Premiere Pro CC 2018: AMD Threadripper 2990WX & 2950X Performance
Written on September 19, 2018 by Matt BachIntroduction
The last time we did extensive benchmarking of AMD's Threadripper CPUs in Premiere Pro, they performed extremely well but just barely fell short of their Intel counterparts. With the launch of the new Threadripper 2990WX 32 core and the 2950X 16 core CPUs, however, it is time to give AMD another look to see if the relatively higher core counts of these CPUs will allow them to pull ahead of the Intel X-series CPUs. This will be very interesting to test as (counter to what many marketing departments want you to believe) more cores is not always better. In fact, it can be extremely difficult for developers to get their software to effectively use more than a handful of cores let alone the crazy number of cores on the new Threadripper 2990WX which can sometimes result in lower performance when you have more cores.
If you would like to skip over our test setup and benchmark result/analysis sections, feel free to jump right to the Conclusion section.
Test Setup & Methodology
Listed below are the test platforms we will be using in our testing:
To thoroughly benchmark each processor, we used a range of codecs across 4K, 6K, and 8K resolutions:
| Codec | Resolution | FPS | Camera | Clip Name | Source |
| CinemaDNG | 4608x2592 | 24 FPS | Ursa Mini 4K | Interior Office | Blackmagic Design [Direct Download] |
| RED | 4096x2304 (7:1) |
29.97 FPS | RED ONE MYSTERIUM | A004_C186_011278_001 | RED Sample R3D Files |
| RED | 6144x3077 (7:1) |
23.976 FPS | WEAPON 6K | S005_L001_0220LI_001 | RED Sample R3D Files |
| RED | 8192x4320 (9:1) |
25 FPS | WEAPON 8K S35 | B001_C096_0902AP_001 | RED Sample R3D Files |
| H.264 ProRes 422 HQ ProRes 4444 DNxHR HQ 8-bit |
3940x2160 | 29.97 FPS | Transcoded from RED 4K clip | ||
Rather than just timing a simple export and calling it a day, we decided to create six different timelines for each codec that represent a variety of different type of workloads. For each of these timelines we tested both Live Playback performance in the program monitor as well as exporting via AME with the "H.264 - High Quality 2160p 4K" and "DNxHR HQ UHD" (matching media FPS) presets.
Lumetri Color
Heavy Transitions
Heavy Effects
4 Track Picture in Picture
4 Track MultiCam
4 Track Heavy Trimming
Live Playback - Benchmark Analysis
After looking through our results, one thing that stood out to us was that the AMD Threadripper and Intel X-series CPUs appear to perform differently depending on whether the media we used was RED or not. Since this was a pretty big factor, we decided to look at the results depending on whether we were using RED media or non-RED media.
The "Live Playback Score" shown in the chart above is a representation of the average performance we saw with each processor for this test. In essence, a score of "80" would mean that on average that CPU was able to play our timelines at 80% of the tested media's FPS. A perfect score would be "100" which would mean that the system did not drop any frames even with the most difficult codecs and timelines.
There is some very interesting data here, but we will start with the results when using RED media. Here, we saw pretty good performance from the Intel CPUs, but the AMD Threadripper CPUs were a decent amount faster. At the highest-end (i9 7980XE vs TR 2990WX), Threadripper was only about 5% faster but with a more mid-range CPU (i9 7900X vs TR 2950X), Threadripper was about 14% faster.
The results with non-RED media are a bit more interesting. With few exceptions, the power of the CPU didn't make much of a difference but what did make a difference was whether the CPU was Intel or AMD. Where the Intel CPUs all scored around 90-95 points, most of the Threadripper CPUs scored around 87 points while the 2990WX only scored 80 points.
What this means is that while AMD Threadripper looks to be about 5-15% faster than Intel for RED media, the Intel X-series are about 5-10% faster for non-RED media.
AME Export - Raw Benchmark Results
[Click Here] to skip ahead to analysis section
Export with "H.264 - High Quality 2160P 4K" Preset
Export with "DNxHR HQ UHD" (matching media FPS) Preset
AME Export - Benchmark Analysis
Once again starting with the results with RED media, at the highest-end (i9 7980XE vs TR 2990WX) Intel ended up being about 4% faster. This is actually almost exactly the opposite of the live playback result where TR was 5% faster. With a more mid-range CPU (i9 7900X vs TR 2950X), however, Threadripper was just shy of 25% faster!
With non-RED media, the results are pretty similar. At the high-end, the 7980XE is about 6% faster than the 2990WX while at the mid-range the 2950X is about 19% faster than the 7900X.
If export performance is a big consideration, your choice of CPU is a bit interesting. At anything other than the highest-end, you should get quite a bit more performance from a Threadripper CPU compared to an Intel X-series. But if you are looking to get the best possible performance, the Intel Core i9 7980XE will be slightly faster than Threadripper.
Is Threadripper 2 good for Premiere Pro?
AMD's Threadripper 2 CPUs are definitely great choices for Premiere Pro, but whether they are better than the Intel X-series really depends on the type of media you work with. Overall, we would recommend Threadripper 2 if you work with RED media, but the Intel X-series may be a slightly better choice if you work with other types of media.
The chart above really doesn't tell the whole story since it is simply an overall compilation of all the live playback and Adobe Media Encoder results. To really solve the Intel vs AMD question, there are three primary comparisons we should be looking at based on the rough price of each CPU:
AMD Threadripper 2990WX vs Intel Core i9 7980XE for Premiere Pro CC
In most situations, we would give the edge to the Core i9 7980XE. For live playback, it does lag behind the 2990WX by about 5% if you use RED media, but it is 13% faster if you use any other type of media. For exporting, the 7980XE will be roughly 5% faster than the 2990WX regardless of the codec.
AMD Threadripper 2950X vs Intel Core i9 7900X for Premiere Pro CC
Here, the Threadripper 2950X is often going to be the better choice. For live playback, the 2950X is about 5% slower than the 7900X for non-RED media, but if you use RED footage it is about 15% faster. When it comes to exporting, the 2950X is about 20-25% faster than the 7900X.
AMD Threadripper 1920X vs Intel Core i7 7820X for Premiere Pro CC
This is actually very similar to the 2950X vs 7900X comparison. For live playback, the 1920X is about 6% slower than the 7820X for non-RED media, but if you use RED footage it is about 17% faster. For exporting, the 1920X is about 22% faster with RED media or 12% faster with non-RED media.
If we had to declare a winner between the Intel X-series and AMD Threadripper CPUs for Premiere Pro, Threadripper definitely takes first place. But the Intel X-series CPU are definitely a close second. Which one you should use is going to come down to the type of media you work with and how important live playback performance is to you. Threadripper has an easy win if you care about performance when exporting or if you use RED footage, but for live playback with non-RED media Intel is going to be faster. Since live playback tends to be the one area that most Premiere Pro users complain about, the ~5% higher performance from Intel (plus the fact that the Intel CPUs will typically be quieter than Threadripper) may be enough to sway many users in that direction.
Choosing the right CPU for your system is a complicated topic, and Premiere Pro is likely just one of many programs you use every day. If you want to see how the 2990WX and 2950X fare in other applications, we recommend checking out some of our other recent Threadripper articles.
Premiere Pro Workstations
Puget Systems offers a range of poweful and reliable systems that are tailor-made for your unique workflow.
Labs Consultation Service
Our Labs team is available to provide in-depth hardware recommendations based on your workflow.
What about Multi-Tasking? Now that we got 32C/64T we can put 1 project to render while working/editing another 1, Do you have Ideas how to Test it?
If I am not mistaken there are software to manage programs core usage so you can split the cores between the 2/3 programs with such powerful CPU and improve the workflow.
Edit: found the software that was mentioned in some forums regarding 2990WX reviews: https://bitsum.com/
Also it was found that Windows is the limiting factor for the 2990WX and Linux performance is much higher- so Lets hope that Microsoft gonna fix it soon :-).
Multi-tasking has always been something we would like to dive into a bit, but it is a really complicated subject that increases the number of variables exponentially. We could just limited each CPU to half or 2/3 the total cores (easy to do by setting it with Windows Affinity), but I think the amount of cores you would want free is going to change depending on how many cores in total you have.
I also saw the reports about issues with the 2990WX on Windows, but I don't think that is an issue in our testing. We did at first see low performance with that CPU, but we fixed it with a combination of updating to the latest version of Windows 10 and using a motherboard that had more robust voltage regulation for the CPU. My bet is that if you don't have a motherboard that has that voltage regulation improvements, Windows power management is limiting the power (and thus Turbo speeds) to protect the system. Linux likely just doesn't have those power management settings.
Thanks for your reply, I think testing 16C+16C on 2990WX Multi-tasking is good enough since the top Intel CPU only has 18C, this will help your customers to decide if 2990WX can improve their workflow VS intel's 7980XE or AMD's 2950X.
Since 2950X performs similar to 2990WX , it's better to give those extra 16 cores different job :-)..
Next year you gonna have even harder time since looks like TR 3000 going for up-to 64C/128T :-).
Any chance, you'll do an "enthusiast" benchmark with CPU + RAM OC instead of a "workstation" benchmark with recommended / factory settings?
Seeing what CPUs can do compared by performance per watt, per GHz or per °C would be great!
To be honest, probably not. Our primary focus here at Puget Systems is selling workstations to our customers (along with providing top-notch support to match), and our articles are a way to support that business rather than being our primary focus in and of itself. Overclocking or other things bringing the system beyond spec is something we don't feel is worth the risk in our workstations. Because of this, we tend to keep our testing within what we would actually sell to one of our customers.
Every once in a while we will do testing that goes outside that box, but right now with all the hardware launches going on we are barely able to keep up with the testing we need to do to ensure we are selling our customers the exact right hardware. When/if things slow down we may be able to go beyond that.
How about the Threadripper 2 running as a dedicated rendering and transcoding machine for Premiere Pro and After Effects projects ?
Since Adobe added IGP-enabled acceleration to Premiere , can you test the 8700k with it enabled? From another test, it makes for a BIG improvement.
Thanks!
We have a post on that actually: https://www.pugetsystems.co... . Basically, you can't compare hardware and software accelerated encoding straight up, however, since hardware encoding tends to be lower performance.
In the 2017 AMD reviews, you explicitly addressed creator mode vs game mode. No mention in the 2018 reviews. And yes, it got even more complicated. What mode was used for the 2018 testing?
We didn't deal with the whole creator vs gaming thing for this testing, we just left it stock (which would mean "Creator mode"). AMD is working on an update to that program that is supposed to go live later this month that is supposed to be much better than their current implementation, so we decided to wait for that before getting too much into it: https://community.amd.com/c... . When that goes live, we'll do some testing to see how much it helps in the applications we target.
Great review!
Would you mind testing again either with this fix from Level1Techs or when Microsoft will fix the issue?
https://youtu.be/M2LOMTpCtLA
Update (plz) the Test with Windows 10 Pro for Workstations for AMD Threadripper 2990. This will use all of the Dies in this CPU on Windows.
Windows 10 for Workstations is only needed if you have more than 2 physical CPUs. Any version of Windows 10 is able to use all the cores available in any single CPU currently on the market, and will be able to do so for the foreseeable future (I believe the limit is 256 cores/threads, but it may be different depending on the architecture) . Windows 10 for Workstations has some "performance improvements", but everything I have seen indicates that they are very minor in the real world outside a few special situations
I'm not really sure how these bench marks are done, I do know that I have a TR 1950x and it has the hardest time playing 4k and sometimes even 1080.
The TR1950x on my timeline usually lags or freezes on one clip when I move the playhead, making live play back horrible, and always crashes when I export to 4k.
Is there some setting or something Im not seeing?
64ram, gtx 1080 card
I'm still researching things myself. Seems it has a lot to do with Intel's hardware decoder, quick video sync that does an extremely good job. And looks to be getting even better in gen11... I was fully expecting to do a threadripper build, but hardware decoding is a big deal to me. It's still unclear to me the impact of GPU decoders and encoders would hide this difference in cpu graphic units. But online reviews/benchmarks do not adequately explore this so far. I'm trying to stitch things together. Even if you don't use a plex server, requirements for those systems can require several concurrent transcode streams doing both the decode and encode. This gap in the CPU's ability to not do hardware decoding for AMD is crippling, at least of modern codecs. The power of threadripper to brute force a software decode hides this, but does not scale. Getting a quardo graphics card that doesn't limit concurrent sessions is an option. But hard to test this stuff without buying the hardware first.