[Criticism] How to choose the best computer for running HitFilm

CNKCNK
edited July 2016 in Everything Else

Hello everyone, the whole point of this thread is to show you guys that the AMD FX CPU's should and can still be considered for HitFilm, as opposed to "make sure it's an Intel".

Source: YouTube

FXHome has said that HitFilm will take advantage of multiple threads as stated by Hendo, "HitFilm's H.264 encoder is multi-threaded and will make use of all cores and hyper-threading if available."

Source: HitFilm Forum

There's a common misconception that AMD FX CPU's are not worth buying. That is only true for gaming, not workstation applications that are able to take advantage of multiple threads

Here's some proof that I have collected in order to show how much value you get from buying an AMD FX CPU, when compared to Intel, including pricing (all builds are in GBP £, but price difference is more or less the same internationally):

AMD FX-6300

 

Intel i3-6100


 

AMD FX-8320E


 

Intel i5-6400


 

AMD FX-9370 (to simulate an overclocked FX-8320E)

 

Intel Xeon E3-1231 V3 (to simulate the performance difference between an overclocked FX-8320E)

The Xeon is equivalent to an i7-4770, but it doesn't have an integrated GPU (iGPU), hence why the Xeon is the closest to the FX-8320E, as it doesn't have an iGPU either.


 

Tom'sHardware benchmark in Adobe Premiere Pro CS6 between the i7-4770K (Xeon) and the FX-8350 (inbetween FX-8320E and FX-9370)

 

 

Thank you for your time,

Kevin

 

Sources used: 

  • uk.pcpartpicker.com
  • www.futuremark.com
  • www.cpubenchmark.net
«1

Comments

  • Triem23Triem23 Moderator

    You definitely did your research. 

  • @Triem23 - I had to, I'm still in shock!

  • edited July 2016

    Sorry to shock you Kevin and, as I've said elsewhere, there are tons of powerful CPUs on the market now from both AMD and Intel.

    I've explained the intended focus and simplicity of the YouTube video on the other forum post so I won't repeat that but I've honestly had no recommendations for AMD in high-end home desktops for editing and graphics in the last few years from any source I know.

    The high end section of the PassMark link you provide is completely dominated by Intel i7s and Xeons but PassMark is a terrible measure of actual PC performance. Synthetic benchmarks have a tendency to be inaccurate to the point of being grossly incorrect.

    I would place far more stock in SYSmark and unfortunately this is where AMD isn't doing very well (at the moment). In their media creation benchmark the top AMD is still lower than some Intel i3s and  well below the chips I recommend in the video. We've certainly had similar experiences here when comparing several AMDs to i5s let alone i7s.

    That said, I place the most stock in reviewers and industry people I trust. After all, we have no association with either Intel or AMD and only want the best systems for our team and our users - we have to buy both for testing purposes anyway so we have no advantage to gain here by recommending one over the other.

    Despite searching all morning I've not been able to find any recognised reviews where a current generation  AMD chip is besting a current i5 or i7 in the majority of the comparisons the reviewers are making. There are places where AMD do well but in our own test and overall on the internet (especially in reviews like the one below which are non-synthetic/theoretical) Intel are dominating the high-end market. For that reason, when we're trying to keep things simple in a YouTube video, they are the logical thing to suggest for a high-end home editing system.

    http://www.tomshardware.co.uk/skylake-intel-core-i7-6700k-core-i5-6600k,review-33276-5.html

    All that being said you can have a completely great system still which has an AMD chip as the CPU. A computer is more than the CPU alone and, as the video suggested, there are other areas which are just as important to overall HitFilm performance. :)

  • CNKCNK
    edited July 2016

    I mentioned this on the other thread, but if that's the case. If you (your company) found that the FX CPU's fall behind Intel i3's, i5's i7's, then HitFilm is not properly coded to support multiple threads efficiently. 

    So, the claims should be removed, because they aren't true.

    I also included the Adobe Premiere Pro CS6 benchmark. A FX 8000 series CPU is equivalent to performance of a Intel i7 (3x the cost).

    Tom'sHardware owns Anandtech, so they're both reliable.

     

    I did directly say that you guys are either not aware, or misinformed. I think the latter, because well the amount of Intel bias out there is huge. 

    People buying i5's or Haswell or older i7's, the primary use being workstation applications (HitFilm) are straight up throwing their money away.

     

    "Despite searching all morning I've not been able to find any recognised reviews where a current generation  AMD chip is besting a current i5 or i7 in the majority of the comparisons the reviewers are making."

    I think I know why, people might not know that QuickSync decreases render times but the quality output is much lower.

    I've seen a video of a i3 6100 beating an FX-6300, so yes, A TON of misinformation out there.

  • edited July 2016

    I'm echoing what the industry, reviews and non-synthetic tests are saying about the relative performance of the i5 and i7 vs AMDs offerings at the moment but it is also the results we see ourselves (not only in our software development).

    If you could point us at scientifically performed reviews showing current AMDs beating current i5s and i7s I'm very happy to read them and discuss them here. This would be great for the industry as a whole, all my first PCs were AMD chips and it made for a healthy and competitive environment.

    I'm totally with you that there are cases when more cores can help which is why I recommended 4 in the video. Encoding footage can be one of those places (your example for CS6), but this doesn't mean it'll help with everything all the time. In fact, as you can see in the full report I posted, it definitely doesn't overcome the sheer performance of the i5/i7 even when it has less cores.

    Utilizing and maximizing all the cores all the time simply isn't the usual use case for any chip. The SYSmark rating below show a better indication of the relative speed of these chips than the PassMark (although I still wouldn't entirely base my decisions upon it).

    As you can see, the AMD only begins to feature fairly low on the chart and all the top positions are taken by Intel - I've missed out huge gaps here just to get down to the AMD chips as there were many Intel chips above them. For further comparison, and to update the example you gave, here are some more charts from tomshardware last year.

    http://www.tomshardware.co.uk/charts/cpu-charts-2015/-23-Blender,3714.html < The AMD chips are slower than the recent i5 and i7s and many older models too.

    http://www.tomshardware.co.uk/charts/cpu-charts-2015/-31-Adobe-Premiere-Pro-CC,3722.html < More up-to-date version of the chart you posted. The AMDs 8 cores help here but its still beaten by 4-core i5 and i7s.

    http://www.tomshardware.co.uk/charts/cpu-charts-2015/-30-Adobe-After-Effects-CC,3721.html. Again outclassed by the i5s and i7s.

    What is common to all these lists is that the top spots are taken by Intel as are the majority of the higher performing results. AMD does do well in some places but CURRENT i5s and i7s do well everywhere which is why (for the sake of a simple video) I maintain they are the right ones to recommend... but that doesn't mean I'm slating AMD and don't think they bring value to the market.

    The fact that tomshardware exclusively recommends AMD chips for people wanting to spend less than $200 on the CPU shows they are beating Intel at that price point. But by that same token, all the chips they recommend above $200 are Intel for good reason.

  • CNKCNK
    edited July 2016

    Yeah I don't trust any of that at all. Please go out and test this yourself, I don't have the money to spend.

    @Aladdin4d Owns an FX-8320, so if someone with an i3 that's at least newer than Ivy Bridge, then we should create a benchmark for it to run using a script or a final render, and we would see then how optimised HitFilm is for multiple threads.

    If the i3 wins, then all claims that HitFilm properly supports multiple threads are false.

    Adobe while not primarily OpenGL, is the polar opposite of HitFilm.

    Not a single i5 today would beat the FX-8000 in raw horsepower, it just doesn't make any sense what so ever.

    My current CPU (A8-5500) is too old to compete in raw horsepower with even a Pentium CPU, so there's nothing I can do other than provide common sense to this discussion...

    Hopefully it doesn't get ignored and forgotten. 

     

  • edited July 2016

    "If the i3 wins, then all claims that HitFilm properly supports multiple threads are false."

    There are far too many factors in HitFilm (and over video editing software) including all the different parts of the pipeline, the operating system, third-party encoders and decoders, compiler optimisations which may or may not be impacted by core count far less than the single core performance.

    Furthermore, if you had an 8-core i7 vs a 4-core i7 there are very few cases where you will get double the performance - if you're lucky you'll get 50% more.

    "Not a single i5 today would beat the FX-8000 in raw horsepower, it just doesn't make any sense what so ever."

    If you've got 4 fast cores, vs 8 slow cores then it is possible for the 4 core to win... especially if it has reduced overhead and boosted single core clock speeds. Plus, having lots of theoretical horsepower doesn't mean that it can be used, hence the PassMark results. An 8-core i7 would theoretical have  double the horsepower of a 4-core i7 but this never adds up to real world gains.

    Please see all the stats above which include other multi-threaded software like Blender, After Effects and Premiere, and even if you go look at something very specialised like Handbrake. All this software supports multi-threading - the results for all those products, just like HitFilm, show the Intel chips being faster (at the moment) despite less cores.

    Anyway, I hope you can see we have nothing against AMD and why we kept things simple with our recommendation. The reason I pointed out tomshardware was so people could make up their own minds and hopefully visit the great Best CPUs for the money page if they want to know more. :)

  • Triem23Triem23 Moderator

    @JoshDaviesCEO You also definitely did your research. 

  • CNKCNK
    edited July 2016

    So basically what you're saying and what I've always said/thought is that HitFilm acts more like a game than say Premiere Pro when it comes to take advantage of the CPU and the amount of threads a CPU has.

    Premiere Pro is pushing all threads to 100%, HitFilm lets the GPU do it. Is it because this happens, that FX falls behind the Intel in HitFilm? 

    Intel's and AMD's frequency are nearly identical generally/most of the time, so that is the only explanation that makes sense, Intel CPU's, at the same frequency, pull far ahead of AMD.

    That's exactly the question I've tried to answer using my research. It's weak, but it took some time to make for me.

    No answer in sight yet though.

  • edited July 2016

    @KevinTheFilmmaker , when you posted: "Yeah I don't trust any of that at all. Please go out and test this yourself, I don't have the money to spend." you lost the argument.

    Claims (from both sides) require evidence and all you're saying is: "I don't believe you and I've got my fingers in my ears." Josh says they test AMD and Intel in house and that his findings agree with everyone else. What more could you want him to say?

    I couldn't care less one way or the other if AMD or Intel "wins", but multiple sources seems to indicate that on balance: Intel is likely to be better at more things, some of which might be video editing and some might be playing games and it's not just limited to performance with Hitfilm.

    If everyone else says it walks, looks and quacks like a duck...it's a duck. :)

  • That's not proof either though. We need to test this.

  • CNKCNK
    edited July 2016

    Oh look, totally not expected at all!

    "In AMD's testing SYSmark AMD has found that SYSmark provides data that is not comparable to the difference in real world performance when it compared AMD and Intel CPUs, giving a performance delta of 50% for a comparable system when other benchmarks like PCMARK only give a delta of 7%.

    The US' Federal Trade Commission, the FTC, now required Intel to post a disclaimer when using data from this benchmark, stating that SYSmark may only be optimized for Intel platforms. "

    Here's proof that the quote is real, and that AMD are bashed by Intel and are misleading consumers.

    Directly from FTC: https://www.ftc.gov/sites/default/files/documents/cases/2010/08/100804inteldo_0.pdf

    "Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchase, including the performance of that product when combined with other products. "

  • Triem23Triem23 Moderator

    While Josh hasn't (and doesn't need to) give us exact numbets, he has said multiple times that FxHome does have Intel and AMD systems and that they, overall, are seeing better performance on Intel. I have to assume this counts as testing? 

    The problem with a "benchmark" scene is every other variable: RAM speed, bus speeds, storage speed... What other software is running? Silly things like Antivirus or custom control mapping software can steal processor cycles all over the place. One thing dedicated test sites do is maintain some identical machines, then swap out the part being tested in order to eliminate variables, but a community benchmark can't do that! 

    Kevin, saying you don't trust the results from Josh citing Tom's Hardware was a mistake that did "cost you" the argument. For one thing your sources cited Tom's Hardware, so... You kind of invalidated your own research there!

  • You're quoting something from 2009/2010... If that's still relevant to today's CPUs then the people to blame are AMD for not improving their CPU architecture.

  • CNKCNK
    edited July 2016

    Yes but it's still not accurate for AMD CPU's.

    Please understand that I'm not trying to say AMD is better than Intel. 

    I do want a yes/no whether or not AMD runs worse in HitFilm because it doesn't saturate all it's threads, because Intel has got a big core advantage.

     You two are taking what I say completely out of context.

    The Tom's bench he linked

    "


      • 3.7 GHz (Turbo 3.9 GHz), DDR3-1600, 1 MB L2, 10 MB L3
      139

      • 4 GHz (Turbo 4.2 GHz), DDR3-1866, 8 MB L2, 8 MB L3
      139

      • 3.5 GHz (Turbo 3.9 GHz), DDR3-1600, 1 MB L2, 8 MB L3"

       

      The FX 8350 is inbetween the i7 3770k and the i7 4820k.

      That's a matter of common sense, not where the benchmark comes from.

       Fact: FX 8350 is on par with Intel i7 and can compete for less 1/3rd the price.

      My original question has still not been answered. Why?

  • edited July 2016

    Well I have been running Hitfilm on this laptop for a long while and it is a far cry from being intended for video editing/3d animation but it chugs along just fine with its i3 processor. Yes it could be a lot faster with a SSD and a GPU and only thing changed on it is Win10 and 8 gigs of ram, but I am pretty much happy with its performance for what it is and how old it is. http://support.hp.com/us-en/document/c04008973 ; But I guess it comes down to which type of person you are. Being an Intel person or an AMD person is kind of like being a Chevy or a Ford person. Both have their perks but a die hard fan of one isn't going to come out and say that yes the other is better.

  • Aladdin4dAladdin4d Moderator

    Well things have morphed since I was last here! Anyway as I said in the other thread I don't think there's any question Intel is faster but you're going to pay more to get that speed and according to some of the research on here it looks like $200.00+ for just the CPU. If that's out of your budget then you're looking at something from AMD. I haven't even seen the video being talked about yet but I would remind others the original thread that sparked this debate was started by an individual with a proposed $400.00 budget. 

  • CNKCNK
    edited July 2016

    I need to improve my english, I don't think people understand me.

    Since Josh says they tested it, then he's verifying that HitFilm prefers faster cores over more but slower cores.

    @Aladdin4d ;

    No, I'm buying a new CPU soon, I want to know the answer to this question. That's why I made a new thread. :P

  • I really wish there were some official hitfilm benchmarking test. I remember back in the day people used part of the HP wand tutorial, having something like that would help a bit (although there would still be a million variables to take on account).

  • edited July 2016

    @KevinTheFilmmaker

    As I state on other threads I did like informed rational purchase. I also like original thinking and outliers, so that I tend to prefer smaller and more aggressive and innovative companies than the enstablished mammooth. I'm not and AMD fanboy, but for sure I like to see them succeed with good products, with healthly competition.

    In case of AMD cpu the problem is coming all from their Buldozer architecture down from 2011. Simplifying they aggregate cpu and fpu unit in 2+1 cluster. We can say 2 AMD core are more like 1.5 Intel, or even less. This first generate horrible performances considering Windows scheduler was unable to manage correctly this unusual arch. Fixed this problem the architecture still generate lower single thread performances, something was never fixed in all following implementations. The fact they need higher clock to compete, and need so a lot of power, didn't help.

    Sw developers need to optimize their code to fully exploit this more multithread oriented app, but first this was and still is more complex than single thread implementation, and second with AMD market share at 20% there was less push to do so. With also more load shifting to GPU there was even less push.

    The market relegated these CPU on the lower side, with some exception where sw side there was some specific optimization or their good APU that for the cost offer some decent GPU experience.

    AMD can do better and will do in the upcoming Zen, in the meantime I'm not ashamed to suggest Intel CPU to anyone if it's not really budget constrained.

  • CNKCNK
    edited July 2016

    I understand that and I appreciate the explanation, but it doesn't answer my original question. If an i5 beat a FX in HitFilm, then why are they telling people to buy a CPU with more threads for more performance, because threads are equal, so if it takes advantage of Intel threads, why not AMD?

    If an i7 is better than an i5 in HitFilm, then A FX 8000 should be better than an i5, why isn't that the case?

    Is it because HitFilm offload to GPU so that the frequency of the CPU's (AMD/Intel) are more or less the same, which Intel has a big advantage at due to the much better single core performance?

    None of you understand my question and i't starting to get frustrating, how can I explain this better?

     

  • Again this is just a review but explains a little in the process between amd and intel. http://www.digitaltrends.com/computing/here-we-explain-the-basic-differences-between-intel-and-amd-cpus/

  • @FlyingBanana78 - For gaming, yes, but the question I'm asking is about multiple threads in HitFilm.

     

  • @KevinTheFilmmaker

    https://en.wikipedia.org/wiki/Amdahl's_law

    So, so summarize, more cores will increase speed only when almost all parts of a process can be parallelized. Video editing/encoding/decoding have a lot of things that cannot be done in parallel, so single  thread efficiency will always affect the process more.

    This is a common misconception, when people ask for GPU encoding. Cannot be done, unless you use GPU hardware encoder, but this gives you a lower quality output.

    Marcin

  • CNKCNK
    edited July 2016

    HitFilm's H.264 encoder is multi-threaded and will make use of all cores and hyper-threading if available.

    With that out of the way, then the only possible outcome is the GPU bottlenecking the CPU before it's able to encode a frame, right?

    So, because of that, HitFilm runs better on the Intel CPU's, because the threads aren't pegged at 100%, which is when Intel CPU's pull ahead in multithreaded applications.

     Are we on the same page? 

     

  • edited July 2016

    IMHO H264 encoding it's not a good example about the whole package speed on a kind of CPU. Video encoding and decoding can be hw accellerated so does impact less on global performance, also can more easily parallelized.

    But as example particle simulator (the most interesting part for my needs) it's not multithreaded, so here single thread performance is massivelly important. Looking around for some faster solution I did find suprisingly few multithreaded one, meaning this is not an easy parallelizable task.

    This is also true for many other high end packages such as Cinema 4D, where also single thread performance is vital in viewport during design and simulation phase, and multithread is massive only at rendering stage. Maxon  - C4D developer - is taking years and years to develop a new core, supposelly fully multithreaded. When you can find processes offloaded to GPU mean they can be massively multithreaded, and looking around in every direction I did find few of it. It's a trend, but just in the long term.

    Bottom line single thread performance is vital in not trivial situations, that's also why upcoming AMD Zen is claiming +40% single thread performance vs previous generation, returning finally AMD also in the CPU game with some credible products.

  • Triem23Triem23 Moderator

    Kevin I think you're on to something with GPU load as well. Remember, Hitfilm is using the CPU mostly for video decode/encode, storage I/O and particle physics computations. Pretty much everything else is going to the GPU. And, as others have said, much of Hitfilm seems to be single-threaded.

    One point I want to re-state: There's thousands of viable PC part combos out there and only so many to cover in a 5-minute video. Also, Josh was specifically discussing a higher-end system. So they made their picks and went with the video. I think if Josh had a "Now we're looking more at a higher-end system today--there are tons of other configurations that will run Hitfilm, but today we're aiming at the best," then we wouldn't be having this discussion. So there's a slight failure to communicate in the script.

    @Aladdin4d and anyone else who missed it: The starting point for this discussion is this week's video from the Hitfilm Channel.

    https://www.youtube.com/watch?v=zWFGg9R0m64

  • CNKCNK
    edited August 2016

    Then why say something like this?

    You don't need expensive computing hardware to run it either, with company founder Josh Davies saying that it will work well on the "£250 [$363] computer that you'd pick up for your kids at PC world."

    Im very confused as to what to believe. Theyre 100% not hitting the biggest user base of HitFilm recommending i7s and i5s.

    I understand (at this point) that not all parts of HitFilm is multithreaded, but I never not even once mentioned that, I from the start was talking about the H.264 encoder.

    The best way to get to the bottom with this is to test it. An FX-8000 vs an i3, use particle system and so on, including final render.

    I'm genuinely curious.

    About the Sysmark scores, well it's still an Intel benchmark today, I don't see how it's relevant how old it is.

     

  • I'm sorry that this argument seems to be going around in circles and the focus seem to be changing so I'm going to leave it after this post. You know what people say about arguing on the internet. :)

    As I said before, I like AMD as a company - and when they had a performance lead over Intel they were the chips we purchased but that is a long time ago now.

    To clarify... you don't need expensive equipment to run HitFilm, but faster (more expensive) kit undeniably runs our software (and all other software) better/faster - nobody has tried to suggest otherwise, this is why people buy beefy computers.

    The video was about the best home computer to run HitFilm, not the cheapest - this was a question we are asked a lot and I feel that we answered that in the video. If you want to create the cheapest computer that can run HitFilm decently then the answer would be different - but that isn't the question we get asked all the time.

    The i5 is a mid-level chip and isn't an expensive outlay for anyone serious about video editing and graphics. You're not denying the tomshardware stats which still put the FX behind (or way behind) the chips I recommended - the new i5 and i7 - which should be enough to put this to bed.

    But just to close things out... I'm sure its possible to find specific cases (like encoding video) where some encoders are highly parallelised so the FX is able to keep up with the i5 (maybe even beat it) but unfortunately these cases are undeniably rare hence all the reviews and the majority of the stats favouring the i5. Most general usage functions on a PC are not highly parallelised, even when it comes to editing, which is why Intel's faster cores do so well - the FX isn't a good compromise here compared to what the i5 offers and that is just as true in Premiere. When software is more parallelised Intel currently has a better answer in the i7 which is more expensive but still not a huge outlay for people who are seriously getting in to filmmaking (and often buying expensive computer systems to go with their cameras and lenses).

    HitFilm and Premiere are different software programmed by different teams using tons of different technology - comparing them on a specific task will not really prove anything other than the software is different. We will be using different compilers (which are optimised in different ways) and different encoders/decoders and rendering pipelines therefore I would expect us to be utilising things totally differently to Adobe.

    I know Premiere doesn't fully utilise all my cores generally while I'm using it although it is quite likely during some encoding that it might. HitFilm will be utilising the CPU and GPU very differently and our third-party encoders (unlikely the same as Adobe's) will also be doing different things too which we don't control.

    Therefore to answer your question - as of July 2016, when running HitFilm, which includes computing the user interface, the particle system physics, a chunk of graphics processing and conversions, encoding and decoding video with 3rd party libraries, processing 3D data and models we find (as do the rest of the industry and all the reviews and stats linked) that the latest i5s give better overall performance than the FX.

    There could be cases where the best FX is able to hold its own or even best the i5 (but not the i7) but overall the i5 is faster and more significantly so than the few cases where the FX does well. It is also faster running the other software you might have (as per the reviews and stats) hence it is our recommendation (along with the i7) if you want to make a great desktop computer to run HitFilm and editing software.

    As a final note... the unfortunate position for AMD is that they currently have a platform which is harder to optimise for (many functions are very complicated to effectively parallelise) and over 80% of the desktop market is now using Intel CPUs which are easier to work with... without a fundamental change AMD have a mountain to climb, but they've done that before and I wouldn't bet against them doing it again. :)

    Both Intel and AMD can run HitFilm well - so we can all be happy with that. As always, we'll be continuing to optimise new versions to run better and better on all the latest hardware from both chip makers.

  • CNKCNK
    edited August 2016

    Thank you for all the answers @JoshDaviesCEO

    Josh, as a side note, if you're reading this. It would of helped so much if you added that AMD CPU's are too slow, because if you come from Premiere, or worked with applications that push all thread during final export to 100% and it never drops, then it can be confusing. 

    I understand that I may have wasted your time, but you chose to reply so there's that.

    I'm still confused how so many people confused my original concern and ONLY concern which was H.264 encoding (the screenshot I provided is from Hendo, lead software developer correct?), so why so much talk about single core/thread performance?

    Basically if you argue against that you argue against Hendo, so why are people trying to force the topic to something completely different, that didn't make any sense to me.

    I wanted to find out if the H.264 encoder was bugged or bottlenecked by the GPU. So if you have a very expensive GPU, you wouldn't buy and AMD FX CPU either, which is only when the AMD would pull ahead of the i5/match i7 haswell and older.

    Perhaps if my english was better then people would understand, but none the less, I asked because I didn't know, hopefully no hard feelings for anyone involved. :)

     

Sign in to comment

Leave a Comment