HitFilm 3 Pro realtime performance

edited December 2014 in Express Support

Hi guys,

firstly congrats on Hitfilm 3.. so far, looks great, the new interface is slick, the extra features I haven't really got to yet but overall it looks great.

My question is about what to expect with realtime performance..

I used to run Hitfilm 2 Ultimate on a Nvidia GTX 650Ti 1GB, and was always able to get smooth playback with the viewer set to "Quarter" Render. I decided to upgrade my GPU, I am now running on a Nvidia GTX 970 4GB Overclocked. Even with the viewer set to render "Quarter", I still am getting choppy playback, but the exact same project in Hitfilm 2 Ultimate was pretty smooth at the same "Quarter" render in the viewer.

What am I doing wrong? I was hoping with a majorly more powerful GPU that I now have, I'd be able to at least get the same if not better performance, but at the moment I am going to have to proxy/pre-render everything, as I am doing music videos where I need to see that the playing/singing is in sync.

Is there some setting to force the GPU to be used during playback, or is it all RAM/CPU?

My specs are:

Intel Haswell Quad core i7 3.9GHz

8GB DDR3 RAM

Samsung 256GB SSD

MSI GTX 970 4GB overclocked

Any ideas or help is much appreciated. Thanks. Justin

«13

Comments

  • Performance generally should be equal or better than HF2, so this sounds very peculiar. What are the specifics of the project? Are you able to upload it so we can take a look, or is it tricky to send over?

  • edited November 2014

    Hi Simon,

    yes might be a bit tricky to send over.. I mean, the total project size with all video content is around 15 or 16 Gigabytes in size..

    I just re-opened the identical project file in Hitfilm 2 Ultimate and the playback with render "quarter" was smooth as expected..

    I can try and create a sort of mini project with say just one video in it that has the exact same comp effects and see if that has the issue if you like.. then I'd be able to upload it / send it through to you guys..

    Let me know and I will try. Thanks.

  • If you're able to create a contained example project demonstrating the issue which you can share that'd definitely be very useful. That should enable us to pinpoint what's going on and figure out if it's something that needs fixing. Thanks!

  • edited November 2014

    Hi Simon,

    I was able to create a short test project that shows the issue, or at least the fact that hitfilm 3 is choppy when playing back versus hitfilm 2. I didn't know how to upload here, so I will give you a oneDrive link (public) here shortly. I am just uploading it now.

    From what I can figure, the issue occurs when I have many comp shots in the editor that overlap, or many comp shots that exist at the same point in time. Normally I just keyframe the opacity for shots in and out, rather than cutting them up and committing to certain cuts - does that make sense? What I find is that playback is smoothest (and resource usage least) by clicking the "eye" icon off for a given track, but obviously I want to be able to bring a comp track in or out at certain points, and I can't keyframe the "eye" or visibility/active setting.. can I? But what is happening is that in the test project I am sending you is that I have the same comp shot duplicated 4 times in the editor, and the bottom 3 have opacity set to 0, and the top comp shot opacity 100% although setting the opacity to 0 for the bottom 3 tracks doesn't seem to do anything, as hitfilm seems to go "Ok, we have 4 comp shots. Lets process them all and then we have 3 that need opacity processed to 0". But in hitfilm 2, the playback at quarter render is smooth, but if you open the exact same project file in Hitfilm 3, the playback, still at a quarter render, is choppy.

    So it appears that hitfilm 2 processes multiple comp shots on the editor better than hitfilm 3, or is hitfilm 2 considering that a comp shot on the editor with opacity set to zero means that it doesn't need to process that comp shot at that time? I don't know..

    Which raises a general question: what is the best way to have multiple comp shots on the editor timeline that overlap? Cutting them up seems too final, especially if you then decide that you want to change things later.

    Anyway, sorry for the long spiel.. If you extract the zip file I've attached to a folder, and then open the project file first in Hitfilm 2 and observe, then open in Hitfilm 3 you should see what I am saying. Thanks. I will post the oneDrive link in a minute or two as soon as it has uploaded.

  • Great description, thanks. As soon as we can check out the project file we'll have a better idea of what's going on. The QA team will get right on it.

  • Cool. While the file is uploading, a couple of questions:

    Is it the normal practice to keyframe the opacity of comp shots on the editor as I have described? Or is there a better way, short of cutting and committing? I admit I am a purely hobbyist with VFX and video editing.

    Also, does hitfilm need to process a shot (either composite or straight clip) on the editor that has opacity set to zero? It would be useful if Hitfilm was "smart" enough to go, "Ok, this clip has opacity of zero from time 'a' to 'b' so we can ignore any processing on it from 'a' to 'b'.

  •  Cutting up comps isn't final at all.  On the Editor, composite shots behave exactly as video clips do; so you can very quicky andjust the start or end point of a comp on the Editor using the Edit tools. Slip, slide, ripple, roll, whichever tools you need can be used right on your comp, to adjust its length or content area on the fly.

    It should actually be quicker than keyframing opacity, at least once you get the hang of it, since you just grab a tool, drag the end of the clip, and you are done.

  • Yes that is pretty much what I thought, Axel, thanks. I guess I could just drag starts and ends of many of the clips rather than changing opacity, then if I change my mind later or want to do a different edit, just change the starts and finishes..

    I just realised from your post that there is a much better way of doing things.

    Maybe because hitfilm 2 handled multiple overlapping shots pretty well that I mistakenly thought that setting opacity to 0 meant effectively no processing - equivalent to turning the "eye" or visibility of that track off.

    Cheers.

  • I've found the same thing, that playback in Hitfilm 3 seems to be choppier than in Hitfilm 2. I've placed the same clip in each editor timeline to compare. Then again, I'm just running the trial version of 3, and on an outdated computer. That might be the problem.

  • Ok Simon, I have a link for you to test.

    http://1drv.ms/1zlpGXA

    I did enable "Relative paths" so it should open no worries. Just extract all to a new folder. And it was saved in Hitfilm 2 only (not 3) so it will open in both.

    Hope this helps. Please let us know how you get on. Thanks

  • KirstieTKirstieT Staff
    edited November 2014

    @FingerPeopleMovies - whether or not you're using the demo won't make a difference, but an outdated computer will certainly have an effect on how the software is running. Have you taken a look at the minimum specs for HitFilm 3 Pro to see where your machine falls short?  http://hitfilm.com/pro/specs

  • SimonKJonesSimonKJones Moderator
    edited November 2014

    @jsljustin - brilliant, downloading now to try on my home machine. Will let you know what happens.

    EDIT: Have tested the project in HF2U and HF3P and there is indeed a performance degredation in HF3P. This obviously shouldn't be the case (it should be the same or faster), so the devs will look into it ASAP (it's late evening here now, so it'll be in the morning).

    Thanks again for reporting it and especially for providing such great detail and an example project. That should help us to pinpoint precisely what's causing the peformance difference.

  • Cool Simon, happy to help :) I appreciate how you guys are more than willing to work with us end-users to resolve issues.. its fresh air compared to many companies these days.

    And by the way, the OpenEXR export ability, ofx plugin support and denoiser are just some of the brilliant things you guys have done to massively improve Hitfilm. I only do a music video twice a year (it takes me the rest of the year to do a full CD), but I always enjoy using hitfilm as it is so intuitive and just generally good fun to work with.

    Cheers.

  • No worries, and sorry you've run into this performance issue. Hopefully the devs will have a better idea of what might be causing it (being a non-techie, I wouldn't like to comment one way or the other!).

    We've always tried hard to listen to feedback and respond to it and we're intending to do that to a far greater degree with HitFilm 3. Proof is in the pudding, as they say, so time and your good selves will judge us on that one. :)

  • Is pudding going to be added as a post-release update?

  • Haha @AxelWilkinson...

    Maybe a 3D unified black pudding?

  • Interestingly, I'm seeing people report elsewhere that they're getting noticeable performance improvements when comparing HF2 projects. So there's mixed results coming out, which makes me suspect it may be related to a specific use case(s).

    Anyway, the devs will no doubt have a better idea in the morning. Thanks, everyone!

  • That's interesting to hear Simon.

    I'm no big techie either, but my engineering background and given what you have just posted makes me want to try isolating some of the aspects of the test I gave you to see if the performance hit has something to do with one or several aspects of the comp effects I was using. Might help speed up the discovery of the issue if I can nail a given aspect that might be causing it..

  • edited November 2014

    I haven't done any time comparision of using it, but a render comparision showed HF2 at 8:16 and HF3 at 5:12 (m:s), so it was noticably faster in this case. No proxies were used for HF3 and it was a beta when I did the test. My test used a LOT of particles.

    So it does sound like there are some situations where it's faster and some slower.   Probably depends on the effects used?

  • AphAph
    edited November 2014

    Slower here using the preview. Haven't tried rendering. A very simple composite plays smooth on HF2 but slightly jerky on HF3.

  • I'll confess I didn't see a performance difference between HF2 and HF3 in the sample you provided Justin.  I had been benchmarking HF2 and HF3 with some of the tutorials, and the only consistently slower one I found (from what I tested) was the EPTitles tutorial, specifically the Exploding title, where I had doubled the number of particles specified.

    I suspect we may be seeing differences depending on what OpenGL versions each piece of software is optimized for, further complicated by what/how OpenGL is accelerated by given versions of drivers.

  • edited November 2014

    @Daniloth - excellent point you make. Just to be clear, the realtime preview is GPU accelerated, correct? And I just recently upgraded from a Nvidia GTX 650Ti to the Nvidia GTX 970 which is only a few months old, I mean it was only released in September this year I think.

    But on my machine, however the OpenGL and whatever else is being used between the drivers and Hitfilm, if it is a GPU/hardware-dependent issue, is showing that whatever HF2 did in terms of GPU calls or whatever may be more efficient than the way HF3P is doing it.. basically what you said, I totally agree. (If it is hardware). Simon said many users have reported improvements in performance over HF2.

  • edited November 2014

    Ok just an update.. You probably wont get this till the morning Simon, but this afternoon HF3P, the exact same files / project I sent through is now playing back ok, pretty much the same and quite smooth as HF2U.! Strange! Tested as many things as I could think of, including killing running processes to see if any made a difference.

    The only thing I could think of, as "nothing" as far as I can tell has changed on my machine since this morning - is either my antivirus or malware bytes anti exploit. I'm running both avast 2015 free edition and malwarebytes anti-exploit. Both do realtime monitoring.. Avast does analyse programs as they launch and often when running, especially new programs. Now perhaps Avast has gotten "used" to HF3P so it no longer bothers to look at it (and possibly slow it down) when running.

    That's pretty much all I can gather. So hopefully the dev team will nail it tomorrow. Otherwise, maybe it is the  security software.. lol.

    Good afternoon, good evening and goodnight :)

  • @jsljustin

    if this was a transitory effect can be not coming from OpenGL version.

    I suspect can be coming from proxy creation? Working with an NLE I did need to disable proxy creation since working in background was slowing all my work. When finished all was really fast, but I resolve the problem using a SSD.

    Also when I work on heavy videos I did disable almost any service on my machine (working offline, so no viruses around).

    The difference can be huge (at least on my laptop): from a choppy playback of a tree video multicam shot to be able to work on it without any problem.

  • SimonKJonesSimonKJones Moderator
    edited November 2014

    @jsljustin - peculiar! I definitely noticed a difference on my home machine, so there's something going on.

    I just tested on my work machine and there's no noticeable difference between HF2 and HF3. Will do further testing at home tonight.

  • edited November 2014

    Are any of you who are experiencing performance issues using Chrome? I've noticed that it slows down Resolve renders drastically... Chrome is a huge pig, and isn't satisfied with consuming massive amounts of CPU cycles and ram, it wants to consume GPU cycles also... so I've taken to killing it before starting up GPU heavy tasks.

    On my Surface tablet, if I have Chrome running when I launch Premiere, Premiere often complains about the lack of OpenCL hardware support. And playback stutters like crazy with Chrome running. 

    It might be worth a quick experiment :)

  • edited November 2014

    Strange Simon!

    I almost wanted to see the issue again, but like any good bug or issue, the more you want it appear, the more it just says "no, I'm not showing up".

    About proxies Hitfilmer120549, can someone confirm - proxies are only created if you right click on a comp or other media item and select it to be proxied, correct? I mean, I initially didn't use any proxies with the problem occurring, but at one point I did right click on all the media and tried to proxy them, but then later clicked on them to stop or disable the proxying of them.

    As for this issue, the only reason I suspect my anti-virus is that it only would interfere and possibly slow the processes needed with hitfilm down when the program is first installed. If it was a general system-wide resource issue I suspect I would've seen performance issues with Hitfilm 2 ultimate as well, but I never did. Only for about 4 hours was I seeing performance issues with HF3P, now I am pretty much not seeing performance issues with it, it is pretty much on par with HF2U.

    The other thing I am aware of that can cause all sorts of realtime video and audio playback issues is DPC latency (Deferred procedure calls). I have had audio crackling and other sync issues caused by this, and my particular antivirus is well known to cause this.. but again, if it were intermittent high DPC latency issues, then it would randomly also affect HF2U..

  • I will note that HFP3 is noticeably slower on my aging i72640QM and NVidia 580GTXm than HFU2, but, given enhancements to the 3D space and 3D model particles, etc, I just assumed the render engine was pushing more data. More complex renders need more power, therefore slowdown. If the devs CAN get HFP3 performance up to HFU2 performance across the board, or get it faster, that would be fantastic! 

  • The same peculiar thing happened to me. HF3Pro was much slower than HF2U upon first use, but upon multiple uses it's now on par with HF2U, (and faster in some ways). @jsljustin You mentioned having Avast and its tendency to slow down programs upon first use. I also use Avast, so perhaps that's the problem and could explain why some users found HF3Pro to be slower and others didn't.

  • FingerPeopleMovies

    Good points mate! That was my thinking.. I can't see of any other reason why HF3Pro would suddenly speed up after several uses.

    Simon mentioned he had performance issues on one of his machines with HF3P, be interested to hear if/what antivirus if any he has. I suspect that most antivirus / security software can slow programs down, sometimes just initially, sometimes ongoing.

    I, like you, can only think that it must be avast in our case that had slowed HF3P down.

Sign in to comment

Leave a Comment