Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1659
Article Thumbnail

Premiere Pro CPU performance: AMD Threadripper 3990X 64 Core

Written on February 7, 2020 by Matt Bach
Share:

Introduction

AMD's new Threadripper 3990X 64 Core processor is a very impressive product, but there are very few applications (even in video editing) that are capable of taking advantage of all these cores. The positive side is that even though the 3990X has an insane number of cores, it also maintains a very respectable 4.3GHz maximum Turbo Boost frequency. This means that even if it does not result in stellar performance in any given application, it shouldn't be particularly bad either.

While we have tested the Threadripper 3990X in Photoshop, After Effects, and many other applications, most of them are lightly threaded so we never expected the 3990X to put up record numbers. Premiere Pro, on the other hand, is one of the few applications from Adobe that can take advantage of higher core count CPUs. Having 64 cores on a single CPU is something completely new to the workstation market, however, so there is no real way to know whether or not Premiere Pro will be able to effectively utilize all of these cores without running benchmarks to find out.

AMD Threadripper 3990X 64 Core Premiere Pro Review

In this article, we will be examining the performance of the AMD Threadripper 3990X 64 Core CPU in Premiere Pro as compared to the latest AMD Threadripper, AMD Ryzen, Intel X-series, and Intel 9th Gen processors. If you are interested in how this CPU performs in other applications, we also have several other articles for Photoshop, Lightroom Classic, After Effects, DaVinci Resolve, and more available on our article listing page.

If you would like to skip over our test setup and benchmark sections, feel free to jump right to the Conclusion.

Looking for a Premiere Pro Workstation?

Puget Systems offers a range of powerful and reliable systems that are tailor-made for your unique workflow.

Configure a System!

Labs Consultation Service

Our Labs team is available to provide in-depth hardware recommendations based on your workflow.

Find Out More!

Test Setup

Listed below are the specifications of the systems we will be using for our testing:

AMD Ryzen Test Platform
CPU

AMD Ryzen 9 3950X ($749)
AMD Ryzen 9 3900X ($499)
AMD Ryzen 7 3800X ($399)​​​​​​​

CPU Cooler Noctua NH-U12S
Motherboard Gigabyte X570 AORUS ULTRA
RAM 4x DDR4-2933 16GB (64GB total)
Intel 9th Gen Test Platform
CPU

Intel Core i9 9900K ($499)​​​​​​​
Intel Core i7 9700K ($385)​​​​​​​

CPU Cooler Noctua NH-U12S
Motherboard Gigabyte Z390 Designare
RAM 4x DDR4-2666 16GB (64GB total)
AMD Threadripper 3rd Gen Test Platform
CPU AMD TR 3990X ($3,990)
AMD TR 3970X ($1,999)
AMD TR 3960X ($1,399)
CPU Cooler Noctua NH-U12S TR4-SP3
Motherboard Gigabyte TRX40 AORUS PRO WIFI
RAM 4x DDR4-2933 16GB (64GB total)
Intel X-10000 Series Test Platform
CPU Intel Core i9 10980XE ($979)
Intel Core i9 10940X ($784)
Intel Core i9 10920X ($689)
Intel Core i9 10900X ($590)
CPU Cooler Noctua NH-U12DX i4
Motherboard Gigabyte X299 Designare EX
RAM 4x DDR4-2933 16GB (64GB total)
Shared Hardware/Software
Video Card NVIDIA GeForce RTX 2080 Ti 11GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit (version 1909)
Adobe Premiere Pro 2020 (Ver. 14.0.1)
PugetBench for Premiere Pro V0.86 BETA

*All the latest drivers, OS updates, BIOS, and firmware applied as of January 20th, 2019

A few notes regarding the hardware and software used in our testing:

First, since many of the platforms we are testing support different RAM speeds depending on how many sticks you use and even whether those sticks are single or dual rank, we have decided to standardize on using four 16GB sticks of whatever the "middle" supported RAM speed is (rounding up when applicable). This means that the latest Ryzen, Threadripper, and X-series platforms are using DDR4-2933 while the Intel 9th Gen is using DDR4-2666.

The second thing to note is that we are using Noctua air coolers on all of our test platforms. We are not certain yet if this cooler will have enough capacity to keep the 3990X cool when installed in our workstations (our product qualification team is still investigating that), but in our air-conditioned Labs environment on open-air testbeds, we did not encounter any significant performance degradation with this cooler versus a beefier liquid cooler.

Lastly, there are reported issues with this processor using Windows 10 Pro due to how Windows behaves after you have 64 threads. This likely won't affect Lightroom Classic performance since it can't effectively utilize anywhere near this many cores, but we have further testing using Windows 10 Pro for Workstations planned. If you want to learn more about this, Anandtech has a terrific writeup going over the issue and the problems it causes when doing head-to-head CPU performance comparisons like this.

Benchmark Results

While our benchmark presents various scores based on the performance of each test, we also like to provide the individual results for you to examine. If there is a specific task that is a hindrance to your workflow, examining the raw results for that task is going to be much more applicable than the scores that our benchmark calculated.

Feel free to skip to the next sections for our analysis of these results to get a wider view of how each CPU performs in Premiere Pro.

Premiere Pro Benchmark Analysis

A lot is going on in the chart below, so before getting into it we wanted to provide a key regarding the color scheme.

  • Light blue = Intel 9th Gen
  • Dark blue = Intel X-10000 series
  • Light red = AMD Ryzen 3rd Gen
  • Dark red = AMD Threadripper 3rd Gen

There is a lot to cover here, but to summarize our findings, the Threadripper 3990X is overall not a great pick for Premiere Pro since both the Threadripper 3970X and 3960X are as fast or faster at a much lower price point.

Getting into the details a bit more, live playback performance tends to be the #1 concern for most of our Premiere Pro customers. Here, the 3990X does OK (and does slightly beat any Intel-based CPU), but it is about 6% slower than the Threadripper 3960X 24 Core. This isn't enough of a difference that you will likely notice it in your day-to-day work, but that does make the 3990X almost 3x the cost for slightly worse performance.

Moving on to exporting, this is where AMD's Threadripper line really excels. Even here, however, the Threadripper 3990X only manages to take second place, coming in at about 4% slower than the Threadripper 3970X 32 Core. It is also only a few percent faster than the Threadripper 3960X 24 Core, so in most cases either the 3970X or 3960X is likely a better option since the much lower price of those CPUs allows you to allocate more of your budget towards more RAM, faster storage, or other upgrades that should make a more significant performance impact overall.

One thing we do not currently test that the 3990X should be great at is rendering multiple videos at once. Premiere Pro does not have this functionality natively, and even in Media Encoder you can only render to multiple files from a single source. But if you use software that can bulk transcode media in parallel, that may be an idea use-case for the 3990X.

Is the AMD Threadripper 3990X good for Premiere Pro?

While there may be some niche uses for the AMD Threadripper 3990X 64 Core within the greater video editing industry, the 3990X is underwhelming for Premiere Pro. It is certainly no slouch, but only performs roughly on par with the Threadripper 3960X 24 Core for both live playback and exporting.

Price aside, the 3990X does perform fairly well in Premiere Pro - just no better than a CPU that is 1/3 the price. If your workflow involves other applications that are able to utilize all of these cores, you at least won't be giving up much performance in Premiere Pro by using the 3990X. Unless you know for certain that the 3990X will be worth its hefty price tag, however, you may get more overall value by investing that money into more RAM, faster storage, a second dedicated workstation, or a trip to Hawaii to recharge yourself.

Keep in mind that the benchmark results in this article are strictly for Premiere Pro and that performance will vary widely in different applications. If your workflow includes other software packages (we have similar articles for Photoshop, Lightroom Classic, After Effects, DaVinci Resolve, and many others), you need to consider how the processor will perform in all those applications. Be sure to check our list of Hardware Articles to keep up to date on how all of these software packages perform with the latest CPUs.

Looking for a Premiere Pro Workstation?

Puget Systems offers a range of poweful and reliable systems that are tailor-made for your unique workflow.

Configure a System!

Labs Consultation Service

Our Labs team is available to provide in-depth hardware recommendations based on your workflow.

Find Out More!
Tags: Intel 9th Gen, Intel X-series, Intel vs AMD, AMD Ryzen 3rd Gen, AMD Threadripper 2nd Gen, Intel X-10000, Threadripper 3990X, Premiere Pro
Hwgeek

Great reviews as always, also there is nice comparison on Anandtech regarding the Win10 Pro and Win 10 workstation/enterprise that maybe interesting for you:

"From our multi-threaded test data, there can only be two conclusions. One is to disable SMT, as it seems to get performance uplifts in most benchmarks, given that most benchmarks don’t understand what processor groups are. However, if you absolutely have to have SMT enabled, then don’t use normal Windows 10 Pro: use Pro for Workstations (or Enterprise) instead.


https://www.anandtech.com/s...

Posted on 2020-02-07 15:28:12

Yea, we have a note on that in the "Test Setup" section. First, I don't think just moving to Win 10 Pro for Workstations is going to affect the results all that much since most of the apps we test are not going to be able to use anywhere near this many cores. A few of the rendering ones, yes, but something like Premiere Pro? The other thing we have to sort through is that if we do decide to do something like disabling SMT or Hyperthreading or using Win 10 Pro for Workstations, we have to do it on ALL the CPUs. There are plenty of times where disabling SMT/HT improves performance on even a 6 core CPU, so disabling it just on the 3990X is not a fair comparison. It likely will never hurt performance on 3990X since it has so many cores, but it can give it an unfair boost versus other CPUs that would also benefit from having SMT off in some cases, but work better with SMT on in others.

Same thing with using Win 10 Pro for Workstations - if we use it only on the 3990X, the questions becomes whether it would help performance on other CPUs as well. If it does, and we swap over to that OS for all our testing, then that leaves unanswered what the performance would be on Win 10 Pro (normal), which the vast majority of users are going to be using.

We're really not sure what the right direction to take is to be honest. We could test every CPU with both Win 10 Pro normal and Workstation, and with SMT/HT on and off, but that is 4x the testing which just isn't feasible for us to manage. The way I'm leaning right now is to do this testing as-is (Win 10 Pro with SMT/HT on), then have secondary articles looking at a selection of CPUs with Windows 10 Pro for Workstation as well as with SMT/HT off. Even that is a big pain since testing a second OS either requires an OS reinstall or swapping between multiple hard drives, but at least that is significantly less work than doing it for every single CPU.

Posted on 2020-02-07 15:50:11
Hwgeek

Thanks for your reply, I just mentioned that link only to hear your opinion and maybe it would help for your build systems that you gonna offer with 3990X since now 128T on "Desktop" came to realty :-).
It will be interesting to see if MS will patch Win 10 Pro also after 3990X is out.
Thanks again for the great content :-).

Posted on 2020-02-07 16:00:15

We do plan on offering the 3990X in our systems, but I don't expect we will sell too many of them. CPU-based rendering is one of the few areas we sell a decent number of systems for, and even that has been steadily moving towards GPU-accelerated rendering over the last few years.

I also doubt that Microsoft will patch Win 10 Pro - they already have a product that will work (Win 10 Pro for Workstations), and just like you have to move to a different OS version to get support for more than 2 physical CPUs, my bet is that they will just tell people to upgrade to the Workstation version.

Posted on 2020-02-07 16:03:13
Hwgeek

I gave it more thought and would be interested to hear your opinion:
Regarding future testing- I don't think you should spend the time to test each system with both OS version, I think you just need to split the OS used based on the System thread count:
For system with up-to 64 Threads: Win10 Pro.
For system with over 64 Threads: Win10 Pro Workstation.
There should be no benefit to use Win10 Pro Workstation for systems with less then 64 Threads so no one should complain, Also you could always link to the SMT ON/OFF /Win 10 Pro for Workstations article that you plan to do- so every reader would know what OS to use based on their system.
It's just that 64C/128T single CPU on HEDT changes few things, and if Win 10 Pro for Workstations is needed to get the 100% from it there is no reason that potential buyers will limit themselves with win10 Pro, Also I thinks not many know about this [me included] because beside Anandtech no one tested it.
P.S: interesting why AMD themselves not mentioning that is recommenced to use Win10 Pro Workstation for 3990X.
Thanks again.

Posted on 2020-02-08 07:45:07
Nick Lam

Do you plan on introducing a benchmark which measures the performance of H.265 random scrubbing? Sequential playback is one thing, but as an editor I'm constantly needing to pick random segments of the timeline to playback by scrubbing to get a quick preview of what I am seeing.

Posted on 2020-02-09 12:25:27

Probably not anytime soon. Scrubbing is something we want to test, but it is a lot more difficult than straight playback. Without getting into the technicalities, it just isn't something that is easily (or difficultly for that matter) done in an automated way.

Posted on 2020-02-09 16:02:41
Nick Lam

You could make this part of the test subjective. So it's a matter of taking the same work files and having the reviewer scrub through it and then have them give their opinion on their experience (i.e. how smooth and responsive it was).

Posted on 2020-02-11 19:38:03
Luc

You absolutely need to generate gopro cineform proxies. Once you've done this, scrubbing will be extremely smooth even on low end hardware. Don't be put off by the proxy process. You can generate those extremely quickly with a fast CPU.

Posted on 2020-02-11 00:44:35
Nick Lam

Yes I agree, but I am sill dreaming of a day where we can scrub H.265 or 4K H.264 without the use of proxies, which is why I am curious as to how the newer flagship CPU's would perform in this category.

Posted on 2020-02-11 19:37:16
Luc

I get what you're saying.i didn't use proxies for a long time. My editing workflow became 100x more pleasant after I configured premiere to generate proxies upfront. It's just a checkbox when creating a project. Then wait 5mn until proxies are generated .after that ,not a single stutter when scrubbing.

Posted on 2020-02-12 01:55:23
Nick Verzilli

I tell this to colleagues all the time, but they just don't get it. Very frustrating. "I shouldn't have to make proxies!" they say. "Well, you're right, but you should want to make them" is my only defense.

Posted on 2020-02-20 15:46:31
Bin

Best article for video editing!
Can i ask you a question? I got a Ryzen 5 3600 at standard clock, 2x8gb @3200MHz, 1 250GB NVMe (OS), 2 1TB 7200RPM (Media) and Radeon R5 250 and still lagging when i'm doing live playback or scrubbing a time lapse from a sequence of photos with size approx. 5000x3000 px. What should i do to improve the live playback and scrubbing performance? Thanks in advance!

Posted on 2020-03-14 07:23:27

Image sequences are going to be pretty tough to handle in real-time and upgrading your CPU is probably the only way to improve performance. The better workflow, however, is likely to either transcode the image sequence to something like ProRes 422 first, or to use proxies.

Posted on 2020-03-16 15:43:12
Bin

Thanks! I will try it

Posted on 2020-03-17 01:40:55
Nick Lam

It'd be good to know what the CPU % utilization is for the above tests. I have a XEON 2011-v4 18 core / 36 thread and all 36 logical cores are used up (100%) when exporting 1080p H.264 to Prores 422, so I'm curious to know how well CPU % utilization scales in various exports.

Obviously, if the 3990x is pretty much tied with the 3970x, does it mean it only uses around 50% CPU utilization?

Posted on 2020-05-05 19:57:50