with v CS5 and each update.

                     

                    It seems to me and I'm not at all knowledgable on this, but  that some of the same points that make the mercury supposed to be so good for playback would also apply to rendering, maybe the muti-thread category, 64 bit computing, and taking advantage of muti-core.

                     

                    Also, even if Todd thinks my 8 GB of ram is wimpy, it is a right big lot compared to only a couple of yrs ago especially in the 32 bit world where most of us were then.  And its really a whopping increase to me who only recently entered the world of multicore and started doing editing on 64 bit.  So was locked in at 3 GB on my P4s.  I've had several unix and linux machines for yrs that had (amd) 64 bit architecture but not up to much softwarewise for editing video.

                     

                    But anyway, if CS4 is out preforming in rendering out a project... on the same machine as CS5 then of course whatever amount of ram is not a factor.

                    • 7. Re: About rendering differences CS5 vs CS4
                      Harm Millaard Level 7

                      Harry,

                       

                      Let me try to put your mind at ease regarding rendering performance.

                       

                      In the new PPBM5 Benchmark the tests were improved to better reflect reality. One of the tests is exporting a 1 hour SD timeline to DV AVI, consisting of around 550 instances of a short clip. No effects, no transitions. Exporting this 1 hour timeline takes around 60 - 75 seconds on fast systems, so that means around 50 - 60 times faster than RT. That is commendable performance.

                       

                      To run this test, you need to have the full version of PR CS5, not the trial.

                       

                      Now, my 'worries' are that this test takes longer and longer with each point release, going for instance from 65 seconds with 5.00 to 69 seconds with 5.01 to 72 seconds with 5.02 to 74 seconds with 5.03. This is of course negligent on a 1 hour timeline, but the fact that the render time increases by nearly 15% over some point releases is something I do not like to see. If anything, I would have liked to see 65 with 5.00 and 65- with 5.03.

                       

                      Hope this puts my remarks in better perspective and alleviates your worries.

                      • 8. Re: About rendering differences CS5 vs CS4
                        Ann Bens Adobe Community Professional & MVP

                        Harm, why are you testing dv-avi when most users have switched to avchd, mpeg2/4 etc? Do you still use a SD camera?

                        • 9. Re: About rendering differences CS5 vs CS4
                          Harm Millaard Level 7

                          To test disk speed. All HD formats entail heavy CPU usage. SD DV does not and that is also the reason not to include effects or transitions, since those would complicate the interpretation of results. Writing a sizable file (13 GB) to disk is a good measure of disk performance.

                           

                          While your question is valid, remember that a lot of people still burn to DVD and then DV export is still an important factor to take into consideration. Final delivery on DVD is still the most common format.

                           

                          FYI, I still use DV, HDV, AVCHD and XDCAM.

                          • 10. Re: About rendering differences CS5 vs CS4
                            Jeff Bellune Adobe Community Professional

                            I guess I've grown as complacent as most in overlooking the lying claims
                            that are allowed to advertisers and just figure it's par for the
                            course, which is really too bad.  We shouldn't allow that kind of stuff
                            to go on.  It is really kind of disgraceful that we put up with it
                            although I guess it would be really difficult to police if a company
                            were hell bent on exagerating.

                            It's also incumbent on the customer to know exactly what is being claimed.  Hardware MPE is effective only on CUDA-capable nVidia graphic adapters, and only certain of those have been certified by Adobe.  Other CUDA-capable nVidia cards have to be added to a list of cards in a text file (called the "hack" around here) to get hardware MPE.  But hacking the list is definitely a caveat-emptor moment.   I believe Adobe has been very clear about the need for nVidia graphic adapters.

                             

                            Todd has given you a great explanation of rendering vs. exporting, so I won't repeat that info here.  However, it is the responsibility of the customer to know the difference between the two.

                             

                            With an ATI card, you have absolutely no chance at all of experiencing any of the hardware MPE goodness.

                             

                            My laptop's specs are almost identical to yours.  I think that 8 GB of RAM, while not a *lot* (as Todd says), is sufficient for most tasks in Pr.  However, I do have an nVidia 250M that I've "hacked" for use in Pr, and I can tell you that when you apply hardware-accelerated effects to clips in a sequence, playback of that sequence is almost always in real time without having to render.  Real-time playback of an effects-heavy sequence is, IMHO, "blazingly fast".

                             

                            Sidebar: policing of TV advertising by the FCC is an ongoing effort, and they must be somewhat successful at it, otherwise you wouldn't hear a company claiming that "no other product does this better" as opposed to "our product does it better than any other product".  For a company whose product performs only as well as the competition, the former statement is true; the latter is a lie.

                             

                            -Jeff

                             

                            PS - you may want to file a Feature Request to have Adobe make hardware MPE available on ATI cards as well, using a standard like OpenCL.