Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/909
Article Thumbnail

Premiere Pro CC 2017 AMD Ryzen 7 1700X & 1800X Performance

Written on March 3, 2017 by Matt Bach
Share:

Introduction

Whenever a new generation of CPUs is launched, the main question everyone wants answered is exactly how well they perform in the applications they use. In the case of AMD's Ryzen, there are also a lot of questions surrounding how they compare to the processors available from Intel - especially in programs like Premiere Pro which can utilize a high number of CPU cores. For quite a while now Intel has held a dominant position in nearly every computing market, but there is a lot of hype around Ryzen due to the fact that you can get eight CPU cores for half the cost of an Intel processor of the same size.

There are a wide variety of tasks we could test in Premiere Pro, but in this article we will specifically be looking at:

  1. Rendering previews (standard and VR projects)
  2. Exporting to 4K H.264 (standard and VR projects)
  3. Exporting to 4K DNxHR HQ
  4. Exporting to 8K H.265
  5. Performing a Warp Stabilize Analysis
Update 3/9/2017: Added benchmark results for the AMD Ryzen 7 1800X (results in original article were placeholders) and improved Warp Stabilization benchmark section.

If you would rather skip over our analysis of the individual benchmarks, feel free to jump right to the conclusion section.

We also have a number of other articles looking at the performance of the AMD Ryzen 7 1700X & 1800X CPUs in other applications including:

Test Setup

To see how the new AMD Ryzen CPUs perform in Premiere Pro, we will be testing with the following hardware:

These test configurations include three different platforms along with five different CPU models. For Premiere Pro, we would typically recommend one of the "High End" Intel CPUs since Premiere Pro can make effective use of higher number of cores. With their eight CPU cores, this puts Ryzen right in the sweet spot for Premiere Pro.

Most of the media we will be using is available from the Sample R3D Files and were 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 correction, 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 correction, 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.

Render Previews - Standard Footage

Rendering previews is something that you hope to never 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.

AMD Ryzen 7 1700X 1800X Premiere Pro 2017 Benchmark Render Previews

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
96.5% 99.3% 105.9%
Compared to
Ryzen 7 1800X
95% 97.5% 103.8%

Starting off with standard footage and projects, we found some very interesting results when rendering previews. There is quite a bit of back and forth between the various CPUs, but overall Ryzen is extremely competitive compared to the Intel CPUs. Interestingly, the 1800X was only about 2% slightly faster than the 1700X which is much less than we expected given the clock speed difference between the two CPUs.

Compared to the Intel CPUs, Ryzen was around 4-5% faster than the Core i7 7700K and a small 1-2% faster than the Core i7 6850K. Neither the 1700X or 1800X could match the i7 6900K, however, which was about 4-6% faster than the two Ryzen CPUs we tested.

Render Previews - VR Footage

VR might not be here quite yet, but it is certainly gaining in popularity so we wanted to include it in our testing. Since it is not widespread, however, we opted to keep the results separate from our testing with standard footage. Note that we opted to test both with the built-in effects and transitions along with using the Mettle plug-in. We typically avoid testing plug-ins since it opens up a huge amount of potential testing, but in the case of VR we thought it was prudent to include some testing with Mettle as that is one of the most popular plug-ins for editing VR footage.

AMD Ryzen 7 1700X 1800X Premiere Pro 2017 Benchmark Render VR Previews

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
91.7% 100.2% 108.1%
Compared to
Ryzen 7 1800X
90.1% 98.3% 106%

Overall, Ryzen was decently faster than the Core i7 7700K and pretty much matched or was a bit faster than the slightly more expensive Core i7 6850K. Just like in the previous section, Ryzen fell behind compared to the Core i7 6900K with the 1700X coming in at about 8% slower and the 1800X 6% slower than the 6900K.

Export to 1080p H.264

1080p H.264 might be replaced with 4K at some point, but for now it is still among the most popular resolutions and codecs. Unlike the 4K and 8K export testing we will be doing in later sections, we opted to stick with only 4K source footage rather than going all the way up to 6K or 8K since if you working with 8K footage you probably are not going to be terribly concerned with 1080p exporting performance.

AMD Ryzen 7 1700X 1800XPremiere Pro 2017 Benchmark Export 1080p H.264

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
98.4% 105.6% 108.6%
Compared to
Ryzen 7 1800X
97.3% 104.5% 107.6%

When exporting to 1080P H.264, we found that Ryzen was on average a bit faster than the Core i7 7700K - although this was mostly due to the great performance with 4K RED footage. If you were to take out that result, the 1800X was actually about 2.4% slower than the 7700K while the 1700X was about 4% slower.

Compared to the "High End" Intel CPUs, the Ryzen CPUs were about 5% slower than the i7 6850K but again the RED 4K test skews things for the better. If we ignore that one test, Ryzen was actually about 8-9% slower than the i7 6850K. Looking at the i7 6900K, Ryzen took about 8-9% longer to export to 1080p H.264 which incidentally is about the same as what we saw with the 6850K when we excluded the RED 4K test.

Export to 4K H.264

With the rise of 4K, we opted to spend a good portion of our testing on exporting to 4K with a wide range of source resolutions and codecs. In addition, while we will focus on exporting to 4K H.264 in this section, we will also look at 4K H.264 for VR as well as 4K DNxHR HQ 8-bit in the next two sections.

AMD Ryzen 7 1700X 1800X Premiere Pro 2017 Benchmark Export 4K H.264

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
95.9% 101.2% 109.7%
Compared to
Ryzen 7 1800X
95.7% 101% 109.5%

When exporting to 4K H.264, the results for Ryzen are a bit odd. While the 1800X should be faster than the 1700X, the actual difference in this test was only a tiny .2%. This means that if performance when exporting to 4K H.264 is your primary concern, there is very little reason to use the more expensive 1800X over the Ryzen 7 1700X CPU.

Compared to the Intel Core i7 7700K, we found that Ryzen was around 4% faster on average. However, Ryzen fell behind the i7 6850K by a single percent and was about 10% slower than the i7 6900K.

Export to 4k H.264 VR

VR is gaining in popularity and one of the most commonly used resolution and codec is 4K H.264. As we stated in the "Render Previews - VR Footage" section, we tested using both the built-in effects and transitions as well as the Mettle plug-in. We typically avoid testing plug-ins since it opens up a huge amount of potential testing, but in the case of VR we thought it was prudent to include some testing with Mettle as that is one of the most popular plug-ins for editing VR footage.

AMD Ryzen 7 1700X 1800X Premiere Pro 2017 Benchmark Export 4K H.264 VR

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
90.8% 98.1% 112.9%
Compared to
Ryzen 7 1800X
88.6% 95.6% 110.1%

Overall, Ryzen is in a bit of an odd position when exporting to 4K H.264 for VR content. It is quite a bit faster than the i7 7700K and even beats the i7 6850K by a healthy amount, but it is about 10-13% slower than the i7 6900K. For it's price point Ryzen performs great when exporting VR content, but if you are looking for better performance at a bit higher cost there are a lot of arguments for using the i7 6900K instead.

Export to 4K DNxHR HQ 8-bit

While H.264 might still be one of the most common codecs to export to, it is not a great choice if there is additional editing that needs to be done in a separate application. For that, DNxHR (and DNxHD) is a popular codec to use on PC video editing workstations.

AMD Ryzen 7 1700X 1800X Premiere Pro 2017 Benchmark Export 4K DNxHR HQ 8-bit

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
90.9% 95.7% 102.5%
Compared to
Ryzen 7 1800X
90.7% 95.6% 102.3%

Interestingly, exporting to 4K DNxHR HQ 8-bit is another time where we saw minimal performance difference between the Ryzen 7 1700X and 1800X CPUs. On average for this test, the Ryzen CPUs were about 9% faster than the i7 7700K and about 4% faster than the i7 6850K. They also did very well against the slightly more expensive 6900K coming in at just 2.5% slower.

Export to 8K H.265

8K is not common quite yet, but there are already a number of people working with 8K footage so we wanted to include at least some tests for exporting 8K footage to 8K H.265.

AMD Ryzen 7 1700X 1800XPremiere Pro 2017 Benchmark Export 1080p H.264

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
92.6% 98.1% 114.2%
Compared to
Ryzen 7 1800X
88.4% 93.6% 108.9%

To be honest, even though we didn't include the 6950X in this round of testing (since it is so much more expensive than the Ryzen CPUs), if you are working with 8K projects you most likely have the budget for the Intel Core i7 6950X CPU as that is currently the fastest overall processor for Premiere Pro. Still, it is interesting to see that Ryzen is able to outperform both the i7 7700K and the i7 6850K by a good amount in this test. The Core i7 6900K is still around 9-14% faster, but for $500-600 cheaper Ryzen actually performs better than we expected for 8K projects.

Warp Stabilize 4K H.264

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. 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 a 8K ProRes 4444 clip in the next section.

Since warp stabilize is not well threaded (it does not take great advantage of multiple CPU cores), one method people use to decrease the time it takes to complete the analysis is to split their clip into multiple parts and analyze all of them at the same time. This essentially forces Premiere to make more effective use of all the CPU cores in your system. Since this is trick employed fairly commonly, we decided to not only time 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.

Special thanks to Adrian in the comments who suggested we split our test clips into sections to improve performance!

AMD Ryzen 7 1700X 1800X Premiere Pro 2017 Benchmark Warp Stabilize 4K H.264

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
117.3% 105.3% 114.8%
Compared to
Ryzen 7 1800X
110.5% 99.5% 98.6%

Warp Stabilization is a very interesting test, especially when we split the clip up into multiple sections. With just running the stabilization on a single clip, the 7700K is much faster than the Ryzen CPUs - although Ryzen does beat the i7 6850K and 6900K. However, if we divide the clip into two the 1800X actually becomes faster than the i7 6850K and 6900K but still lags behind the i7 7700K.

Once we divide the clip into 4 or more sections, the i7 6850K and especially the i7 6900K really start to pull away from the i7 7700K and the Ryzen CPUs. Interestingly, while the 1800X does end up being faster than the i7 7700K when you divide up the clip into this many parts, the difference is only by a few percent.

Warp Stabilize 8K ProRes4444

AMD Ryzen 7 1700X 1800X Premiere Pro 2017 Benchmark Warp Stabilize 8K ProRes4444

Average Relative
Performance

Core i7 7700K Core i7 6850K Core i7 6900K
Compared to
Ryzen 7 1700X
162.6% 158.0% 183.3%
Compared to
Ryzen 7 1800X
149.7% 147.8% 171.6%

Oddly, while Ryzen did pretty decently in our 4K warp stabilization tests it did not fare so well when we increased the project resolution to 8K. This time, the i7 7700K was better than the Ryzen CPUs even when we split our test clip into multiple sections and ran the analysis on each section at the same time. Once again, the i7 6850K and i6 6900K CPUs were overall the best, at times finishing the stabilization more than twice as fast as the Ryzen CPUs.

One interesting thing to point out is that when we divided our clip into 16 sections, it actually took longer to perform the analysis with every single CPU we tested. This is probably due to the fact that we are somewhat taking a brute force approach to multi-threading so there is probably some thread scheduling issues going on. However, since this occurred on all the CPUs we tested and was not limited to just Intel or AMD we went ahead and left the result in.

Conclusion

AMD Ryzen 7 1700X 1800X Overall Premiere Pro 2017 Benchmark Performance

After going through all our testing, Ryzen is in an interesting option for a Premiere Pro workstation. It is about 6-8% faster than the Core i7 7700K for exporting and rendering previews - although the i7 7700K can have a sizable advantage for warp stabilize depending on your project resolution and whether you are willing to split your clips into multiple sections. Since the 1700X and 1800X are about 20-30% more expensive than the i7 7700K, this means that if you are on a budget you have a bit of a choice between the i7 7700K and the new Ryzen CPUs. One thing to keep in mind, however, is that in a number of lightly threaded applications like Photoshop and Lightroom we found that Ryzen is quite a bit slower than the Core i7 7700K. So the choice is to pay 20-30% more for 6-8% faster export and preview generation times, but potentially much lower performance for warp stabilize and in other applications like Photoshop.

Compared to the 6-8 core "High End" Intel CPUs, the two Ryzen CPUs we tested fared pretty well. They should be pretty much on par with the Core i7 6850K for exporting and rendering previews (for about $100-200 cheaper as well), but are on average around 30% slower for warp stabilize. If you use warp stabilize in your projects, that is likely going to be the tipping point where you should definitely use a i7 6850K over the Ryzen CPUs.

Compared to the Intel Core i7 6900K (which is about $500-600 more expensive than Ryzen), you are looking at about 7-10% lower performance for exporting and rendering previews. Once again, warp stabilization is heavily in Intel's favor with the 6900K clocking in at about 45-50% faster than Ryzen.

Overall, our take on Ryzen for Premiere Pro is that while it presents some interesting opportunities and can be competitive with Intel for many tasks, it doesn't quite earn a place on our Premiere Pro recommended systems. This isn't to say that a Ryzen workstation might not be the right fit for you, but there are a couple of reasons that (right now) it just barely misses the mark for us:

  1. If you are on a tight budget, the Intel Core i7 7700K is likely a better choice for most users. While the 7700K is a little slower than Ryzen for some tasks in Premiere Pro, but it is much faster for other tasks (like warp stabilize) and in applications like Photoshop. The i7 7700K is a bit cheaper as well which frees up some of your budget for more RAM, larger storage, etc.
  2. For higher-end workstations, the 6900K (and the 6950X for that matter) is a good amount faster than the Ryzen CPUs. In addition, the 64GB RAM limitation of the Ryzen platform is a concern as 64GB is becoming much more commonplace in our Premiere Pro workstations. With 4K and VR projects gaining steam, not having the room for future RAM expansion is not ideal.

While these compromises keep Ryzen from earning a spot in our Premiere Pro configurations, this is certainly a big step forward for AMD and we will be watching them much more closely in the future. It is exciting to have more competition and options, and for the first time in a long time, we can imagine a future where AMD will be back in our product line - even if that day isn't today.

Tags: Premiere Pro, CPU, Processor, Ryzen
Kev Parsons

Great Article!

Posted on 2017-03-04 19:12:17
Banquos Ghost

Wow!.....glad I sold my AMD stock before it crashed...." buy on the rumor, ( I did), SELL on the news,( I did ) " With all the pre-release hype,I thought there was a great possibility of Ryzen smoking all the Intel CPUs with PPro,but, alas...it was not to be. I was surprised that the 7700K Kaby Lake performed so well across the board on every task which did not involve RED 4K and higher. Like in gaming, the Kaby Lake appears to be the BEST CPU for the "regular 4K " codecs and lower footage...especially now that it is only $300.....and the tests did not even push it hard with the overclocking it can achieve on all cores.
Next, is the new 1080ti too much for the 7700K ??...can the 7700K fully drive that card,or, is the 1080ti overkill ? Is a lower grade NVidia card a better and more efficient match ?

Posted on 2017-03-05 03:15:46
PuiuCS

besides the 8K warp stabilize performance, everything seems to be good. i have no idea where your negativity is coming from.

Posted on 2017-03-17 05:53:48
Banquos Ghost

As in the past, Premiere Pro has NOT been " optimized" for use with AMD products......Puget's testing SHOWS this. Adobe CC has generated record profits this year as more users come on board....over 400 million dollars worth. Yet, AMD FAILED to make sure it's new Ryzen was fully supported and utilized by this important software vendor....even Photoshop and " lightly threaded" Adobe apps are not benefiting from Ryzen over Intel's 7700K...this is a SHAME !! Other production software has proven to work with Ryzen so well that even an over clocked $300 1700 with the provided stock cooler EQUALS the performance of Intel's $1700 flagship 10 core 6950X !!!
Furthermore, after soaring to $15.50 a share on the Ryzen hype, the AMD stock continues to tank as rumors that Vega will be inferior to the 1080 to, DESPITE the 600 million dollar purchase of AMD stock by Goldman Sachs and AMD's joining the S&P 500. Last night, in after hours trading, NVidia soared $2 a share on the sales performance of the 1080ti, while AMD sunk lower. Adobe has a cozy relationship with NVidia, as only NVidia cards provide CUDA acceleration on the long list of effects, transitions, and operations which can be as much as a ten times increase in performance! !! So, neither wants to help little AMD......too bad for US !!! If Black Magic was smart, they would IMPROVE the free Resolve to take advantage of this situation by optimizing for Ryzen and by equalling the features of PPro, then maximize Resolve's performance with THEIR cameras and codecs !

Posted on 2017-03-17 14:21:11
PuiuCS

you are asking for the impossible, it's a completely new product, the majority of specific software optimisations always come after launch. and if you work with video then going with ryzen is still a better choice than the i7 7700k. it's not like you'll be using only PPro.
please stop making stupid stock market observations as they are irrelevant. and the price isn't "tanking", it's been going up and down for a while now. FYI the R5 lineup is looking to be great. soon they will also launch the RX 500. the stock value will continue to fluctuate.

Posted on 2017-03-17 18:59:20
Banquos Ghost

"impossible" ??...really?? The main good thing about Ryzen and Premiere Pro is that the CPU runs easily at 30 to 40 % during renders, while the 7700K is almost at 100% usage. This indicates that Ryzen should have more "headroom" concerning how many video tracks it can handle,etc.
Adobe NEVER got around to "optimize" their software for AMD in the past....they let the AMD CPUs under perform by requiring certain " instructions" found only on Intel CPUs and they never allowed AMD video cards to come close to the performance of the NVidia GPUs concerning CUDA acceleration vs. Open CL. As Adobe would have even MORE users because of the cheaper Ryzen, there should have been an effort made to insure maximum performance between PPro and Ryzen....there WASN'T, and there may NEVER be...as it was in the past.

Posted on 2017-03-18 07:05:40
Corey

Realistically Adobe will optimize for the market leader which is undeniably Intel and NVidia. If AMD can snag a majority of the market Adobe will be forced to optimize for them. Until AMD get market share, there is no point optimizing for a small % of their market. Still unoptimized Ryzen does extremely well and given there is so much head room still just goes to show how powerful the chips are.

Posted on 2017-04-24 03:25:41
Banquos Ghost

I saw a YouTube video yesterday where a Ryzen owner is claiming a huge performance increase because of the most recent Adobe update. He reports a 24% reduction in render times and an increase of CPU usage from around 50% to 75%. In addition, much has been learned about Ryzen performance being linked to successful over clocking of the CPU AND the memory speed. Currently, users are getting the highest memory speeds using no more than 16GB system memory. MOBO makers are promising a new BIOS soon which will enable a 32GB kit to reach over 3000 MHz and thus unlock Ryzen's true performance for everyone. AMD has achieved a major victory here, providing high multicore performance to the public for a fraction of the cost of Intel's expensive offerings. In addition, as shown on " Linus Tech Tips ", just using ANY decent GPU with at least 4GB of video memory dramatically improved PPro performance, compared to " CPU only " Mercury playback and acceleration. With MUCH cheaper AMD choices for a GPU, the ONLY reason to choose NVidia now appears to be the list of Premiere's " CUDA only" accelerated effects and operations. Someone needs to dig specifically into those particular effects and operations to see how important they are, and what exactly is the performance difference when using them with either AMD,or,NVidia GPUs.

Posted on 2017-04-24 17:07:49
terozzz

Just use that warp so that every core does the work, bam. Ryzen is the winner again....

Posted on 2017-04-11 17:55:13
chap

if you run AE without multiprocessing it would be faster then?

Posted on 2017-05-21 18:05:48
Peter

This post aged well... lol

Posted on 2018-10-06 13:36:38
Adrian

Hello! Thank you for testing :) I wish to say something about wrap stabilizer. Most people know, that it's a background process, single threaded. That's why most people stabilize many clips at a same time. For example, if I stabilize one 30second clip it took 2min 15s to analyze it. Processor load is between 15-25%. (4core i7). But if I stabilize 4 clips with a total lenght of 30s it took 1min 20s to stabilize them all. Processor load is 100% while analyzing. Thats why i7 7700K deals better in your wrap stabilizer test than ryzen - it has more powerful single core. But the situation may be different if you get eight 10s clips side by side, check them all and add wrap stabilizer effect at all of them in same time. This can show true comparision: 8core ryzen vs 4core i7 in wrap stabilizing. If you have some time, please, do that test. I always analyze many clips at the same time. If you go google and ask how speed up wrap stabilizer, on every forum they say to stabilize at least 4 clips in same time to get 100% processor load. I know that because of nature of that process (background), while stabilizing one clip I can do other things cause 3 cores are free to work. But, when it comes to stabilize something I always have many clips to stabilize. Doing it one by one would take hudge amount of time, thats why most people are stabilizing many clips at same time, not only one.

Please, for test use 8 clips placed side by side because Ryzen has 8 cores, I'm very interested in how they deal with it when you push them to 100% load :)

Posted on 2017-03-06 11:15:24

That would be an interesting test, I'm not sure why we didn't consider doing it! I'm not sure we will be able to add it for this round (figuring out how to start multiple warp stabilizes and timing completion accurately may end up being harder than it sounds) but i like the idea for future testing. Thanks for the suggestion!

Posted on 2017-03-06 17:57:55
Ralf

Hi Matt,
Adrian is completely right here. Splitting clips before warp
stabilizing does the trick for any premiere pro users. Took you seconds
and will save you hours.

Here is a good example how it works

https://www.youtube.com/wat...

Posted on 2017-03-06 20:22:30
Geeer Fak

HOW CAN YOU BE SO INEPT? NO! that's now how it works, it's one thing to not understand how premiere works, but it's another to be told over and over that what your doing is incorrect and remain willfully ignorant, and ontop of that act like you know what your talking about. IF YOU DO THIS IS WILL WRECK YOUR CLIPS, THEY WILL NOT MATCH AND IT WILL DO THE OPPOSITE OF STABILIZE THEM, IF IT WORKED THAT WAY ADOBE WOULD HAVE DONE IT. NOW FUCK OFF YOU PIECE OF SHIT.

Posted on 2017-04-17 01:37:10

Hey now - no matter what you may disagree about, language like this will NOT be tolerated on our forums. I am giving you 24 hours to edit your post and make it more respectful; if you choose not to, I will remove the comment entirely.

Posted on 2017-04-17 04:45:36
Adrian

It would be great to see how Ryzen performs with 8 clips with wrap stabilizer vs i7, not only with one clip. Except export, stabilizing is the only thing that many PPro users are waiting for while working on their footage. Unfortunately it is something to do with every movie, so we are doing stabilization every day, on many clips. It takes some time, so any adventages are worth the money. I want to buy new PC in 2-3 months, playback 1/1 is ok on my PC I got now, applying effects, changing settings, everything is in real time. I'm waiting only for export, once a week, and for stabilize clips, everyday. I would be thankful for test with 8 or more clips. It can change my mind about what to buy next.

Posted on 2017-03-07 14:42:20

Hey Adrian, just letting you know I went ahead and took your advise about splitting our test clip into sections for warp stabilization. It was easier than I feared to automate testing for, and since I already had to re-run some testing with the 1800X I went ahead and just ran it on all the CPUs we tested in this article and updated that entire section at the same time I put in actual 1800X numbers.

The results are pretty interesting. It definitely improves performance drastically, but the i7 7700K still fared very well compared to the Ryzen CPUs. I would say Ryzen was pretty much the same with our 4K project, although the 7700K still was much faster with our 8K project. The interesting thing was how much better the i7 6850K and 6900K did - they went from being a lot slower than the i7 7700K to being a good amount faster. While we didn't include the 6950X in this article, I did run some testing on that separately since I was curious and it ended up being around 5-10% faster than the i7 6900K in most cases if you were curious.

Overall, if you were aiming for the price-point of the Ryzen CPUs I would recommend going with the 6850K instead. Pretty close to the same price but it should be roughly on par for most tasks but much faster for warp stabilize.

Posted on 2017-03-09 21:50:58
Adrian

Hello! Thank you very much for your work. I have now clear situation and know what to buy. Thank you also for "special thanks" for me, i'm glad I could help with anything :) I decided to go with 6800K, it is just 200MHz slower than 6850K, same cache, it has enough pcie lines for my purposes, and 6800k costs 2000pln, 6850k is about 2900pln, it is almost a 50% more for 200MHz. I want to OC that i7 to 4.0-4.3GHz. I work in Photoshop, Lightroom and PPro. Everything works fine in my PC i have now, only export/wrap stabilizing, exporting images in PS/Lr are things I'm waiting for, and after your tests its clear that 6 core i7 overclocked will be best choice. That 6800 will be faster for most things I do in that programs, and will cost less than ryzen 1800X, so my choice was simple :)

Thank you one more time. Regards!

Posted on 2017-03-11 23:23:56
Adrian

PS. You may just check 8 clips on timeline and grab wrap stabilizer effect on them, it will start stabilizing all the clips in same time. You can stop the clock when last one is ready. To be more accurate, you can use 16 or 24 clips too. Ryzen can have some adventages with 8 clips because it has 8 cores, so every clip will be analyzing by one core. With i7 every clip will use one thread. So you can give both more work to do, one test with 8 clips, one test with 16 or 24 :)

Posted on 2017-03-07 14:51:56
John Spirou

with what cpu did you had smoother playback with native H.264 footage in the timeline ?
Can you have full resolution playback with any of those cpus (H.264 footage again)
?

Posted on 2017-03-07 00:02:15
John Brier

My most important question! I have a terrible seven year old system where timeline playback is horrid. I don't want to experience that at all again if possible so this is my most important performance metric. I don't care how long it takes to export.

Posted on 2017-04-10 20:40:51

We didn't include live playback testing the Ryzen articles, but we did in a more recent Mac vs PC performance test. Here is a link to the appropriate section of that article, which will hopefully help answer your questions:

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

Posted on 2017-04-10 20:45:53
John Brier

Wow that was a fast response! Thank you that does help. I see the tests were all at 4k or above "Maximum resolution." I don't even use 4k yet so I can only imagine at 1080p I would have mostly full resolution and no dropped frames with these higher end i7 processors. Thank you!

Posted on 2017-04-10 20:57:10
Dragon

It would be useful to know how these CPUs all compare with more than one task. E.g. start a conversion of a half-hour 4k video clip in Premiere and then move to say Lightroom and run performance tests there while the video conversion is running in the background. At the end, also time out the video conversion. This would better reflect real usage and would help to expose the capability of the multicore processors. I suspect the 7700k would fall back in that kind of test, but the Ryzen/6900 comparison would be very instructive.

Posted on 2017-03-08 04:07:31

Hi, and thank you for doing all of this informative testing.

As I understand it, you generally recommend against CPU over-clocking because of the unfavorable performance/reliability trade-off. I wonder what you think of overclocking with respect to Ryzen?

I ask for a few reasons. One is just that AMD made such selling point of all Ryzen models being unlocked. Another is that overclocking gains affect the price-to-performance value of the offerings within AMD's own product stack, since getting 1800X-level performance from an overclocked 1700, for instance, as reviewers like Gamers Nexus (https://goo.gl/gGcE7D) have reported, would make the 1800X look like a poor choice. The last is that overclocking affects some comparisons with competing Intel offerings, insofar as it may narrow performance gaps due to Ryzen's clockspeed and IPC handicaps.

I don't disagree with you in general about overclocking, and I definitely don't have a favored side in the Intel vs AMD battle. I'd just like to hear your thoughts, since you guys bring a sobriety to this issue that is nice after watching too many frothing Youtubers!

Thanks again :-)

Posted on 2017-03-09 05:43:06

Overclocking is an interesting topic because you are right that we generally don't recommend it for our customers, but it isn't something we really dislike or have a problem with in general. In fact, we do sell overclocked systems - just only to customers who have fully accepted and understand the potential downsides associated with overclocking. What it comes down to is that if you don't mind tinkering with your system every once in a while, overclocking is perfectly fine and a great way to extend your budget a bit. But most of our customers don't want to tinker and simply want the machine to work 24/7 with minimal downtime. So for them, the small chance of there being a problem isn't worth the performance gains from overclocking.

One thing that you should keep in mind is that for whatever reason reviews tend to compare overclocked CPUs to other CPUs that are not overclocked. If you have a certain performance point you are trying to hit that could make sense but in most cases you are simply going for the most performance possible within your budget. In that case, make sure you consider how the 1700 overclocked would compare to the 1800X overclocked (or the i7 7700K, i7 6850K, etc). I haven't dug too much into overclocking results for Ryzen, however, so I can't really help out when it comes to specifics in this case.

Posted on 2017-03-09 18:58:23

Hi Matt, thank you for the thoughtful reply - cheers!

Posted on 2017-03-09 20:57:34
HammeredToast

Good work. It would've been nice, I think, to see a vid of the projects rendering though. I mean, at least for me. It's one thing to read about it, it's another to witness it.

In the end, I still think I'll end up going Ryzen. Hopefully drivers and bios updates help it out a bit more.

Posted on 2017-03-09 23:47:12

We're going to be experimenting with videos in some upcoming articles so we'll see how that goes. The problem is showing something like exporting in a video. Sitting there watching a progress bar move for a couple minutes isn't really all that exciting. We've got some ideas though, and of course are always open to any suggestions you (or anyone else for that matter) might have.

Posted on 2017-03-09 23:53:25
Max

Can you please tell me that if AMD ryzen 5 will be better than Intel i5s for after effects purpose. Please let me know

Posted on 2017-03-10 12:41:29
PuiuCS

the i5s have slightly higher clocks, but are only 4c/4t CPUs. if you are building a cheap workstation, going for the 6c/12t R5 should be better. the R5 1600 is 219$ and also has the big "good" cooler bundled. or you could go for the more expensive 1600x that has higher clocks but no cooler bundled. you can pair them up with the cheaper B350 motherboards and still be able to OC them. i don't know if you need ECC memory, but you do get support for it from AMD.
but the best deal so far looks to be the 1500x. it has good clocks, the big cooler and is an 4c/8t CPU priced at 189$.
in terms of clock speeds the difference between the i5 7600k and r5 1600x is of 200MHz and the difference between the i5 7500 and the R5 1500x is of 100MHz

Posted on 2017-03-17 05:36:13
dra6o0n

You used different RAM that has different specifications.
Two of the columns doesn't have ECC and the last one did.

Also try doing a test with SMT on and off, as Windows 10 has a scheduler issue that is still not fixed yet (otherwise there would be a lot of youtube videos and news articles telling you that it's fixed).

This scheduler issue makes the OS thinks that the SMT enabled 'threads' are cores or something.
https://mspoweruser.com/amd...

"Unfortunately it seems Windows users on the latest version of Microsoft’s desktop OS will not be able to get the maximum benefit from the high performance, good value solution.

A bug has been discovered in the Windows 10 Scheduler which limits the performance of AMD Ryzen CPUs. The issue is due to the difference between how Windows treat threading between Intel and AMD processors.

For Intel hyperthreading it appropriately prioritizes the main thread and gives secondary threads lower priority. With AMD processors it treats all the threads as equal, meaning lower priority tasks are treated the same as higher priority tasks and overall performance suffers.

Another bug also incorrectly identifies the amount of cache available per thread, with the Windows 10 Scheduler thinking each thread has 136 MB of RAM rather than the actual 20 MB available in the L2+L3 cache combined.

At present the issue is only believed to affect Windows 10. Windows has had scheduler issues in the past, so we are sure a solution is only a software update away.

In the mean time users can disable SMT in the BIOS to improve performance in gaming, but for more multi-threaded work it may be better to leave it enabled."

So if in any case that the testing eventually utilized single threading at one point, the performance of the SMT enabled Ryzen cpus will take a nose dive.

Posted on 2017-03-12 02:08:19

Only the last configuration supports ECC Registered memory, and the Intel 7700K processor had the slowest memory - so if anything, Ryzen had a slight advantage from the faster RAM speed it was tested with.

As for SMT, it is debatable whether that is part of Ryzen's lower performance. Some other sources online indicate that SMT is not causing issues with Ryzen or mis-assigning threads to its cores:

https://www.pcper.com/revie...

We do plan to re-test Ryzen in a few months, though, after the platform has had a chance to mature. We'll see if anything changes :)

Posted on 2017-03-12 04:31:26
Jay Jardin

Is there a chance to get O/C results in as well? I just feel bad for those cpus nobody runs at stock speed.

Posted on 2017-03-12 18:39:31
PuiuCS

as a quick reminder: when it comes to such workstations, your mileage can vary a lot depending on your own specific workloads. always research performance around what you are working on.

Posted on 2017-03-17 05:47:44
Jon

Thank you for testing, very helpful :)

Posted on 2017-03-21 07:37:06
Milosx

On one of the demonstration videos they were showing that 7700K's H.265 decoding at 8K had more than double the performance compared to 6th gen i7, but your results do not demonstrate that. Are you sure that the testing software is able to use the hardware acceleration for H.265 on 7700K?

Posted on 2017-03-22 22:17:32

This is specially testing Premiere Pro so I'm not sure how much the improved decoding makes a difference or if Premiere even makes use of it at all.

Posted on 2017-03-22 22:21:45

Any decoding stuff related to the CPU is probably testing the video capabilities of the Intel graphics built into the processor. That would be for media playback, not for editing.

Posted on 2017-03-22 22:36:38
Elisey Seleznev

How the hell 7700K is cheaper than R7 1700?
Z270 mb for 7700k will cost you 130$ minimum. R7 1700 motherboard can be cheapo B350 - 60 bucks or smth.
R7 1700 costs 330$, 7700k costs 350$. Also R7 1700 TDP is 65W. Does intel still cheaper?

Posted on 2017-03-23 13:07:47

We tested the R7 1700X, not the vanilla 1700. It goes for about $400, which is definitely more than the i7 7700K. As for motherboards, if you want to save money there are other Intel chipsets like H270 as well. But you get what you pay for in terms of quality, so I don't recommend cheaping out on the motherboard.

Posted on 2017-03-23 14:20:30
Darien

At the time of your post it was a true statement but 4 month later the price have drop and with a few bios updates and better memory support the Ryzen 1700x is now the better and cheaper option. As of now Microcenter got the bundle of the same 1700+ & Prime X370-Pro for $399.98 while 7700 & the cheaper than tested Z270-AR goes for 409.98 The actual ASUS PRIME Z270-A cost as much as than the entire bundle intel or Amd bundle

And the price and spupport keep getting better I got mine 1700+ & Prime X370-Pro a week before for $30 higher. I think maybe be time to recommend Ryzen for premiere as CPU option.

Posted on 2017-07-19 17:45:38

Unfortunately for AMD, Intel's release of Skylake-X has taken any wind out of the sails for Ryzen in Premiere Pro. Threadripper might change things (though I doubt it) but for now the Skylake-X chips win on price:performance ratio. Here is a link to our updated article:

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

Posted on 2017-07-19 23:22:41
Darien

What wind? The new platform has been taken pretty negative by the industry and the lower mid level cpus that have been release can't even take full advantage of the x299 platform, until the not yet release higher end cpu come to market this make no sense.

Again to surpass the 1700x you have to jump to the 8 cores which are $600 plus a new motherboard which you can't at the moment use all the pci lanes and pay minimum of $300, while for less than $450 you get a 1700x + x370 motherboard and now with better memory support.

Posted on 2017-07-20 01:17:40

Not really a surprise, but I'm going to agree with William - I don't see any reason to use Ryzen for Premiere Pro unless you simply want to support AMD. There is nothing wrong with that at all, but Intel has the lead currently from both a raw performance and value standpoint. Make sure you read that article William linked https://www.pugetsystems.co... - the performance data is much more up to date (and includes the Skylake-X and Kaby Lake-X CPUs) than what is in this article and it is what I am going to base my points off of.

(Just an FYI, I didn't mean for this to turn into such a big post, but I keep finding myself answering this exact question over and over so I decided to write an in-depth one that I can simply copy/paste in the future)

For raw performance, you have to go down to the 4 core i7 7740X on the X299 platform to get the same performance as the Ryzen 7 1800X. That is a $339 quad core Intel CPU versus a $499 8 core AMD CPU. The motherboard cost can make a big difference, but most quality X370 boards start around $100. X299, on the other hand, is around a $240 starting point. So CPU+motherboard cost comes out to $580 for Intel vs $599 for AMD. Virtually the same cost for the same performance. But if you can afford the extra $50 to go up to the i7 7800X you now have the Intel system for about $30 more than an AMD system that is 7% faster in Premiere Pro. Not to mention that if you ever want to upgrade to a higher-end CPU in the future, you could upgrade to the i7 7820X, i9 7900X or one of the CPUs that aren't out yet.

Now, if you want to take sales and special deals into consideration, that really muddies the price-to-performance argument since pricing is constantly changing. Usually, however, you can find a 1800X plus a X370 motherboard for somewhere around $500. The Skylake-X CPUs you can't find many sales on right now because they are so new, but you can get a X99 motherboard plus a Core i7 6850K for around $550. So a $50 difference in price but the i7 6850K is about 5% faster than the Ryzen 1800X in Premiere Pro which is a pretty good tradeoff. If you wanted to compare using the 1700X, that is usually around $430 for a CPU and motherboard. But at that performance level, you are comparing more against the i7 7700K which you can find along with a Z270 board for usually around $420 combined. Again, sales pricing is always changing, but really you are looking at close to price parity between Intel and AMD unless you manage to catch a major deal one way or the other.

In terms of PCI-E lanes, the i7 7740X only has 16 lanes, but honestly for most users that is more than enough especially considering the X99 chipset has an additional 24 lanes built into it. That is more than enough for a GPU and a couple M.2 drives. And if you need more lanes, the i7 7800X has 28 lanes plus the 24 on the chipset for another $50. True, you can't run two GPUs at full x16 speeds, but you can't do that on Ryzen's 24 PCIe lanes either (plus Premiere Pro sees basically no performance gain with multiple GPUs so there isn't much of a reason to do so anyway).

For RAM support, the i7 7740X supports 64GB of RAM - exactly the same as the Ryzen CPUs - whereas the i7 7800X and higher support 128GB (or 512GB if you get a motherboard that supports Reg. RAM like the GIgabyte X299 Aorus boards). In terms of RAM speed, the officially supported RAM speed for Ryzen is actually just DDR4-1866 if you want to use 4 sticks of dual rank RAM or DDR4-2133 if you use 4 sticks of single rank (source: https://community.amd.com/c... ). Compare that to the DDR4-2666 of the i7 7740X or the DDR4-2400 of the i7 7800X. Yes, you can use higher frequency RAM with Ryzen, but you can do that with Intel as well.

Really, what it comes down to is that if your budget falls within the Ryzen 1700X/1800X or Core i7 7700K/7800X range, you could go either the AMD or Intel route and get roughly the same performance for your dollar in Premiere Pro. So if you want to support AMD, go for it! As a workstation manufacturer, however, there are a number of things we have to take into consideration that a person building their own computer doesn't. If we can't get any higher performance with AMD over Intel for the same dollar and Intel has a higher performance ceiling, Intel is the way we are going to go. We could list Ryzen alongside Intel, but that would effectively double our product line which doubles the number of parts we have to qualify, support, stock, and be able to provide a solid support experience on.

Posted on 2017-07-20 03:04:11
Darien

More like if you want to spend extra money on Intel x299 for premiere instead of investing than money on a graphic card where you will see the major difference on premiere.

And please stop using outdated pricing, I'll reply to the comment with specific current pricing and there is no reason to choose a 7700k over 1700x, less performance overall for more money.

My comments was specifically about the 7700k and 1700x.

Posted on 2017-07-20 04:07:52
Banquos Ghost

As time goes on, people are discovering that the performance of Ryzen can be improved with the successful use of higher speed memory....something which is currently difficult to do, depending on the exact motherboard, the exact BIOS and microcode. One user has managed to successfully use 3600 speed RAM, but, he had to include raising the base clock to 123 MHz . However, this led to Ryzen outperforming the 7700k and the 6900k in almost all games.
Ryzen architecture is unlike Intel's in terms of connecting it's internal modules, or, " clusters". Lower memory bandwidth introduces latency and slows performance. The current use of 2166 speed memory gimps performance compared to what can be achieved with memory over 3000 speed. Hopefully, this situation will be improved soon to unleash the full Ryzen performance for everyone by easily allowing the use of 3600 speed memory. If that happens, all the above tests will need to be redone.

Posted on 2017-03-29 17:33:48
Streetguru

Hopefully Adobe can get their multi-threaded stuff in order, shouldn't the 8 core intel parts/R7 parts be destroying the 7700K in everything?

Posted on 2017-04-13 17:19:54

Unfortunately, multi-threading really isn't something that is easy to implement perfectly - especially in something as complex as Premiere Pro. Now that the GPU is getting more and more heavily thrown into the mix as well, it is harder and harder for Adobe to get their code to scale efficiently. I think what we might be seeing is that many of the calculations that scale well are being offloaded to the GPU (since modern GPUs have thousands of cores) so what is left on the CPU is tasks that are really difficult or impossible to scale efficiently. I would love to see dual Xeons and other high core count systems improve performance, but I don't think it is something that will happen in the near future. Really hope Adobe proves me wrong though!

Posted on 2017-04-17 23:15:35
Streetguru

I keep hearing that older versions of premier execute multi-threading just fine and have better scaling, may be something worth testing.

There's also the semi-random test of a single PC user setting 4c/8t to premier, and running a game with the other 4c/8t, pretty niche I suppose.

At the very least I hope you guys keep the R7 1700/X at 3800mhz in your test suite even if you don't build systems with it, because for consumers it's a great alternative to the 8 core X99 chips at $329, very useful results for them.

Posted on 2017-04-17 23:57:34

Yea, older versions did scale better. But keep in mind that GPU acceleration is much, much better today so while scaling is worse, the actual raw performance should be better as long as you have a decent GPU. So in the end you should be able to get better performance with a more affordable workstation with current software/hardware.

Multi-software testing is something we'd like to get into at some point, but it is pretty difficult to make it universal. At this point we still have a number of software packages that we want to get our benchmarking process nailed down for, but after that hopefully we can look into this.

Posted on 2017-04-18 00:09:48
Kevin OReilly

Hi Matt, one thing that I noticed while I'm using Premiere is that CPU scaling goes down specifically when using the Lumetri color effects. I'm not running anything more than a 12 thread machine, so I haven't tested to the extent you have. Something about the Lumetri code path limits the CPU/GPU scaling and this is very obvious when exporting to SD MPEG because often you don't see full utilization of either the CPU or GPU. That's not to say that all GPU accelerated effects share this limitation. Most of the GPU accelerated effects do not limit CPU scaling as far as I can tell. So for our workflow where we need long SD MPEG clips rendered quickly, we purposely do not use the Lumetri color panel or Lumetri looks. We typically use fast color correction, contrast/brightness, and obviously we are also downsampling. When we export to SD MPEG or H.264 1080p, either the CPU or GPU (depending on the system) will always be pegged at 100%. Contrast that with using the Lumetri color effects and you will often see that neither the CPU or GPU are running at 100%.

Posted on 2017-07-03 03:24:52
Florin ✓ᶜᵃʳᵉᵇᵉᵃʳ

Please redo the tests using Premiere update from April ... some guy on YT was boasting around 30% improvement on Ryzen with the new update... wonder if he`s just BS`ing or not.

Posted on 2017-04-23 09:39:02
Steve

This the vid?
https://www.reddit.com/r/Am...

Posted on 2017-04-24 17:23:17
Florin ✓ᶜᵃʳᵉᵇᵉᵃʳ

No, not that.

This one, and since the guy is technically a nobody .. i just wish some other trusted reviewers with the hardware could make the tests.

https://www.youtube.com/wat...

Posted on 2017-04-24 17:25:55
Steve

Agreed. Never heard of him.. but those results are quite substantial if true!

Posted on 2017-04-24 17:28:25
Steve

Oh wait.. here it is! Very interesting! https://www.youtube.com/wat...

Posted on 2017-04-24 17:25:38

Hello, this is great! I have read that since the latest update of premiere pro, few days ago, it works much better with Ryzen CPUs. Is that true? Thank you! :)

Posted on 2017-04-27 10:03:11

I've looked through all the update notes and even talked to some Adobe guys at NAB last week. As far as I can find, there is nothing Ryzen-specific that was changed. Performance across all hardware has likely been improved a bit, but I really wouldn't expect anything crazy. Adobe may have improved their multi-threading a bit, but that is going to apply to Intel CPUs just as much as it would to AMD CPUs (If not making the 6-10 core INtel CPUs even faster than Ryzen since their single-core performance is better). It might make the 1800X a percent or two faster than the 7700K but I wouldn't expect more than that.

Posted on 2017-05-01 18:24:52
Adrian

Hello again! I was waiting for these tests because I wanted to buy new PC with Ryzen 1800X or i7 6800K. After this arcticle was published my decision was clear - i7 6800K. Now it is time to buy new rig and.. I have some doubts.... I watched many other tests of ryzens and premiere pro and most of them show quite different results than your tests... For example, in your export test I see that i7 7700K is quite equal to ryzen 1800x when exporting to h.264, DNxHR or ProRes, even in 4K or higher resolution, ryzens are also slower than i7 extreme models. Only when exporting to RED ryzens are faster. In most tests on different websites I saw that ryzens are much faster than i7 7700K, often faster than i7 6900K even in export 1080p h.264 footage to 1080p h.264 files... Now I'm little confused.. Your test shows that when exporting 1080p/4K h.264 to 1080p/4k h.264 ryzen 1800X is slower than 7700K.. In other tests online i see that 1800X is much faster, equal or better than 6900K. Why results on other websites are so different than yours? Are they missing something? I will be grateful for answer.

Posted on 2017-05-10 17:46:23

The differences you see are most likely the result of what exactly is being tested. I don't know exactly which tests you saw so I can't comment with anything specific, but generally the other benchmarks you will find online are extremely limited in their scope. Either they just test exporting a timeline with a single H.264 clip and no effects to H.264 (basically just transcoding) or use something like the PPBM6 benchmark which has it's uses but is pretty dated and the test timelines are not very real-world.

Given the amount of variation we saw between one test and another in our own testing, this is why we try to benchmark a wide array of codecs and resolutions across a variety of tests. I would bet that if you looked up exactly what those other benchmarks were testing and compared it to what we saw with the same source/export codec and resolution the results would be a bit closer. However, also keep in mind that we test with things like Lumetri Color, cross dissolves, multicam, and other things that you would be doing in the real world - if the other benchmarks are simply testing a single clip on a timeline, there could still be significant differences between their results and ours.

Posted on 2017-05-10 18:03:22
Adrian

I think this is the problem. In many tests they don't tell us in what codec is footage, sometimes even in what resolution. But when that info is provided I see that in most cases ryzen beats 7700K/6800K if export is with cpu only software mode. When they enable gpu acceleration situation is like in your tests. You can see what I'm talking about here: http://winfuture.de/videos/... in 3:27s. GPU disabled - ryzen beats intels without problems. GPU acceleration enabled - ryzen is often slower than 7700K. This is not okay to hide info about footage and target codec/resolution, and if GPU is used in test. This creates opportunity to exalt ryzens. I'm working on 1080p/4K footage from canon/gh4, i'm exporting it to 1080p h.264. Always using GPU. In my case Ryzens are unprofitable. By reading that specially crafted tests I could think that 1800X will give me power of 6900K which is not true in my case. It is good that you do your job properly, on other sites it is a lottery to know what they are testing...

Posted on 2017-05-12 22:23:58

Turning off gpu acceleration is really odd since nobody ever runs Premiere Pro in software only mode. I think what it is is that so many reviewers are interested in the theoretical differences between CPUs and lose site of the real objective: determining which is better in the real world.

Honestly, this is why we do very targeted testing looking at specific pieces of software rather than catch-all articles that try to cover everything. Those definitely have their place (we refer to them ourselves quite often) but we prefer our software-first approach. Not as universal, but much more accurate for the software packages we cover.

Posted on 2017-05-13 00:49:45
Adrian

Good job, I'm not interested in synthetic benchmarks or other tests that have nothing in common with real life results. Thanks for your work guys!

Posted on 2017-05-14 15:31:22
AlbertS

Adrian - without knowing what other tests you refer to I would put my trust in the Puget test results.
It's always a tough call but my suggestion is hang on a bit as Intel is due to announce a new range next month in response to Ryzen. Forecast is new Kaby Lake replaces the i7-7700k and new Skylake X 6, 8, and 10 core Cpus. We are hoping for higher performance and more aggressive pricing.

Posted on 2017-05-10 19:55:45
AlbertS

Intel due to release new Kaby Lake replacing i7-7700K and new Skylake X 6, 8, & 10 core
cpus next month. X99 and 1151 sockets will be obsolete replaced by X299. Rumors are improved performance and (hopefully) better pricing in
Response to Ryzen. i7 will be replaced by i9.
They even expect even expect a new 12 core Intel Core i9-7920X with 44
PCIe Gen 3.0 lanes "which are enough for running a range of multiple
discrete GPUs and fast NVMe storage devices".
http://wccftech.com/intel-s...

If
I was a gamer I might consider Ryzen, but for editing I'll stick to
Intel. I hope AMD survives but I have my doubts. Nvidia have tramped
them in the GPU world and even the recent surge in Ryzen demand is not
enough to make them profitable yet. A lot of PC software has not been
optimized for Ryzen yet and may never be as long as Intel dominates the
market. Time will tell.

Posted on 2017-05-15 06:56:35
Karolis Sipavicius

Hi, I watched one video from tech reviewer that Adobe's April update improved Ryzen performance significantly. Do you think it would be possible to run the tests again and see if there are any changes? Thanks!

Posted on 2017-05-25 05:18:03
Aelwyn

Between CPU and GPU, does anyone happen to know what makes a bigger impact on *live* preview/playback in Premiere Pro?

Posted on 2017-05-26 02:15:45

What is best for live playback performance is actually pretty much identical to what is best for exporting - really it is simply important to have a good balance between the CPU and GPU. It changes a bit depending on how many GPU accelerated effects you are using (and how heavily they tax the GPU), but in general pairing a CPU like the i7 6950X to a GTX 1080 Ti works well. Or a i7 6900K to a GTX 1080, i7 6850K to a GTX 1070, or a i7 7700K to a GTX 1060. If you use Lumetri color correction you might want a prioritize the GPU a hair more but it is rarely necessary to pair something like a GTX 1080 Ti to a i7 6850K/7700K. Except in very unusual situations, you won't run into a situation where the CPU won't be a bottleneck in that case so a high-end GPU is simply not going to be able to be fully utilized.

Posted on 2017-05-26 03:04:47
Romey

Has there ever been a test done with Cpu/igpu+Gpu vs Cpu + Gpu? If you have a cpu with onboard graphics ability, there is usually an accelerator option that will allow you to process multimedia files better. This along with the gpu acceleration should make for some amazing numbers. I remember,a hardware site tested "quicksync" igpu vs software and quicksync destroyed software when transcoding a dvd quality file to h.264. Something like 1'1/2 minutes for software and 10 seconds for quicksync<if i="" remember="" correctly="">. This was some years ago. In 2017 things must be better so have you done any tests comparing igpu+gpu acceleration vs say, 10 core x99 platforms? I will admit, that back then it was seen that quality was not the same when using acceleration. Software was "finer". Not horribly worse was acceleration, but discernable differences when viewed with a professionals eye. I still feel that it could be something to look into since transcoding is sometimes the most time consuming thing done on video.
Thanks for listening.

Posted on 2017-05-29 06:57:35
AlbertS

Intel iGPU graphics with Quicksync is slowly being used by a few NLEs like Edius for AVC/HEVC edit and playback. Software codec decoding is CPU bound and limited multi-threading. Magix (who own Sony Vegas) claim 4k AVC, HEVC, and HEVC10 editing using Intel Graphics QuickSync Video (QSV) on a i7-770K only. No need for a 8-10 core CPU and expensive Nvidia 1080. 4k editing editing on cheaper hardware would be a breakthrough. I'd like to see it tested. I found a user on their forum who claims 4K realtime playback with 4k preview on a Skylake CPU.
https://software.intel.com/...

Posted on 2017-06-07 10:22:05
david

can you please do an updated one as adobe released an update that may push ryzen ahead of intel now? anyway you could also do a dual gpu vs single gpu as I am sure that changed as well?

Posted on 2017-05-26 07:42:49

Hey David, sorry for taking a while to respond but I wanted to make sure you saw our latest Premiere Pro CPU article that included Ryzen: https://www.pugetsystems.co... . Overall, not much of a performance difference relative to the Intel CPUs. I believe performance did increase, it just did so fairly equally for all CPUs.

As for single vs dual GPU, that is something we actually do spot checks on every 6 months or so, but we still haven't seen much if any performance gain with multiple GPUs. We typically wait until we have useful results to publish before putting out an article, but we might go ahead and do an article on it in the next couple of months just to get the information out there that there still isn't a reason to use multiple GPUs fo Premiere Pro.

Posted on 2017-07-06 18:26:12
David

Thank you for the review, it's very valuable. I was reading the discussion about warp stabilization with interest and wanted to ask if there is a clear, conscious movement from Adobe towards multi-threaded support. It is clear that if it is as easy to split a clip and put warp stabilizer on each little chunk, that such a step could be automated by an Adobe programmer (whereby warp stabilizer would split a video file into small chunks and then assign each one to a thread). Yet, we don't really see such a tendency from Adobe as compared to with BMD's Resolve for example. As professionals who are monitoring developments in the field, are you seeing active and meaningful movements towards multi-threading from Adobe? Thanks.

Posted on 2017-05-27 14:59:57

David - Multi-threading is kind of weird at the moment. This is all speculation since I haven't heard too much officially from Adobe, but my impression is that their heavy development into GPU acceleration is what is affecting their multi-threading performance. GPUs tend to do very well at tasks that are highly parallel (things that scale well across lots of cores) so what Adobe has been doing is simply moving those tasks completely off the CPU and onto the GPU. This is tasks like Lumetri color correction, blurs, etc. The result is that most of the tasks left running on the CPU are those that are really difficult (or impossible) to multi-thread efficiently. So as more and more of these parallel tasks are moved to the GPU, the overall CPU multi-threading performance in Premiere has been getting worse.

This should result in overall better performance even with cheaper workstations since a powerful GPU is much less expensive than a dual Xeon or other high CPU core count setup. I think this is really the right path to take on Adobe's part, but I totally hear you that it would be ideal if we could get both great GPU acceleration and great CPU multi-threading performance.

A side note on DaVinci Resolve, we actually started testing Resolve before NAB but decided to hold off on finishing until the new version gets released since they are supposed to be making a lot of performance improvements. We did get to compare a bunch of CPUs though, and found that with most footage (H.264, ProRes, DNxHR specifically) the CPU scaling was really similar to Premiere Pro. A 8-10 core CPU gave the best performance and they were even faster than a dual CPU setup. The only exception was with RED footage where dual Xeon gave some decent performance improvements. So having not terrific multi-threading performance isn't really just an Adobe thing, it is something we are seeing across a wide range of software packages.

Posted on 2017-06-06 22:55:48
David

Very informative. Thank you.

Posted on 2017-06-07 09:44:33
Hugo Scali

They should re-test since the performance change is quite large with memories at 3200mhz.

https://s3-us-west-1.amazon...

Posted on 2017-06-06 22:11:43

Hey Hugo, we are planning on including at least the Ryzen 1800X in our next round of Premiere Pro testing (likely when the new Intel X-series CPUs come out) and will update to DDR4-2666 at that time. It is very unlikely we will go up to DDR4-3000 or DDR4-3200, however, since those are not officially supported by AMD. Even DDR4-2666 is only supported in single rank mode, so even going up to that speed is something we typically wouldn't do. Source: https://community.amd.com/c...

I'm sure higher speed RAM would improve performance a bit, but keep in mind that our testing is first and foremost to help us determine what hardware to offer to our customers. And since going outside of spec is something we usually heavily avoid (unless there is an extremely compelling reason to do so) something like higher frequency RAM is pretty low on our list of things to test. If you build your own computer and are willing to take on whatever risks come with being our of spec, that is 100% your choice. But the majority of our customers come to us because they are not the type of people who want to tinker or even think about their computer. They just want it to work so they can get their work done - hence why reliability is even more important to us than raw performance in most cases.

Posted on 2017-06-06 22:42:20
Hugo Scali

I totally agree with the stability of workstation equipment. But the Gskill Flare X memories that are currently 3200mhz cl14 are certified by ryzen and supported by the motherboard assemblers so they have a good vote of confidence. Since it is known that currently ryzen is dependent on the frequency of RAM to improve the connection between the CCX through the infinity fabric so currently AMD included support in the microcode of agesa 1.0.0.6 frequencies of 4000mhz that is coming out in These days in the new bios for now are beta

Posted on 2017-06-06 22:53:50
lin lin

Please make sure the test is done with 2017.1.2 or with the newer version. I have notice they have optimized their multi core rendering speed on my 5960X, 5820K and the new ryzen 1700. With cc2015.2 , the CPU utilization is always 10 to 20% lower than cc2017.1.2 and it also takes longer to render the final outputs. thanks

Posted on 2017-06-26 04:51:58

We actually just published an article including the new Skylake-X CPUs alongside Broadwell-E and AMD's Ryzen 7 using Premiere Pro 2017.1.2: https://www.pugetsystems.co...

Pretty much, with Skylake-X there is not much of an argument for Ryzen 7 since the i7 7800X is faster and cheaper than the Ryzen 7 1700X and almost as fast as the Ryzen 7 1800X. The i7 7820X is also much faster than the Ryzen 7 1800X, although it is $100 more.

We did end up sticking with DDR4-2400 with Ryzen since that is the fastest that is officially supported in dual channel. We actually tried to use faster RAM just to see if it made a difference, but we had pretty significant stability issues with DDR4-2666 and above. From what I've read, this is still a somewhat common issue on that platform.

Posted on 2017-06-27 00:06:13
Hugo Scali

Hi, they should start implementing in the next tests with memories of 2667mhz since they are supported by AMD on their home page.

https://www.amd.com/es/prod...

Posted on 2017-07-23 18:13:57

Hey Hugo, AMD does list 2667MHz RAM as being supported, but note that is it listed as the "max memory speed". AMd hides it for some reason, but if you dig deeper into it AMD actually only supports DDR4-2667 if you are using just 2 sticks of single-rank RAM: https://community.amd.com/c... . So that will limit you to 32GB max and you will be very limited in what model of RAM you can use in the system.

The full list of officially supported RAM for Ryzen goes:

-DDR4-2667 if using 2 single-rank RAM sticks
-DDR4-2400 if using 2 dual-rank RAM sticks
-DDR4-2133 if using 4 single-rank RAM sticks
-DDR4-1866 if using 4 dual-rank RAM sticks

The DDR4-2400 RAM we are using is dual-rank and we are using 4 sticks of it, so we are technically already using two levels faster RAM than AMD supports. Using DDR4-1866 RAM doesn't seem fair, however, so we went ahead and tested with DDR4-2400 even though we typically only use the RAM speed that is officially supported by the manufacturer. We could get in special RAM that would meet the requirements to run it at DDR4-2667, but being limited to just two sticks is simply too restricting for us. If Ryzen was much faster than Intel at the same price points we might try to make it work, but especially since the new Skylake-X CPUs make Intel faster for Premiere Pro at either a lower or similar price level, there really isn't much motivation for us to do so. https://www.pugetsystems.co...

Posted on 2017-07-23 21:20:14
Hugo Scali

The problem with this information is that it is March 2017 and with the current bios agesa 1.0.0.6a the thing changes a lot, amd published a new list with the memories that are currently supported:

Https://www.amd.com/system/...

Posted on 2017-07-23 22:15:42
Damien

I think AMD have done pretty good considering these tests were performed on release day hardware and firmwares. There have been numerous hotfixes and mobo updates since which seems to be improving Ryzen perf. :-)

Posted on 2017-06-11 06:21:05

Hey Damien, I've been waiting to reply since I've been working on updated testing including Skylake-X and Ryzen using Premiere Pro 2017.1.2 and all the latest BIOS/driver updates: https://www.pugetsystems.co...

It is day one testing on Skylake-X so things of course might change over time, but the results are really interesting. Ryzen actually did improve since this article was published in March, but the much lower cost of the Skylake-X CPUs compared to the previous generation Intel CPUs makes them overall faster than Ryzen even at the same or lower price points. It will be very interesting to see if things change once Threadripper comes out, especially since by then we should have the 12-18 core Intel CPUs as well.

Posted on 2017-06-27 00:12:48
AlbertS

For those of us who edit 4K H.264 most NLEs struggle with the codec software decoding for timeline preview/playback which is totally CPU bound. It's not the NLE, it's the codec. H.265 is even more CPU challenging (8 core may not be enough) but gaining popularity due to better quality and smaller file sizes at the same bitrate and is already available on the GH5 and DJI 4k cameras. Many non-pros, like me, have 4k cameras and UHD TVs - but the cost of an 8 core CPu and a 1080TI makes editing hardware expensive and transcoding to an intermediate is a pain taking time and resulting in large projects to backup. The i7-7700K benchmarked above has a "free" iGPU, which rates poorly for gaming but has powerful media capabilities for realtime decoding of 4k H.264 and H.265 HEVC including HEVC10 (10bit color).
http://www.anandtech.com/sh...
"The major feature change in the Kaby Lake-U/Y media engine is the availability of full hardware acceleration for encode and decode of 4K HEVC Main10 profile videos...The Video Quality Engine also receives some tweaks for HDR and Wide Color Gamut (Rec.2020) support....Intel claims that Kaby Lake-U/Y can handle up to eight 4Kp30 AVC and HEVC decodes simultaneously. HEVC decode support is rated at 4Kp60 up to 120 Mbps"

Edius Pro has used some of the functions of the iGPU but it appears MAGIX (who now own Sony Vegas) have made full use of the Intel iGPU in Movie Edit Pro.
https://software.intel.com/...
Is it possible to edit H.264 and even H.265 on the timeline using hardware accelerated decoding, with a full res 4K preview on a 4kK monitor using only a consumer level i7-7700K and no Nvidia GPU card? If true, this would be a significant breakthrough in the world of 4K editing.
I have tried to find out if MAGIX claims are true on their forum and from support with no luck. I know it's a bit off topic, but is there any chance Puget could test MAGIX on the i7-7700K rig used above (free trial download) for two reasons:
1) to test iGPU decoding capabilities and see if it works - if so, why don't more NLEs make using the Intel provided API??
2) I have been postponing my Desktop upgrade for 4K waiting for Ryzen and i9 (both need an expensive discrete GPU card) - but maybe a Kaby Lake will do the job?
- or possibly the 6 core Coffee Lake with Intel Iris iGPU coming later this year.

I'm concerned that newer CPU technology may not solve our future NLE needs as silicon technology hit the wall about six years ago and stalled a 4-5Ghz and the once predicted (by Intel) 10Ghz CPU will not happen in the foreseeable future. More cores at higher cost and slower clock speeds benefits are limited by app software parallelism and multi-threading limitations of decoding highly compressed camera codecs. Intel's "new i9" CPUs are not new - but just based on existing Xeons with ECC removed at better prices thanks to AMD. i9s are marketed as "consumer" CPUs but what consumer needs a 16 or 18 core CPU?

Last, the HEDT market is shrinking so future 10nm and 8nm technology will be focused on the explosion in mobile devices and lower power for longer battery life, but no significant performance increases. For those interested in why we shouldn't expect much more power in the next few years see the 2011 3 part article below. New technology like carbon nano may be the answer but typically it takes new tech 4-6 years to reach production.

http://www.edn.com/design/s...

Posted on 2017-06-26 08:43:46
Jeremy Ables

I am just curious if you are going to repeat any of your tests now that the ryzen 1920/1950x cpu's have been released? If you are expecting better/different results since in the your premiere pro article you stated that premiere isn't really optimized for more than 8-10 cores? Having only seen one motherboard at the moment and the chip having 64 pcie lanes, do you think the 64gb of ram will be lifted, haven't seen any information on that yet? Last but not least none of the current AMD motherboards have thunderbolt 3 have you seen anything that says will be different, could an add in card be possible as I know asus makes one but compatibility with AMD is an issue. Finally, if threadripper does work well would this be a potential system build?

Posted on 2017-08-10 22:33:12

We definitely will be testing Threadripper 1920X and 1950X (and the 8 core 1900X when it launches), but it will probably be at least a few weeks. No promise on exactly when since we will probably be focusing on the tasks that Threadripper should excel at (like 3D rendering) first. Based on all the reviews I've seen, however, I don't expect Threadripper to beat Intel at similar pricepoints for Premiere Pro. It might be able to match it in some workflows (with RED footage is the most likely), but Premiere just values higher clock speed and single/medium threaded performance too much to make the higher core counts on Threadripper worth it. I could totally be wrong (which is why we do this kind of testing!) but I would be surprised if Threadripper ended up being a better choice than Intel for Premiere Pro.

RAM limit is actually 128GB on Threadripper (although it is probably only 64GB if using mATX) which should be more than enough unless you are working with 8K footage so I don't think RAM will be an issue. The 64 PCIe lanes is really nice, but since Premiere Pro doesn't really take advantage of multiple GPUs even the 28 lanes on the new Intel Core i7 should be more than enough unless you also want to use a PCIe sound card, RAID controller, and a couple other PCIe devices. More lanes is always nice, but I don't think that should be a major consideration for a Premiere Pro workstation except in very specific situations.

As for Thunderbolt, we haven't heard anything about it being available with Threadripper but I don't think they've said anything public one way or the other. Since Intel owns Thunderbolt, however, I wouldn't hold my breath.

Posted on 2017-08-10 22:49:54
Jeremy Ables

Thanks for replying so quickly, you confirmed/cleared up my thoughts on threadripper, but I look forward to the testing nonetheless.

Posted on 2017-08-10 22:57:13
Oni

This article needs to be updated. The AMD chips are now 20-30% CHEAPER than Intel counterparts, not more expensive. Perhaps the reviewers are waiting for the AMD 2000 series?

Posted on 2018-04-04 20:02:26

We actually have several newer articles that include the Ryzen CPUs with this one being the latest: https://www.pugetsystems.co... . On pricing, we tend to list simple MSRP since the price is constantly changing based on who you purchase the CPU from and any sales going on. Just for an example, right now the Ryzen 1800X is $340 on Newegg while the Core i7 8700K is $350. While I'm sure there are better prices elsewhere, that is essentially the same price, but the Core i7 8700K is slightly faster in Premiere Pro.

Posted on 2018-04-04 20:15:22
Oni

The article still states that AMD chips are (20-30%!!!) more expensive, and you just admitted they are now cheaper! Who are you trying to fool?

The 1700x is now $30 cheaper, on average, than the 7700k, and the 1700 (non-X) series can attain a very similar overclock for even less, which you never bother to test! And your articles state the 1700x is better on everything except warp stabilize! The reasons for the 7700k recommendation are outdated. Period.

You also don't note the performance decreases because of the Spectre and Meltdown exploits, and their patches.

I can forgive not talking about the AMD 2700/2800 series so early, which will be the same MSRP as the 7700/8700 with improved performance while Intel sits on their hands until 2019. But the other stuff? Lazy or biased.

Posted on 2018-04-11 23:32:11

Hi Oni,

I'm sorry this bothers you, but we don't go back and update everything on old articles. The article above was written over a year ago, and it has been superseded by newer content that we've put out over that span of time. Rather than trying to go back and update dozens or hundreds of older articles all the time, we simply put a warning at the top of articles that are getting a bit long in the tooth - if possible, with a link to a newer article on the same topic. If you look at the top of this article, you will see just such a notification. And if you follow that, and any additional notices like it, you will find that the most recent Premiere Pro article we have (at the time I am writing this comment) is this one:

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

That still pre-dates the Ryzen 2000-series chips, but it clearly shows Intel's Core i7 8700K being far faster than any of the original Ryzen models. The 7700K you complained about is no longer our recommendation, and hasn't been for something like half a year. Ryzen models like the 1800X and 1700X may be a little cheaper at times, but would you really want to save $50 or even $100 on a computer system if it meant losing 5-10% performance? That doesn't seem like a worthwhile trade-off to me, for such a small savings.

If you'd like to see how the newer Ryzen chips compare, stay tuned. We should have results for those out in the coming weeks. Then you can take the data and make the decision that you feel is best, whether it lines up with our conclusions or not :)

Posted on 2018-04-12 15:20:25
Oni

By your own reports that performance tradeoff is only in limited applications. So yes, absolutely personally I would save the $50-100!

I could also spend $1,000 more for 10% improvement, if it was within budget. Or $2,000 and start a render farm for 50%+ improvement.

In any case, this article and your latest one still misrepresents the current prices in the reecommendations themselves. In this specific case, the recommendation, by your own words in the article, is solely based upon a price gap that does not exist. And stopped existing only a few months past publication!

I agree with you that price fluctuations can be volatile. Which is a good reason why to never make such strong recommendations based on minor price differences, don't you think? You outdate your recommendation reasoning as soon as you go to press!

Posted on 2018-05-22 11:54:05
disqus_az3XOMsfnT

Dude, chill. The article is old. The data is old. Why are you here complaining about old data on an old article? Should they come back and update this article every time there's a price change? Do you intend on revisiting old articles about the Athlon 64 and complaining that the price and performance information is no longer accurate?

Posted on 2018-05-29 06:40:05
Oni

Cool. I'll answer.

Why?
1. It is obvious by older and newer articles an unwarranted editorial bias against AMD that is not backed up by evidence, even in their own articles. Someone needed to say something, because:

2. This suppposedly old, outdated article still has high visibility and traction. Else, why are you here, reading it? Are you a historian? Are you also researching old articles on the Athlon 64?

3. While this is nearly a year old, the article's conclusions were actually outdated very soon after publication. Typical editorial discretion and honesty would mandate an insert at the top [This article is outdated, look at new article here: ...] This is what many other services do.

Hell, they could insert a simple program to put real-time prices of the chips in inline text. But it would have made their recommendation reason here look silly, fast.

However:
4. As emphasis, even in newer articles, the same bias is apparent. Why should you trust new ones if the old were biased? The newer article linked in previous replies, actually, does not compare the same devices and does not compare using the same methods. Unlike what he alleged, it wasn't an "update" to this one.

As for chilling. I'm at peak chill :)
You are mistaking the length of my responses for anger. Rather, I want people who, unfortunately, still look at this article in 2018+ to know its problems. They can also better decide if this is a trustworthy source for the future, or if they just need to be more critical of their editorial bias.

If it is so old as to be irrelevant, as you and others allege, then no should see this article or our discussion. So no need to worry there either :)

Posted on 2018-06-21 06:27:28