Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1034
Article Thumbnail

Premiere Pro CC 2017.1.2 CPU Performance: Core i9 7940X, 7960X, 7980XE

Written on September 25, 2017 by Matt Bach
Share:

Introduction

Over the last 6 months, both Intel and AMD have been making sweeping changes to their CPU lineups. Starting in March, AMD launched their Ryzen 8 core CPUs that for the first time in years made them competitive in the workstation market. In June, Intel fired back with their 6-10 core Skylake-X CPUs which in most cases re-cemented Intel as the top choice largely due to a massive price drop. A few months later in mid-August, AMD responded with their 12 and 16 core Threadripper CPUs with a 8 core model quickly following a few weeks later. Then after that, Intel launched a single 12 core Skylake-X CPU (Core i9 7920X) on August 28th. And today, we have another three CPUs from Intel: the Core i9 7940X 14 Core, Core i9 7960X 16 Core, and Core i9 7980XE 18 Core.

If all these CPU launches within such a short time period makes your head spin, you aren't alone. Just keeping up with all these new CPUs is pretty much a full time job! To kick off our latest series of hardware testing, we are going to benchmark the new Intel 14, 16, and 18 core CPUs in Premiere Pro to see how they compare to the existing Intel Skylake-X CPUs as well as AMD's Threadripper CPUs. If you wish to view all of our Skylake-X CPU testing, you can find a full list here.

There are a wide variety of tasks we could test in Premiere Pro to see how these CPUs perform, but in this article we will specifically be looking at:

  1. Rendering previews
  2. Exporting
  3. Performing a Warp Stabilization
  4. Live playback performance

Our testing includes test footage with resolutions of 4K, 6K, and 8K using six different codecs (more information in the test setup section). In total, we ran nearly 90 unique tests on 9 different CPUs resulting in more than 800 data points. If you would rather skip over our analysis of the individual benchmarks, feel free to jump right to the conclusion section.

Test Setup

Our test platforms for the Skylake-X and Threadripper CPUs are listed below, but we did want to point out that the RAM configuration changed a little bit depending on the CPU. For the majority of the CPUs we used DDR4-2666 RAM but since the Core i7 7800X only natively supports DDR4-2400 we tested with that RAM instead.

Before getting into our testing, we want to point out is that while our test platforms are using a single hard drive, that is not actually what we would typically recommend to our customers. We have found in our testing that using at least a two drive configuration with the media cache and scratch files on a secondary drive can make a big impact when it comes to importing footage and tasks like conforming audio. However, since we will not be testing any of these actions in this article we opted to use a single drive simply to cut down on the number of variables.

Most of the media we will be using is available from the Sample R3D Files and transcoded to the various codecs we wanted to test.

4K

EPIC DRAGON
3840x2160
23.976 FPS

6K

WEAPON 6K
6144x3077
23.976 FPS

8K

WEAPON 8K S35
8192 x 3456
50 FPS

To test exporting and rendering previews we used a moderately complex timeline involving multiple clips, Lumetri Color, multicam footage, and some other effects like a logo overlay, Gaussian Blur and Cross Dissolves. If you want a more in-depth look at what our timelines look like, we recorded a short video explaining our test process:

Our 4K VR testing was performed using the "Sample 1 - Ring road motorbike ride" footage from the Autopano Video Benchmarking page. We tested using both some built-in Premiere Pro effects (Lumetri Color, text overlay, and cross dissolve) as well as using the Mettle Skybox 360 VR Tools and Skybox 360/VR Transitions plug-ins to apply a number of effects such as Denoise, Rotate Sphere, Sharpen, and Iris Wipe. We typically try to avoid using plug-ins in our testing, but since Premiere Pro only has basic support for VR at the moment we felt it made sense to also look at the popular Mettle plug-in for VR projects. In addition, Adobe recently acquired Mettle Skybox which means all Creative Cloud customers will be able to use this plugin for free by the end of the year.

Rendering Previews

The first task in Premiere Pro we want to look at is rendering previews. This is something you never really want to have to do since it interrupts your workflow, but if you do complex editing it is sometimes unavoidable. Because of this, being able to render previews as quickly as possible is often an important part of a Premiere Pro workstation.

Premiere Pro Skylake-X Core i9 7940X 7960X 7980XE Benchmark

[+] Show Raw Results

Since we are comparing 9 different CPUs across 16 different projects that include a range of resolutions and source codecs, it would take us a long time to go through the results one by one. Because of this, we decided to compile all the results into an overall average for each CPU compared to the Intel Core i7 7800X. As the lowest cost CPU in both the Skylake-X and Threadripper lines, it should be a great comparison point to judge the other CPUs against. If you wish to examine the raw results yourself, you can do so by clicking on the "Show Raw Results" link under the chart.

Overall, the new 14, 16, and 18 core Intel CPUs are a bit faster than the CPUs that are already available, but it is a pretty minor improvement. In fact, compared to the Core i9 7900X 10 core or Threadripper 1950X 16 core, you are only looking at about a 3-5% gain in performance. Interestingly, we actually saw a small performance drop with the Core i9 7980XE 18 core which, as you will see in the following sections, is an unfortunate trend across all our testing when using that CPU.

Exporting

Exporting is one of the biggest time sinks for a Premiere Pro user and is often the go-to metric for measuring performance. For this test, we looked at 35 different combinations of source footage and export settings. This includes 4K, 6K, and 8K resolutions along with H.264, DNxHR HQ, ProRes 422 HQ, ProRes 4444, RED, and H.265 codecs. In addition, we will also be looking at VR projects using both built-in effects as well as with the popular Mettle plugin.

Premiere Pro Skylake-X Core i9 7940X 7960X 7980XE Exporting Benchmark

[+] Show Raw Results

Just like in the previous section, since we are comparing 9 different CPUs across 35 different projects that use a range of resolutions, source codecs, and export settings, we decided to compile all the results into an overall average for each CPU compared to the Intel Core i7 7800X. If you have the time and will, feel free to examine the raw results yourself by clicking on the "Show Raw Results" link below the chart.

Once again, the Core i9 7940X, 7960X, and 7980XE show some performance improvement, but not anything drastic compared to the existing Intel and AMD CPUs. The Core i9 7940X is a decent 7% faster than the Core i9 7900X but only 3% faster than the Core i9 7920X. The Core i9 7960X is technically another percent faster on top of this, but a single percent is well within the margin of error for this test. At the top end, the Core i9 7980XE actually had a small performance drop, coming in a hair slower than the Core i9 7940X.

Warp Stabilize

While exporting and rendering previews may be the easiest and most common thing to benchmark in Premiere Pro, we have received a lot of feedback that performing a warp stabilize is another task where high performance is important for some users. We have found that the time it takes to complete a warp stabilize analysis does not vary much on the source codec, but the resolution does make a big difference so we will be testing with a 4K H.264 clip as well as an 8K ProRes 4444 clip.

Since warp stabilize is not well threaded (meaning that it does not take great advantage of multiple CPU cores), we also opted to split our test clip into multiple parts and analyze all of them at the same time to force Premiere into making more effective use of all the CPU cores. This is a trick some people use to speed up the time it takes to analyze a single clip, but it is also a great indicator of performance when you have multiple clips that need to be stabilized. With this in mind, we not only timed how long it takes to apply a warp stabilization effect to a single 10 second clip, but also how long it takes if the clip is split into 2, 4, 8, and 16 "sub-clips" that are all analyzed at the same time.

Premiere Pro Skylake-X Core i9 7940X 7960X 7980XE Warp Stabilize Benchmark

[+] Show Raw Results

While we were able to get away with an overall average in the last two sections, the results for warp stabilize depends so heavily on the number of clips you are analyzing that we decided to show both the average result with a single clip and with 16 clips. Feel free to examine the raw results if you would like to see the results for 2, 4, and 8 clips, but the results for those fell between the single and 16 clip results about as you would expect so we opted to not muddy the chart with even more data.

Starting with a single clip, the Core i9 7940X and 7960X both performed about on par with the other Skylake-X CPUs which is to be expected as they have similar Turbo Boost 3.0 frequencies. Interestingly, even though the Core i9 7980XE has the same Turbo Boost 3.0 frequency (which allows two cores to run at 4.4GHz), the Core i9 7980XE was significantly slower at stabilizing a single clip. We're not quite sure why this is, but it is possible that the Turbo Boost 3.0 software simply isn't working properly and is not assigning one of the two "strong" cores to Premiere Pro during these tests.

Stabilizing 16 clips at once allows the new CPUs to show some decent performance gains of around 10% over the existing CPUs. What is odd, however, is that the Core i9 7940X was actually the fastest for this, beating the Core i9 7960X by 4% and the Core i9 7980XE by 2%.

Live Playback

Live playback performance is a challenge for us to accurately test since whether you can play a timeline at full, half, quarter, etc. resolution is highly dependent not only on your source footage but also what effects you have applied to the timeline. To try to keep things universally applicable, we opted to test 10 different projects using 4K, 6K, and 8K footage with multiple codecs across three relatively simple timelines. What we wanted to see is if we would be able to play the timeline at either full or half resolution without dropping any frames - even if it was just one or two at the very start of playback.

Basic

  • 4 clips in series
  • No effects
  • No transitions

Lumetri Color

  • 4 clips via multicam sequence
  • Lumetri Color Correction
  • No transitions

Lumetri & Cross Dissolve

  • 4 clips via multicam sequence
  • Lumetri Color Correction
  • Cross dissolve

Premiere Pro Skylake-X Core i9 7940X 7960X 7980XE Live Playback Benchmark

[+] Show Raw Results

For a general overview of how each CPU performed for live playback, we created a scoring system based on how many timelines each CPU was able to run at either full or half resolution. Every successful playback without dropping any frames at full resolution is 1 point while every playback at half resolution is .5 points. However, if there is a codec you typically use that we happened to include in our testing, we highly recommend looking at the raw results to see how each CPU performed for that specific codec.

Using our score-based system, you can get a pretty decent idea of how each CPU might affect live playback performance. Overall, the Core i9 7940X did extremely well and is now the top CPU for live playback from an overall score perspective. The Core i9 7980XE 18 core also did fairly well in this test as it was able to match the previous top score of 21.5 from the Core i9 7900X.

Conclusion

Based on our testing, one thing is very clear: we are definitely hitting a wall when it comes to how many CPU cores the current version of Premiere Pro can take advantage of. We saw evidence of this not long ago when we looked at whether you should use a Dual Xeon workstation for Premiere Pro (hint: don't) but until these high core count CPUs became available it was unclear whether that was a high core count issue or simply an issue caused by the complexity of dual CPU setups..

Technically, these new CPUs are faster in Premiere Pro, but what is surprising is how minor the performance improvements tended to be:

Premiere Pro Skylake-X Core i9 7940X 7960X 7980XE Overall Benchmark Results

What it comes down to is that at the very top end, the Core i9 7980XE 18 core simply isn't a great choice for Premiere Pro. In a vacuum, the performance is great, but it is slightly out-performed by both the 14 and 16 core models which means there is little reason to pay the extra money for it. Even the Core i9 7960X 16 core you would have a tough time arguing for as it was at most 1% faster than the Core i9 7940X 14 core, but much worse at live playback. There are certainly some use-cases where these CPUs would make sense - such as if you also do a lot of 3D rendering - but Premiere Pro simply can't take advantage of the extra CPU cores.

Overall, we would say that the Core i9 7940X 14 core is the only new CPU that might actually have a place in a typical Premiere Pro workstation. It isn't significantly faster for exporting and rendering previews than the Core i9 7900X or Core i9 7920X, but the improved live playback performance alone will make it a great choice for some users.

Tags: Premiere Pro, Skylake-X, Threadripper, 7980XE, 7960X, 7940X
CybrCat

Good review. Can you also add the i7-6950x to the comparison please?

Posted on 2017-09-25 16:15:24

We didn't include the 6950X is this round of testing, but it is in the testing we did a few months back: https://www.pugetsystems.co... Generally, the relative performance between CPUs doesn't change much over time unless they have vastly different core counts and there were major software changes, so it should be fairly accurate to simply apply those results to this testing. Basically, the Core i7 6950X should be within a percent or two of the Core i7 7820X.

Posted on 2017-09-25 17:25:20
Dylan

ExtremeTech were suffering some problems with Intel Speed Boost on the 7980xe - did you experience anything similar?

Also - it would be interesting to see how the CPU's compare when overclocked

Posted on 2017-09-25 16:44:52

Oh yea, Turbo Boost 3.0 has been problematic on pretty much every Skylake-X CPU released so far. It actually isn't a CPU issue, but rather an issue with almost all motherboard BIOS's. We have a pretty good relationship with Gigabyte (who we use for most of our motherboards these days), however, so we were able to work with them to get a fixed BIOS from them before we ran our testing. I don't believe the EVGA mATX board we use ever had the issue at all, but I can't speak for ASUS, MSI or any other boards we don't carry as to whether they had the same issue or not.

At this point, we actually have a quick script that we run to ensure the CPU is running at the right speeds. It applies a load to various cores (checking for both speed on individual cores for Turbo Boost 3.0 and to ensure that the cores are running at the right speeds with all cores loaded) and logs the frequency of each core to ensure they are running at the right speeds. Really, it is just a combination of Prime95, Windows affinity, and CoreTemp. At this point, I know all our workstations are running at the right speeds, although I don't think the BIOS fixes we have are available for public download quite yet. They probably want to get the fix applied to all their boards before putting it up.

As for overclocking, there is a chance we may do some overclock testing at some point, but that is actually pretty unlikely. With Turbo Boost 3.0 basically overclocking two of the cores already, we have found that you have to push an overclock a lot further than you used to have to in order to get decent performance gains. The issue there is that going that far is definitely getting outside of a "stable" overclock in our opinion. Not bad enough that enthusiasts will stop by any means, but far enough that it doesn't really fit with the majority of our customer base. Performance is always good, but our customers tend to value reliability even more and are usually not the kind of people who want to tinker in the BIOS to tune an overclock over time. Again, we might do overclock testing sometime, but our articles are first and foremost to help ensure we are offering the right configurations to our customers - and if overclocking isn't right for them then it is fairly low on our to-do list.

Posted on 2017-09-25 17:21:16
Dylan

I am planning a build soon so will have to test your experiences for myself, but for well threaded exports like ProRes or Dnx I would expect to see a decent boost from an overclock - but I haven't seen a Turbo Boost 2 CPU in action yet so its based on dated experience.

There are anomalies in your testing that it would be good to know your opinion on - Testing 6K red basic playback produces a fail on the 7980xe but with Lumetri produces a pass... If the CPU is running correctly why is this?

Unless my understanding is incorrect Red footage is compressed RAW - and it is compressed much like a zip file would be. For this reason cores can be very important when 'unpacking' Red footage.

When it comes to a basic encode it doesn't really seem like going beyond 10-14 cores is worth it, but more high end cinema cameras are now using compressed formats to bundle up raw, and the decoding of this footage is equally as important a consideration as the encode.

Your tests with the 4k red footage (which is 11:1 - so a high compression ration compared to the 6 and 8k) should therefore show an improvement in exporting as you add cores, as unpacking the footage for the subsequent encode should scale across the cores - however we actually see that beyond 10 cores the export slows down as you lose a bit of clock speed. It would be good to hear your opinion on why this is - does Red footage simply not scale across cores as well as a traditional 'zip' or is there something else at play. Were you seeing any bottlenecks elsewhere like the GPU for example or even the SSD if you were reading and exporting to the same drive?

Finally - I know you are targeting these tests at specific markets and you know your customers but I would love to see Resolve benchmarks as well.

Posted on 2017-09-25 19:08:21

The pass/fail with Lumetri is odd, but it happens on a couple CPUs and we specifically double and triple checked those results since they are so odd. No idea why it is happening, but I can guarantee that that result is accurate. Weird, but accurate. My best guess is that since Lumetri is a GPU accelerated effect, having it on does some sort of re-balancing between the CPU and GPU which frees up the CPU a bit which improves overall performance.

As for higher core counts, the main benefit I see is in terms of live playback at 6K and higher resolutions. Even then, however, the Core i9 7940X 14 core was as good or better than the 16 and 18 core CPUs so there is definitely a wall we are hitting. I don't think we are hitting any other major system bottlenecks since we are using plenty of fast RAM and running off a 3.5GB/s NVMe drive. We've done some testing in the past looking at storage speeds in Premiere Pro https://www.pugetsystems.co... and while that testing wasn't with these new CPUs, the fact that we saw no difference between a traditional SSD and a NVMe drive 4-5x the speed makes me extremely doubtful that we are bottlenecked from storage speed.

The one place we might (and it is a big might) starting to be limited is in terms of GPU power. We have done plenty of testing that shows that Premiere Pro sees little to no benefit from multiple GPUs, but with these new CPUs getting into crazy core counts that is a possibility. Re-testing multi GPU is pretty high on my list right now just in case something has changed. I really doubt it, but better to do the testing and find out for sure. I think it is more likely that dual GPU still won't be great (its hard to code for and why would Adobe spend effort on something not many people have), but when NVIDIA/AMD comes out with faster cards we might start to see more of a benefit from the 16 and 18 core CPUs.

I honestly think we are just hitting the limit of how well Premiere Pro scales using the software implementation Adobe is currently using. Keep in mind that we almost doubled the number of cores a traditional workstation might have in a matter of months, so it isn't too surprising that Adobe will need a bit of time to catch up. Also, as far as Resolve goes I'm actually starting on that right now. We were waiting for Resolve 14 to go live and now that it is that is something we definitely want to get tested. We will be putting much more of an emphasis on live playback performance with different complexities of color correction, but given the work they've done on the video editing side we'll probably do a bit of testing for that as well.

Posted on 2017-09-25 19:38:00
Dylan

Thank you very much for your feedback - In very different systems (I point this out as it may make no difference to the above discussed) dual GPU support (I think you can even go beyond dual GPU) can be a significant boon - think codecs like ArriRAW which barely stress the CPU at all but are all about the GPU. I would be very interested indeed to see how Red footage (and other compressed RAW as well) performs under these new CPUs - if it is something you are working on I may well hold off any purchases until you get a chance to have a look.

Posted on 2017-09-25 20:01:04
Jay Smith

I was hoping that the 14 core's higher clock speed would help make it a more useful CPU, and that definitely seems to be the case.

I'm also confused about the whole Lumetri 6k playback, but if just adding a corrective LUT helps playback, then thats not a bad problem to have.

8k Prores 4444 playback with these new cpus is so ridiculous!

Posted on 2017-09-25 20:17:23
Higher Order

Any lack of clock speed can be remedied by overclock & custom loop water cooling. Although Intel's choice of using pigeon poop thermal paste between IHS and the cpu die instead of solder might make things difficult.

Posted on 2017-09-26 10:22:15
Ray Gralak

Matt, I wonder if the limiting factor might be the cooling in many of these tests. Each CPU may be throttling at about the same cumulative power utilization (determined by the specific CPU's characteristics and the maximum TDP of the cooling solution used). Thus it could be that the relatively slight performance differences between the different core i9 models may be only due to variations in the quality of each individual processor.

It might make for an interesting analysis if you took five 7900X's (for example) and run some comparative tests against each other with an economy heat sink, a high-end air-cooled heat sink, and a high-end liquid cooled (CPU + VRM) solution. This testing might show any variances between the CPUs of the same type and also how much cooling effects performance.

-Ray

Posted on 2017-12-31 00:11:52

Hey Ray, cooling comparison and testing is something we do quite extensively when qualifying components for our workstations, but we tend to not do too many articles on the subject. One thing I can tell you is that thermal throttling is not an issue at all in our testing. We qualify our cooling to be able to keep the CPU running at the maximum allowed Turbo speeds (depending on the type of workload and number of cores used) in any application you may run. Premiere is actually very easy on the CPU compared to the applications we use to stress the CPU like Prime95 and Linpack.

Ensuring the CPUs are running at the correct Turbo speeds has actually been a pretty big issue for us on X299 (not because of cooling but because of BIOS issues) so we are constantly checking to make sure the CPUs are running at the correct speeds.

Posted on 2018-01-02 18:52:40
Dylan

Something else that would be interesting with these articles is a way to work out fps of the transcode speed - I was just trying to do so by looking at your test set up but the numbers I'm coming up with don't seem to make sense.

Reason for this is its a very easy metric to understand and compare with - when just comparing various CPU's a total transcode time is accurate. But its hard to compare that to a Trashcan sitting on your desk for example. If you had 6kRed>DNxHRHQ is 'x' FPS I could then run the same thing on my system and get an immediate idea of how much better or worse the performance was than 'my current' rig.

Posted on 2017-09-25 21:43:46

We could do that for exporting and rendering previews by just dividing 1440 (60 second project at ~24 FPS results in about 1440 total frames) by the the final time for the 4K and 6K projects, but we opted to stick with simple "time to completion". Mostly because since we are already doing a relative comparison with percentages there isn't much need to but also because even if we got say "20 FPS" that only applies to our exact project. Unless someone is using the same media with the same transitions and effects, it isn't really an apples to apples comparison. If we added a basic timeline with no effects or transitions it might work, but we don't plan on doing that at the moment. A single benchmark run with one set of hardware already takes something like 8-10 hours, so in order to add something like that we would have to take out quite a bit of other things we are currently testing. Otherwise, there is no way we would be able to keep up with all the hardware and software updates.

Live playback is where I wish we could do a FPS result, but from what we can tell Premiere Pro prioritizes playing back at real speed (it takes 10 seconds to play a 10 second clip, even if half the frames are dropped) instead of frame-by-frame. Unfortunately, Premiere doesn't have any sort of FPS counter like Resolve does. Live playback is where I see a FPS result being the most meaningful

Posted on 2017-09-25 22:20:28
Dylan

Does the CUDA render engine info (ctrl-shift-f11) not give the info needed for fps etc on playback?

I thought media encoded gave an fps count during render, but not used it in a while. I understand what your saying about needing to match the set up. It would be nice to have a vanilla transcodes only test but i understand your time constraints.

Posted on 2017-09-25 23:20:01

Thanks for reminding me about that - I completely forgot about it. If I remember right, the last time I tried that was a few years back and I found it to be really inconsistent from run to run. One time it might give 16FPS, the next time 18FPS, then 14FPS. That was one of the reasons why we decided to move to a pass/fail system.

I just gave it a quite test on my machine, however, and so far it seems to be much better now. So something improved that made Premiere Pro much more consistent when dropping frames. I'll have to look into this in more detail since I just tested with a bare clip and I know we'll want to include more realistic editing projects as well but so far I think this should be a decent method. I'm not looking forward to doing this testing manually (automation is pretty much the only reason we can do this large of tests) but it would be much better than the current method I'm using.

Thanks for the suggestion!

Posted on 2017-09-26 17:24:02
Higher Order

Not trying to be dick here but you have run these tests with 2400Mhz ram speed. Ryzen gets huge benefits running higher ram clock because the infinity fabric communication speed is also connected to ram speed. Increasing it automatically increases uncore speed so you get huge gains all over the place, whereas Intel gains are minimal due to its architecture. This has been documented in reviews for numerous times.

I think you should have run 4x8GB (32GB total) 3200Mhz single bank memory in quad channel mode for both systems. Ryzen would surely do a lot better then. My guess that it would sit between 14 and 16 core Skylake X cpus.

That's my planned system anyways. I bought some 4x8GB Trident Z RGB 3200 CL14 and will overclock my 1950x to 4Ghz + slap an EK monoblock on it for 7/24 water cooled use. Since Threadripper is binned, 4Ghz is guaranteed they say.

Posted on 2017-09-26 08:42:18

If you are talking about Threadripper (we didn't test normal Ryzen in this article), DDR4-2666 is the officially supported RAM speed for those CPUs. Anything higher is technically overclocking which we tend to avoid unless absolutely necessary. I actually have a blog post on this that I was going to publish when we launched our Threadripper workstations in the next few days, but I'll go ahead and make it viewable if you have the URL. It explains some of the constraints we have as a business selling to end users that people like you who build their own PCs don't necessarily have to take into account: https://www.pugetsystems.co...

One thing I am curious about, do you have links to any reviews/testing that shows Threadripper (not normal Ryzen) performing much faster with higher speed RAM? I could only find a handful of tests and the majority of them just looked at 3D Rendering in things like Cinebench. Even in that ideal situation, they all showed at most a 6% gain with DDR4-3200 over DDR4-2666.

Posted on 2017-09-26 17:09:21
Higher Order

It looks like those are all about Ryzen, rather than Threadripper. In the several articles I've read covering Threadripper, the biggest difference from just increasing memory frequency was around the 6% gain Matt mentioned. Ryzen seems to get more benefit, at least in some applications, though many of the tests in those articles are still at or under 6% increase. Moreover, some of those also talk about the CPU itself being overclocked as well - which of course will increase performance even further.

As Matt indicated, explained even better in the link in his last post, we value stability and reliability as much as performance (or even moreso!) - hence our testing at stock clock speeds and with the manufacturer rated memory frequency. Enthusiasts may choose to go higher, but we'd rather build systems we can stand behind for years :)

Posted on 2017-09-27 20:48:19
Hexxys

Well technically Threadripper *is* a Ryzen part. One that makes even more liberal use of AMD's new interconnect, which runs at 1/2 the speed of system memory. Faster RAM means a faster interconnect which means lower latency between components which means better performance.

Posted on 2017-09-28 09:47:46

Yes, they are both built on the same Zen CCX architecture... but the benchmarks seem to show less benefit for Ryzen Threadripper from faster RAM compared to Ryzen 3/5/7 (the 4-8 core CPUs). Moreover, and more importantly, AMD only rates these CPUs for up to 2667MHz memory. Going higher is technically overclocking, since part of the interconnect between the parts of the CPU runs at the same speed as the RAM... and this also voids the warranty from AMD on the processor. That is something we don't consider worthwhile, since for us and our customers the stability and longevity of systems is just as important as their performance.

Posted on 2017-09-28 16:21:06
Hexxys

Interesting. A Threadripper part is just two Zeppelin dies connected by IF, so it is only logical that it would benefit from faster RAM speeds every bit as much as a single Zeppelin die part.

It's your prerogative to avoid overclocking, though in my personal opinion you're greatly exaggerating the risks and understating the performance benefits. But, these aren't my tests and it isn't my company, so I'll just leave it at that. Good day to you!

Posted on 2017-09-28 23:19:06
pöllöpoika

No its not overclocking to put 3200mhz memory. You are just paid by Intel media ;)

Posted on 2018-07-31 17:05:48

I know that a lot of users put in fast 3200MHz (or even higher) memory, but in order to have full support from hardware manufacturers we have to stick with what they officially support. Here is info about AMD's memory specifications for Ryzen Gen 2 processors:

https://www.pugetsystems.co...

Posted on 2018-08-02 19:11:25
pöllöpoika

so you seriously think that Threadripper is different than ryzen? It's based on same Zen architecture as Ryzen. Threadripper is same as 2 1700X Ryzens clued together with inginity fabric. Threadripper is a brand ;)

Posted on 2018-07-31 17:04:00
Sc Das

Very tunnel-vision lackadaisical comparison -- Do users buy a PC for just Premier Pro ??

See this comprehensive one -- https://us.hardware.info/co...

Just add any cpu to the right "Add Product" tab in the table

Posted on 2017-09-28 18:19:46

We have articles like this for a number of different applications. Just in the last round, when these 14-18 core processors launched, we had articles for Premiere Pro, After Effects, Photoshop, Lightroom, Cinema 4D (the rendering side - basically Cinebench), V-Ray, Arnold, and Keyshot. You are right that a lot of folks use multiple programs, but rather than have one massive article that tries to address a ton of applications at once we find that breaking them up by software title makes it easier for readers to find the information that matters to them :)

Posted on 2017-09-28 18:29:52

Hey Sc Das, We actually do have a number of customers where Premiere Pro is their primary work tool, so for those users their purchasing decisions really are simply based on Premiere Pro performance. Of course, that isn't a large number of people so we also have similar articles looking at Photoshop, Lightroom, After Effects, Arnold, Keyshot, V-Ray, Remake, Revit, OctaneRender, Unreal Engine, etc. You can view all our articles at https://www.pugetsystems.co...

Keep in mind that our articles are primarily focused on helping to ensure that our customers are purchasing the right hardware for their needs and are not really intended to be an all-encompassing hardware review. We are planning on doing shorter "summary" type articles or videos that look at the big picture in the future, but what we have learned from our customers is that most professionals are really only concerned about performance in one or maybe two applications. A video editor might be 99% concerned with performance in Premiere Pro and while they do use Photoshop or After Effects in their daily work, performance in those applications is not critical. A the same time, a Photographer might really only care about Lightroom performance even though they dabble in Premiere Pro.

What it comes down to is that we want to really focus on performance in specific professional applications since that is where we really see a lack of information. General performance is already covered very comprehensively by other sites like Toms Hardware, Anandtech, Hardware.info, and many others, so we don't see much of a need to repeat their great work.

Posted on 2017-09-28 18:35:30
CobaltGamer

So if you are a video editor the best Skylake-X CPU you should buy is the i9-7940x

Posted on 2017-09-30 22:18:44
Jamie Grant

I currently have the 7820X and i am dropping frames in live preview with gh51080p LONG GOP footage. So i'm guessing the 7940x will be able to perform a lot better.

Posted on 2017-10-02 06:48:10

Yes, the i9-7940X is the best CPU you can purchase right now for Premiere Pro at least. If you do a lot of work in other programs like After Effects then it might not be, but if Premiere is where you need the performance than that is as good as you can get right now.

Posted on 2017-10-02 18:00:24
pöllöpoika

Buy threadripper, this review is fake.

Posted on 2018-07-31 17:07:00
☠☠☠

Hey Matt! Would it be possible for you to add the Xeon W-2155 10-core CPU to these new Adobe tests that you did on Sep 25? I see that William George did some GPU rendering tests using the W-2155 with the C422 Gigabyte MW51-HP0 motherboard on Sep 6, and I am very highly interested in seeing how the 10-core i9-7900X compares to the 10-core Xeon W-2155 with the Adobe software that you tested for the HCC i9 CPUs on Sep 25. Thanks!!!!

Posted on 2017-10-01 04:08:44

Right now we don't have any plans to test the Skylake-W CPUs in Adobe. William did some testing with them more due to the couple extra PCI-E lanes they have compared to the normal Skylake-X CPUs to see if that made a difference (it didn't) but at the moment we don't plan on carrying them really at all. The main difference between the two CPU lines is that the Xeon Skylake-W CPUs have 4 more PCI-E lanes (not an issue 99.9% of the time), supports ECC RAM (which rarely matters for workstations these days), and doesn't have Turbo Boost 3.0. They do have higher Turbo Boost 2.0 speeds, however, so the Skylake-W CPUs might be slightly faster for highly parallel tasks like 3D rendering but slower for lightly threaded tasks. Premiere Pro is kind of right in between so they should perform pretty much the same.

The problem with the Xeon W CPUs is that they are much more expensive. Just as an example, the 10 core Xeon W-2155 is $1440 versus the Core i9 7900X at $999. So even if the Xeon W-2155 ended up being a tiny bit faster, for the same cost you could get a Core i9 7940X which will for sure be faster. In addition, you can no longer use Xeon CPUs in a normal X299 board like you could in the past, so you are limited to the few C422 motherboards that are available. That might not be an issue if one of those boards has everything you need, but you do still have to deal with the large price disparity.

Posted on 2017-10-04 03:40:58
Ray Gralak

Matt, I would say the biggest difference is Xeon-W CPU's can support 4x the memory (512GB) of Skylake-X.

-Ray

Posted on 2017-12-30 23:55:23

RAM support is actually kind of interesting. The official spec for the Core i7/i9 CPUs is that they only support up to 128GB of RAM, but that is actually an artificial chipset limitation, not something native to the CPU. There are a number of X299 boards out there that can use Registered RAM to allow the use of up to 512GB on Core i7/i9 CPUs including the Gigabyte X299 Gaming 7 board we use quite often: http://www.gigabyte.us/Moth... . We don't do this terribly often since most of our customers don't have a need for >128GB of RAM, but it comes up on average about once a week and we haven't had any issues so far.

The one thing you do get with the Xeon-W CPUs is ECC functionality, but to be honest with modern RAM that really isn't necessary for most workstation applications. If you are doing engineering simulations you might want that just to be safe, but for content creation RAM it doesn't do much for you.

Posted on 2018-01-02 18:46:54
davewolfs

Can older dual E5-2687W V2s still keep up with these chips?

Posted on 2017-10-01 23:01:44
Joe Roberts

Love your work. I always struggled to work out what the real cost-benefits of certain hardware upgrades for Adobe were, so this site is making my life vastly easier. Thanks!

Posted on 2017-10-02 01:49:38

I think if you factor in usage of Neat Video Reduce noise, Lumetri, Warp and Film Convert at the same time... the more cores you have the better.... Neat can utilize the cores Premiere Pro does not.

Posted on 2017-10-03 19:51:43

Hey Joe, you are exactly right that different plug-ins can really change how the software performs. We've considered including different plug-ins in our testing but we have found that there really isn't such a thing as a "common" plug-in. There are some that are popular, but even those are only used by a relatively small number of Premiere Pro users. We added Mettle to our testing specifically for VR, but beyond that there is simply way too many different plug-ins out there for us to reasonably test. It is something we might address in the future, but it opens a huge can of worms that we aren't quite ready to take on at the moment.

Posted on 2017-10-03 21:09:56
Joe Roberts

Ah, thanks for pointing that out. I forgot that those are commonly used in our office (and currently kill playback performance!). That almost certainly helps to justify one of these CPUs.

Posted on 2017-10-12 03:03:20
Jacob Pawloski

I pre-ordered the 7980xe about a week ago and then came across this article. I heavily use Warp Stabilizer every day in my production workflow. Reading a 15% drop in single warp stabilizer just made my heart drop. Can you guarantee that these results are accurate? Are you verifying Turbo 3.0 kicked on all cpu's you tested?

All other single core tests on the internet for the i9 have just about the same single core performance. Multicore performance is scaling with more cores. It just doesn't add up in my head why the 7980xe is doing a terrible job stabilizing 1 and 2 clips... Keep in mind I ONLY work with 1080 & 4k footage. I plan on a mild-moderate overclock as well. I'm thinking I need to downgrade my pre-order now....

Posted on 2017-10-04 01:17:33

Hey Jacob, I know for a fact that Turbo Boost 3.0 is properly set up and working as intended. We actually had a lot of problems with Turbo Boost 3.0 when X299 first launched that required us to work with motherboard manufacturers to fix so I've made it a habit to verify that the CPUs I'm using are Turbo-ing to the proper speeds before running my tests.

Now that you mention it though, it is strange that the 7980XE is the only Intel X-series CPU that doesn't score ~100% when stabilizing a single clip. The problem with Turbo Boost 3.0 is that it only applies to two "strong" cores on the CPU and relies on Intel's software to prioritize what tasks are run on those strong cores. It is supposed to prioritize whatever is in the foreground by default as well as any apps you set in the software (I actually do this for my testing just in case). I wonder if there may have been a problem with the warp stabilize not getting priority like it should have. Premiere itself is actually doing a lot of different things in the background even when doing a warp stabilize so it might be that these other things are monopolizing those strong cores (even though they are not a heavy load) which leaves the warp stabilize to run on one of the normal cores.

We always run multiple loops of our tests and take the fastest result, so if this is the case then it is surprisingly consistent. I'm not sure if I'll be able to investigate this immediately, but I'll try to look into it soon to see what I can discover. Based on everything I know, I suspect that the 7980XE result might be inaccurate but until I can thoroughly test it I can't be sure. However, you still might consider changing your order to a Core i9 7940X instead. Its cheaper and at least in Premiere Pro is faster pretty much across the board - especially for live playback performance.

Posted on 2017-10-04 03:29:34
Jacob Pawloski

Matt, thanks for the quick response! I occasionally run after effects which is one of the reasons I was going after the 7980xe. Performance & time saved are most important in my workflow. I'm currently running a 6900k, so even a bump from 8 to 14 cores should see a noticeable performance increase in after effects. Then again I'm 90% premiere and 10% after effects.

What are the chances Adobe will rewrite the software in premiere to take advantage of more cores? I wish Adobe would state some kind of intent in response to the core war going on with Intel and AMD... I'd buy the 7980xe if they planned to take advantage of the cores in premiere...

I don't think I necessarily liked how the playback performance was added up... I think showing the amount of dropped frames and then adding the collective as a final result would have been waaay more interesting.

I'm super appreciative that you took the time to test and share all of these results!!!

Posted on 2017-10-04 05:15:43

Make sure you check out our After Effects article for these new CPUs as well: https://www.pugetsystems.co... . Also, our slightly older article that includes the 6900K might be useful: https://www.pugetsystems.co... . The testing process between those two articles is the same so you can use the results for the i7 7900X as a go-between to see what you might see if you upgraded to the 7980XE. Unfortunately, from the testing we've performed you will most likely see a performance decrease in After Effects if you upgrade - the 7980XE should be somewhere around 8% slower than your current 6900K.

Since AE version 2015 (when they started adding GPU acceleration and removed the "render multiple frames simultaneously" feature), After Effects has actually become extremely bad at utilizing higher core counts. For most things, a simple quad core CPU is actually the fastest CPU you could use although we often recommend the Core i7 7820X if you need more than 64GB of RAM (the standard quad core CPUs only support 64GB max). The only exception to the high core count issue is if you use the "Cinema 4D" 3D renderer which can still make excellent use of higher core counts. Since you are only working with 1080p and 4K footage, the 8th Gen CPUs with 6 cores that are coming out in a few days might actually be a good fit for you. We should have articles up pretty soon after launch on Oct. 5th, so check back then to see performance numbers.

As far as Adobe optimizing for higher core counts, I'm sure that is on their radar but I wouldn't expect it anytime soon. They've been making a pretty big push on GPU acceleration which in most cases actually brings about significantly higher performance gains than optimizing for higher CPU core counts. GPUs are really good at running highly parallel tasks (like Lumetri Color Correction, blurs, etc.), so a ton of software developers (including Adobe) are simply moving these highly parallel tasks onto the GPU. The result is that the CPU is left with the tasks that are really hard - or flat impossible - to efficiently run in parallel. Because of this, I think the chances of Adobe optimizing for high core count CPUs in the near future to be pretty slim. We are already seeing higher overall performance from much less expensive workstations due to developers leveraging the GPU (which is excellent!) but it does mean that these high core count CPUs are going to become more and more of a niche product that are only useful in certain applications.

I agree about how we are currently "scoring" the playback performance. Originally I was going to do exactly as you suggested and simply log the number of dropped frames, but when I started testing it I found it to be highly inconsistent. On one playback it may drop 10 frames then the next time it would drop 100 frames. I did just rediscover the CUDA information overlay you can enable that gives a straight "frames per second" counter so I may give that a shot in the future. Unfortunately, there is no way to log that information and we rely pretty heavily on automation to run our tests. The more things we have to do manually, the longer it takes to finish the testing, which means fewer tests overall we can run in the finite time we have. Still, I have some ideas to leverage OCR (text recognition) but I've run into some issues with accuracy. If I can get that to work then I think that should end up being the best way to measure live playback performance.

Posted on 2017-10-04 06:13:17
JBravo33

Having issues with my 7980XE and turbo boost. Using asus rampage vi extreme. I’ve updates all bios and utilities and when running the Intel XTU stress or cinebench all cores are maxing out at 3400. Not sure intel boost is working or the reads are wrong my cinebench number are about the same as other reviewers 3320-3380. Any ideas? I don’t wanna attempt to overclock until I get everything settled. I did change ram to 2666. That’s the only thing I changed in bios.

Posted on 2017-10-06 18:38:54

I believe the all-core Turbo for the 7980XE is 3.4GHz so you are right on the money. I'm not sure if you looked at our spec pages to determine what frequency you should be running at, but I just checked and on the 7980XE I think we mistakenly have it listed as 3.8GHz. I'm working on getting that fixed right now and if that is what threw you off, sorry!

Since you are also right in line in Cinebench with what reviews show, I'm pretty confident your CPU is running at exactly the stock speeds it should be.

Edit: Turns out our CPU specs were correct and listed 3.4GHz. But... I had the URL links wrong in the test setup section. The 7940X, 7960X, and 7980XE were all pointing to the spec page for the 7920X. Sorry if that caused the confusion!

Posted on 2017-10-06 18:58:35
Jamie Grant

Thinking about getting the i9 7940x as my current 7820x cant really handle 6k footage. What was the temps like at stock and load with the 7940x?
very grateful for these details comparisons as it has helped a ton.

Posted on 2017-10-11 23:37:18

Temperature is actually a much harder thing to report than most people realize since everything these days has cooling fans that ramp up and down depending on load. So most of the time, CPUs actually end up running at roughly the same temperature when under similar loads. Not always the case of course, but unless the cooler itself is inadequate, the question is actually more about fan speed than temperature. As an example, if we were seeing 80C but the motherboard isn't set to really ramp up the fans until 85C, then that is actually much better than a temperature of 75C with the fans running full speed.

I can tell you that the 7940X will run a bit hotter/louder than the 7820X since it is rated at 165W versus 140W, but how much so will depend on the cooler you are using and general chassis airflow. With the Noctua 120mm coolers we use, I don't think we ever hit 80C in Premiere Pro (which I believe is the point where our motherboards start really ramping the fans) so at least in our systems you would be pretty hard pressed to tell the difference between the 7940X and 7820X.

Posted on 2017-10-11 23:54:05
Jacob Pawloski

Matt, 2018 CC update came out and they are claiming everything has been redone under the hood to make everything faster. Can you confirm if this equates to better multicore support?

https://helpx.adobe.com/lig...
https://helpx.adobe.com/pre...

Posted on 2017-10-20 01:39:40
Jacob Pawloski

Matt Bach - Warp Stabilizer is now "Warp Stabilizer VFX". I swear its a tad faster and the stabilization looks better. I heavily work with warp stabilizer on a daily basis.

I'd love to hear your thoughts about the 2018 CC update!

Posted on 2017-10-20 02:36:51

Interesting to hear they changed the warp stabilizer. Will definitely need to check that out! It looks like they are really going full blast on the 360/VR features, I'll probably need to re-do that portion of the test with some of the new features they implemented.

With that said, just a bit of warning that we probably won't be putting up any testing for this new edition of Premiere Pro for a little bit. Lightroom has some major changes so I want to get that done this week, but I'm on personal leave for quite a bit next month. Then with Thansgiving, Christmas, New Years, etc. It is probably going to be a few months before we get another set of significant articles up.

Posted on 2017-10-23 18:39:26
Ipman Chess

Most software don't use all these cores..while i'm using chess to run benchmarks..chess has no limits in using all cores/threads @100% all the time!
This asmFish engine is programmed in assembler ,is Numa-aware ,is only 114Kb and no install needed..you can download it for Windows,MacOS,ARM & Linux!
You can check my list here : http://www.ipmanchess.yolas...
For any information,just ask or sent a e-mail..
Kind regards,
Ipman.

Posted on 2017-10-27 08:43:39

Hello, I have a problem with premiere PRO and AMD 1950x. While i rendering something the CPU usage do not cross the 30%. How i can force the CPU to 100%? thanks a lot

Posted on 2017-11-09 19:45:23
lp

Thank you for all your great reviews. One question about temperature: With the same configurations, which of the ThreadRipper 1950X and the i7940X will run the highest temperatures? Do you think Adobe will optimize for more cores in the next 1-2 years? In my current setup I have some problems with my drone material from Phantom 4pro. When trying to playback 4K 60fps straight from the camera it drops almost every frame. Do you think the ThreadRipper 1950x or i7940x will handle that codec? (with the same configuration you have run the tests on) Thanks

Posted on 2017-11-10 18:46:08
Rob

So, am I to assume the 7920x is underperforming the 7900x due to lower turbo speed?

Posted on 2017-11-24 19:15:03
Billy Both

This was a great article in helping to determine most appropriate cpu to buy. I am in the market to build a new edit system and wonder if this information has changed with updated Premiere? Is there an update to this article? It would be extremely helpful.

Posted on 2018-08-01 17:20:15