Adobe After Effects CC 2015.3 Pascal GPU PerformanceWritten on September 6, 2016 by Matt Bach
With the launch of After Effects 2015.3, Adobe has finally introduced support for GPU acceleration via the Mercury Playback Engine. Although AE has been able to use the GPU in the past for ray tracing, this method has largely been replaced with the integration with Cinema 4D and is considered obsolete. However, with 2015.3 AE users will now be able to leverage the power of the GPU to increase the performance for a number of specific tasks.
Unfortunately, AE 2015.3 is only able to use the video card to accelerate a grand total of three effects: Lumetri color correction, Gaussian blur, and Sharpen. More effects should be added in the future, but as of the moment only users who utilize these effects will see a difference in performance.
In this article we will be looking at the performance of After Effects 2015.3 when rendering (exporting) and scrubbing the timeline with the latest Pascal video cards from NVIDIA. Specifically, we will be testing the GTX 1060 6GB, GTX 1070 8GB, GTX 1080 8GB, and the Titan X 12GB video cards. In addition, we will also be testing the Quadro M6000 24GB (even though it is based on the older Maxwell architecture) to see if there is any significant performance advantages or disadvantages to using a Quadro video card.
For our test system, we used the following hardware:
|Motherboard:||Asus X99 Deluxe II/U3.1|
|CPU:||Intel Core i7 6900K 3.2GHz (3.5-4.0GHz Turbo) Eight Core|
|RAM:||8x Samsung DDR4-2133 32GB ECC Reg. (256GB total)|
|Hard Drive:||Samsung 850 Pro 512GB SATA 6Gb/s SSD|
|OS:||Windows 10 Pro 64-bit|
|PSU:||EVGA SuperNOVA 1200W P2|
|Software:||After Effects CC 2015.3|
This hardware is similar to the hardware we list in our After Effects Recommended Systems and should be an excellent platform for our testing. While there might be faster CPU options you could use depending on exactly what you are doing in AE (more information on this is available in our Adobe After Effects CC 2015.3 CPU Comparison article), the Intel Core i7 6900K is among the best choice when working with video footage in After Effects which makes it ideal for comparing different video cards.
The different video cards we will be testing are:
|Test Video Cards|
|NVIDIA GeForce GTX 1060 6GB|
|NVIDIA GeForce GTX 1070 8GB|
|NVIDIA GeForce GTX 1080 8GB|
|NVIDIA GeForce Titan X 12GB (Pascal)|
|NVIDIA Quadro M6000 24GB (Maxwell)|
To help with consistency - and since the benchmarks we performed ran for several days - we programmed a custom script using AutoIt to start After Effects, load the relevant project, flush the memory and disk cache, time how long it takes to render with the appropriate settings or scrub the timeline, close After Effects to clear any remaining data from the system RAM, then loop while changing the project file.
The length and source of the files we will be using in our testing are:
|Video Editing Projects||Length||Source||Tested Effects|
|H.264 1080p||15 seconds
(59.94 FPS - 900 frames)
|Provided by: Jerry Berg
Barnacules Nerdgasm - YouTube
-Lumetri Color Correction
|RED 4K||20 seconds
(23.976 FPS - 480 frames)
|Provided by: Mike Pecci
Director & Photographer
In order to make our testing as accurate as possible, we used relatively simple timelines for our testing and only applied one effect at a time. In the past, we've loaded on the accelerated effects to show the maximum difference between cards, but we found that this was not representative of real-world performance gains. This way, we can accurate gauge the performance difference between the different video cards while introducing the smallest number of variables.
Right off the bat, you can easily tell that utilizing GPU acceleration results in extremely significant performance advantages compared to using the CPU alone. At worst, performance was about 1.5 times faster (Gaussian Blur), and at best was about 6 times faster (Lumetri Color Correction).
While the massive gains in performance make it look like all of the video cards perform at the same level, there is actually a little bit of variation. Compared to the GTX 1060, the GTX 1070 was on average about .3% faster (essentially the same), while the GTX 1080 and Titan X were about 1.5% faster. The Quadro M6000 was in between clocking in at about 1% faster than the GTX 1060. This is a very small difference between each card, but it was fairly consistent across each test.
The results for scrubbing the timeline may look very similar to the rendering results, although if you look closely you will see that they are actually even better. This time, the lowest performance gain we saw was about 2x (with Gaussian Blur) compared to using just the CPU which means you can scrub the timeline almost exactly twice as fast. The best we saw, however, was with Lumetri Color Correction where we saw between a 8x and 10x speedup depending on whether the source footage was RED 4K or H.264 1080p. We still are not quite at realtime playback, but at least for H.264 1080p it is getting really close.
As for the average performance between the different cards, compared to the GTX 1060 the GTX 1070 was about 1.6% faster, the GTX 1080 about 1.9% faster, and Titan X about 2.3% faster. The Quadro M6000 was similar to the GTX 1080, measuring about 1.8% faster on average than the GTX 1060.
While GPU acceleration in After Effects 2015.3 may be limited to only three effects at the moment, it is clear that it has tremendous potential to dramatically increase the productivity of After Effects users. For rending, using a GPU allowed us to render in about 2/3rd the time at minimum, and at best rendered about 6 times faster. Scrubbing the timeline was even better with between a 2x and 10x (!) increase in performance. If you use Lumetri Color Correction, this means that what used to take 10 minutes to scrub could now only take about a minute and twenty seconds!
While these performance gains are really impressive, this article is really about helping you decide what video card you should use in an After Effects workstation. With that in mind, here is the average performance gain we saw compared to the GTX 1060 (the lowest end card we tested):
This is just an average of all the tests we ran, but it tells the story pretty well. What it comes down to is that if you want every little bit of performance you can squeeze out of your system, there is a small performance advantage to using a higher end card. It really isn't much - only one or two percent - but it is there. Honestly, however, for a workstation strictly for AE you will likely be better off spending your money on getting the exact right CPU (check out our AE 2015.3 CPU Comparison) and plenty of RAM before getting a higher end video card. Of course, if you use other programs that can utilize the GPU even more effectively (such as Premiere Pro) that may impact your decision, but for AE alone the difference between the different models is small enough that a more powerful GPU shouldn't be at the top of your purchasing list.