Benchmarks time again and have a favor to ask

[Eric Bowen]

I am starting to run benchmarks again with the releasing of all the new hardware this year such new i7 10 Core 6950X & other i7/Xeon Gen 5 E CPU's, Pascal Nvidia cards, higher clocked DDR4 and more. Unfortunately I no longer have my newer R3D 5K & 6 K test media. I am asking if any fellow community members might be willing to provide me with some sample 5K, 6K, & 8K R3D media I can use to benchmark as much of the new and current hardware including mobile as we can get our hands on. I will also be redoing these benchmarks when Adobe releases their coming new CC version supposedly in June. I would like to also get some done with Davinci 12. What I need to really benchmark this and check for dropped frames is around 10+ seconds of media. From there I can duplicate as needed to extend the timelines for the benchmarks. However I need enough continuous media to verify dropped frames on playback so 10 seconds should be enough. Having some different frame rates besides 24 or 30 frame would help as well such as 60 or 100 frame alteast in the 5K flavor up. Please feel free to post links to sample media you are willing to provide, personal message, or email links to my email support@adkvideoediting.com . If you also have some Blackmagic 4K+ DNG or Sony Raw as well and you want to see benchmarks across that hardware let me know. Any media with enough motion to verify frame drops would be great. Thank you for your time and I hope to be able to provide some valuable info again when this is done.

Eric
ADK

 

[Mark Burten]

I'll see if I can pull some 5K/6K stuff from an Epic body and have someone shoot variable speed. Have Alexa footage as well if you'd like Open Gate etc. Hopefully round it up before getting ready for Cinegear. Thanks for all the work...Looking forward to the tests!

 

[Eric Bowen]

Thank you very much and yes Alexa media would be welcomed as well. The more codecs used the better idea given for how different codecs require different hardware specs or perform differently on hardware. I plan on holding onto this media submission for future benchmarks as well since it gives a good base line to compare as new hardware and software releases. I will be posting the results in a thread as I get them done for each hardware platform tested.

 

[Eric Bowen]

Mark can you email me or PM me links to download the media you have. I have another reply as well for some 5k and 6k R3D media. I am waiting on the media to continue the benchmarks. If any community members have 8K, 4K DNG or any other media you want benchmarks on please let me know. Once I have these media files and can finish these benchmarks on the 6950X with the 980ti and start posting them for everyone. Thank you again.

 

[Flo Eckhardt]

Thanx Eric ! Again looking forward to the results! Do you still need r3d in 5k/6k all kind of frame rates ... please pm!

What do you think of establishing a reduser r3d benchmark so we could all compare our machines with files and programs we actually use!? I would be down to help setting this up with a few partners in crime!

 

[Eric Bowen]

Setting up a Red User R3D benchmark is a good idea. My benchmark projects are not actual editing projects since I am trying to push the GPU acceleration for Adobe as an example as much as possible. What I do is take 4 layers of R3D for example and divide each layer into a 25% of the frame ie 1 layer for each corner of the frame. This forces scaling on all the layers which is GPU accelerated. Then I take known GPU accelerated FX using the same 5 on all 4 layers for example. I make sure to use only FX that does not require rendering or analyzing the frame first like warp stabilizer does for example. I couple this with different frame rates of media to force interpolation for example going from 60 frame to 29.97 which is also GPU accelerated. Then I render the media into different forms such as DPX, DNxHR, and H264 for example to show how different codecs render and perform differently. The entire premise is to push the GPU acceleration as much as possible and see how the different hardware performs since this gives the information everyone wants. It does not look pretty but the GPU's often push 80 to 90% plus load on render and 40 to 50% load on playback. I may have to increase the number of FX or change this benchmark depending on how much load the R3D media puts on the 10 Core i7 with the 1080GTX for example since my current R3D 4K media is really not pushing the 980Ti above 40% on realtime playback and 70 to 80% on render. I will know once I have the 5K and 6K media. The 10 Core i7 6950X at 4Ghz with DDR4 3000 is a an absolute monster chip. AE generates ram preview for 20 to 30 seconds of 4K R3D in a few seconds. It's absolutely crazy fast. Need to make it take longer so I can actually time it. I will PM you for media.

 

[Flo Eckhardt]

when I was thinking about a benchmark I was more focused on an actual editing project. But in the end we would anyway need timelines for many programs so your raw gpu benchmark timeline could be part of an "overall score" for r3d benchmarking a system? Would be great if you (Eric) or any other reducer would take the lead ? I would love to help where ever I can ....

 

[Eric Bowen]

Well I am definitely not to the person to ask for an editing project benchmark since I just work on the technical side. It would be more difficult to setup an actual project since there are so many variables and factors in a normal projects such as different media (which can and often changes threading and performance), FX that do not always thread well, and audio which can completely change the benchmark results based on the audio setup and what all FX are used. This is before you get to differences between CPU and GPU acceleration FX or 3rd party plugins many use. Those variables really make it difficult to isolate limitations of hardware versus software and coding. Hence why I dont want to use them for raw hardware benchmarks. Far to many have already had the wrong impressions from similar benchmarks done by others which showed very little difference above Quad core CPU's or low end video cards. Those benchmarks just did not push the GPU acceleration much at all since they used few if much at all that was GPU accelerated and the codecs of media did not thread well. Hence why the raw benchmarks such as I outlined above are far more accurate in hardware performance differences with GPU acceleration and the applications in general in their current state.

 

[Flo Eckhardt]

Well I am definitely not to the person to ask for an editing project benchmark since I just work on the technical side. It would be more difficult to setup an actual project since there are so many variables and factors in a normal projects such as different media (which can and often changes threading and performance), FX that do not always thread well, and audio which can completely change the benchmark results based on the audio setup and what all FX are used. This is before you get to differences between CPU and GPU acceleration FX or 3rd party plugins many use. Those variables really make it difficult to isolate limitations of hardware versus software and coding. Hence why I dont want to use them for raw hardware benchmarks. Far to many have already had the wrong impressions from similar benchmarks done by others which showed very little difference above Quad core CPU's or low end video cards. Those benchmarks just did not push the GPU acceleration much at all since they used few if much at all that was GPU accelerated and the codecs of media did not thread well. Hence why the raw benchmarks such as I outlined above are far more accurate in hardware performance differences with GPU acceleration and the applications in general in their current state.

Ok got it ! I also saw programs not using all too much cpu cores and gpu in real world applications what made me want to establish a benchmark that helps people decide what computer they really need. I try to think about it a little more and probably lurk at the ppbm7 benchmark test ... probably start an own thread ... lets see where this goes .

 

[Eric Bowen]

I should be starting to post benchmarks tomorrow. The thread will be a running thread as I get them done since I have to make new projects for each media type. So far I have the following from community members and Adobe forums besides my own previous media. R3D 4K 24 frame, 5K 24 frame 6:1 and 7:1, and 6K 25 Frame. I also have some 2.4K DNG for reference which performs incredibly well for those who like working with Raw. If community members want some benchmarks with 30, 60, 90+ frame media or 8K media please let me know if you can give me some samples. Also any other media such as Arri or Sony more than welcome. Already have 6900K waiting to bench after done the initial 6950X. Some other platforms should be available for me to soon since all of the benchmark projects will be created and ready to go. Have some very interesting observations for Adobe's performance as well which I will post once the initial benchmarks are done. Stay tuned and thank you again for your time especially those who provided the sample files to make these benchmarks possible.

 

[Mark Burten]

Sorry Eric, we've been swamped here. Trying to see if some overcranked stuff can get uploaded this weekend. I'll do my best to get you some goods for the cause.

 

[Eric Bowen]

NP at all. Working these benchmarks around our call volume myself so taking a while. I will gladly take the media when you have time to add to the benchmarks. This will be ongoing for abit anyway with all the hardware to test and waiting for 10XX series cards to show up as well.