Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1535
Article Thumbnail

Premiere Pro CPU Roundup: AMD Ryzen 3rd Gen, AMD Threadripper 2, Intel 9th Gen, Intel X-series

Written on July 22, 2019 by Matt Bach
Share:

Introduction

In our Photoshop and After Effects testing, we have already seen that AMD has almost entirely closed the gap with Intel when it comes to lightly threaded applications. Unlike After Effects or Photoshop where per-core performance tends to be king, however, Premiere Pro can see at least some benefit from having a higher number of CPU cores. Because of this, we have really been looking forward to testing the new Ryzen 3rd generation CPUs in Premiere Pro as it would make sense for AMD to firmly take the performance lead over Intel.

AMD Ryzen 3rd Gen Premiere Pro Performance

In this article, we will be looking at exactly how well the new Ryzen 3600, 3700X, 3800X, and 3900X perform in Premiere Pro. Since we expect these CPUs to shake up the market quite a bit, we also took this opportunity to do a full CPU roundup. Not only will we include results for a few of the previous generation Ryzen CPUs, but also the latest AMD Threadripper, Intel 9th Gen, and Intel X-series CPUs. And for good measure, we will throw in a 14-core iMac Pro and a current (for the moment) 2013 Mac Pro 12-core as well.

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

Looking for a Premiere Pro Workstation?

Puget Systems offers a range of workstations that are tailor-made for your unique workflow. Our goal is to provide the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Test Setup & Methodology

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

Shared PC Hardware/Software
Video Card NVIDIA GeForce RTX 2080 Ti 11GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit (version 1903)
Premiere Pro CC 2019 (Ver 13.1.2)
Puget Systems Pr Benchmark V0.2 BETA
Mac Test Platforms
iMac Pro 14-core Intel Xeon W
64GB 2666MHz DDR4 ECC
Radeon Pro Vega 64 16GB
1TB SSD
Mac Pro (2013) 12-core, 2.7GHz
64GB 1866MHz DDR3 ECC
Dual AMD FirePro D700 6GB
1TB PCIe-based SSD

*All the latest drivers, OS updates, BIOS, and firmware applied as of July 2nd, 2019

Note that while most of our PC test platforms are using DDR4-2666 memory, we did switch up to DDR4-3000 for the AMD Ryzen platform. AMD CPUs can be more sensitive to RAM speed than Intel CPUs, although in our Does RAM speed affect video editing performance? testing, we found that the new Ryzen CPUs only saw modest performance gains in Creative Cloud applications when going from DDR4-2666 to even DDR4-3600 RAM.

For each platform, we used the maximum amount of RAM that is both officially supported and actually available at the frequency we tested. This does mean that the Ryzen platform ended up with only 64GB of RAM while the other platforms had 128GB, but since our Premiere Pro benchmark doesn't need more than 32GB of RAM to run, this does not actually affect performance at all. We have recently re-confirmed this in our RAM speed article linked above.

However, keep in mind that this is technically overclocking since the AMD Ryzen 3rd Gen CPUs support different RAM speeds depending on how many sticks you use and whether they are single or dual rank:

Ryzen 3rd Gen supported RAM:

  • 2x DIMM: DDR4-3200
  • 4x single rank DIMM: DDR4-2933
  • 4x dual rank DIMM: DDR4-2667

Since we are using four sticks of dual rank RAM (almost every 16GB module available will be dual rank), we technically should limit our RAM speed to DDR4-2666 if we wanted to stay fully in spec. However, since many end users may end up using a RAM configuration that supports higher speeds, we decided to do our testing with DDR4-3000, which right in the middle of what AMD supports.

The benchmarks we will be using are the latest release of our public Premiere Pro benchmark as well as NeatBench 5. We are beginning to include more plug-ins in our testing, and Neat Video (with its terrific stand-alone benchmark utility) is an easy place to start. Full details on the benchmarks and a link to download and run it yourself are available at:

Benchmark Results

While our benchmark presents various scores based on the performance of each test, we also wanted to provide the individual results. If there is a specific task that is a hindrance to your workflow, or a specific codec you use, examining the raw results is going to be much more applicable than our overall scores. Feel free to skip to the next section for our analysis of these results if you rather get a wider view of how each CPU performs in Premiere Pro.

Premiere Pro Benchmark Analysis

Looking at the overall performance in our Premiere Pro benchmark (which combines playback and export tests with the various codecs), the results are very interesting and not quite what we expected.

At the top of the charts, the Intel X-series continues to hold four of the top five spots. The AMD Threadripper 2950X takes the fourth-place spot which is impressive as it is both cheaper and faster than the Intel Core i9 9920X 12 core. However, note that the higher-end Threadripper CPUs do relatively poorly in Premiere Pro, likely due to the fact that they are actually two CPUs merged into one and have some ongoing issues with core prioritization and memory scheduling.

The comparison most readers are likely interested in is the AMD Ryzen 3rd generation versus Intel 9th Gen processors. Looking at just the overall score, AMD has a pretty decent performance lead at the bottom-end, but by the time you get up to the AMD Ryzen 9 3900X vs the Intel Core i9 9900K, the scores are within a percent or two of each other. However, in this case the overall score is actually masking the real story, so we need to dig into the individual results.

We know that our raw benchmark results tables are a bit of a mess, but if you take the time to really dig into them, there are two very interesting conclusions to draw. Simply put, the main difference between the AMD Ryzen and Intel 9th Gen processors is that Intel is better at processing H.264/H.265 footage, while AMD Ryzen processors are better at processing RED footage.

AMD Ryzen 3rd generation vs Intel 9th Gen for playback of H.264 and H.265 media

The Intel advantage for H.264/265 media is pretty easily explained by the fact that Premiere Pro supports hardware accelerated encoding/decoding of H.264/265 media via Intel Quick Sync. AMD does not have this feature (nor does the Intel X-series for that matter), which explains why the Intel 9th Gen CPUs are simply going to be better at processing H.264/265 footage.

In the chart above, we did pull the 8K H.265 59.94FPS results simply because that test in particular is really inconsistent, but even just looking at the 4K H.264 10-bit (27.97 and 59.94FPS) and 8K H.265 (29.97FPS) results, you can see just how much better performance can be with the Intel 9th Gen CPUs. The Core i5 9600K doesn't have a big lead over AMD (and is in fact a hair slower with 59.94FPS media), but the Core i7 9700K is on average 15% faster than the AMD Ryzen 7 3800X and the Core i9 9900K is on average 30% faster than the AMD Ryzen 9 3900X.

What this means is that if you are looking to get the best possible performance from H.264/H.265 media, the Intel 9th Gen CPUs are definitely the best route to take.

AMD Ryzen 3rd generation vs Intel 9th Gen for playback of RED media

Unlike h.264, RED footage, doesn't have any hardware acceleration that is unique to Intel. It is also in general pretty tough to process, which is partly why the raw power and higher core count of the Ryzen CPUs allows it to take a clear lead over the Intel 9th Gen CPUs.

Starting at the bottom end, the AMD Ryzen 7 3600 is about 60% faster than the Core i5 9600K when playing RED media, not to mention a bit cheaper to boot. Moving up the stack, the Ryzen 7 3800X is about 35% faster on average with RED footage compared to the Core i7 9700K, while the AMD Ryzen 9 3900X is 20% faster on average than the i9 9900K.

One thing we do want to point out, however, is that RED has been working on an update to their SDK that will move more of the processing of RED footage from the CPU to the GPU. Rumors were that this was supposed to be included in the December update for Premiere Pro, but was pushed back for unknown reasons. When this update eventually goes live, we may find that the CPU is no longer a major factor for those that work with RED footage. If this is indeed the case (only real benchmarks will tell), then that may remove the performance benefit with RED footage that we currently see with the AMD Ryzen CPUs.

On the other hand, the new Ryzen CPUs are very new, so it is equally possible that there will be performance optimizations that will make Ryzen more attractive in other areas. There is no way to know the future, so we will simply have to wait and find out.

Neat Video Benchmark Analysis

Neat Video is a noise reduction plugin that is currently one of the more popular options available. Plug-ins in general are something we plan on taking a look at more in the future, and Neat Video is a great place to start since they have an excellent stand-alone benchmark called NeatBench. Neat Video can use either the CPU, the GPU, or combine both together, but since using the GPU often overshadows the CPU, we decided to primarily look at the "Max CPU Only" result, although we will include the CPU+GPU result for those that are interested.

If there is any test to get you excited about the new Ryzen CPUs, this is most certainly going to be it. Not only do the Ryzen CPUs out-perform the comparably priced Intel 9th Gen CPUs, they do so by around 60%! They can't quite keep up with the Intel Core i9 9960X and Core i9 9980XE, but considering the fact that those CPUs are 3-4x the cost, even the fact that this is a comparison we are making is extremely impressive.

Why the AMD Ryzen 3rd generations CPUs are so good for Neat Video, we really have no idea. With both the AMD Ryzen and Intel 9th Gen processors, we hit peak performance at around 4-6 cores, so it isn't even the higher core count that is making a difference here. But whatever the reason behind it, it is clear that if you use the Neat Video plug-in, the AMD Ryzen processors should give you a very substantial boost to performance. Even if you use H.264 media (where Intel holds a commanding performance lead), you may end up wanting to use Ryzen simply because you can use proxies for editing, but noise reduction is always going to take forever.

Are the Ryzen 3rd generation CPUs good for Premiere Pro?

Depending on how you look at it, choosing between the new AMD Ryzen CPUs and the Intel 9th Gen CPUs is either a very easy, or a very complicated, question to answer. It is easy because there is a clear winner in specific situations, but complicated because there is no single overall best choice.

From the testing we have done, it largely comes down to whether you meet one of the three following criteria:

Do you work with H.264 or H.265 media?

In this case, the Intel 9th Gen CPUs have a pretty commanding lead (especially with 10-bit footage) due to the fact that Premiere Pro supports hardware acceleration of H.264 and H.265 (HEVC) media via Intel Quick Sync. This feature isn't available on AMD processors (or on the Intel X-series), which makes CPUs like the Intel Core i9 9900K simply the best option for this type of media.

Do you work with RED footage?

Here, the higher raw performance of the new AMD Ryzen CPUs allows it to take a significant lead over the Intel 9th gen CPUs - although they can't quite catch up with the higher-end (and much more expensive) AMD Threadripper or Intel X-series CPUs. However, this is one area that is highly subject to change since RED is working on moving more of the processing of RED media from the CPU to the GPU. This is expected to be bundled into one of the next few Premiere Pro releases, at which point we may find that the CPU no longer makes more than a minor impact on performance when working with RED footage.

Do you use Neat Video for noise reduction?

If you use Neat Video, there is simply no contest: use an AMD Ryzen 3rd generation processor. Performance is up to 60% faster when using only the CPU, but even with Neat Video set to use both the CPU and GPU (with a NVIDIA GeForce RTX 2080 Ti) we still saw a 20-30% performance gain with Ryzen over a comparably priced Intel CPU.

Outside of these three situations, performance is going to be pretty close between the AMD Ryzen 3rd generation and the Intel 9th Gen processors. Something else to keep in mind, however, is that performance is only part of the question - there are a number of other considerations that you may want to keep in mind:

On the Intel side, the Z390 platform has been available for quite some time which means that most of the bugs and issues have been worked out. In our experience over recent years, Intel also simply tends to be more stable overall than AMD and is the only way to get Thunderbolt support that actually works. Thunderbolt can be a bad time on PC, and there are only a few motherboard brands (like Gigabyte) where we have had it actually work properly.

For AMD, the X570 platform is very new and there will be a period of time where bugs will need to be ironed out. However, AMD is much better about allowing you to use newer CPUs in older motherboards, so if upgrading your CPU is something you will likely do in the next few years, AMD is the stronger choice. In addition, X570 is currently the only platform with support for PCI-E 4.0. This won't directly affect performance in most cases, but it will open up the option to use insanely fast storage drives as they become available.

Keep in mind that the benchmark results in this article are strictly for Premiere Pro. If your workflow includes other software packages (We have articles for Photoshop, After Effects, DaVinci Resolve, etc.), you need to consider how the processor will perform in all those applications. Be sure to check our list of Hardware Articles for the latest information on how these CPUs perform with a variety of software packages.

Looking for a Premiere Pro Workstation?

Puget Systems offers a range of workstations that are tailor-made for your unique workflow. Our goal is to provide the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Tags: Premiere Pro, AMD Ryzen 3rd Gen, AMD Threadripper 2nd Gen, Intel 9th Gen, Intel X-series, Intel vs AMD
Nick Verzilli

Been waiting for this for a while! Awesome write-up, thanks!

Posted on 2019-07-23 00:13:53
SMD79

REALLY appreciate this. We're purchasing a new system this weekend for my assistant editor and we couldn't decide on which CPU was the better fit. We cut mostly Sony 4k footage so looks like 9900k is the clear winner. Question...does Quick Sync just work or do you need to some how enable it either in the BIOS or in Premiere or anything like that? How do you turn it on/off?

Posted on 2019-07-23 00:38:01

Best place to start is to check if you have the option to use "Hardware Acceleration" when exporting to H.264 (under Encoding Settings -> Performance). If you can, then all the hardware is good to go. Only thing you might want to check is that under Preferences -> Media that "Enable hardware accelerated decoding" is checked. Not that this box can be checked even if your system doesn't support it, which is why checking the export options is the better way to know if you have things set right.

If you don't have to option to use hardware accelerated exporting, you probably need to go into the BIOS and make sure integrated graphics is enabled (not on Auto, but actually enabled) so that it shows up as a video device in Device Manager. After that, you need to install the Intel Graphics driver, or let Windows update finds it for you. Then it should work.

Posted on 2019-07-23 17:26:51
Rodri

Is the igpu still worth even if we do not export using "hardware acceleration" as I believe it yields worst results in quality rather than using software?

Posted on 2019-08-11 20:28:22

It is also used for decoding, so it improves playback performance when editing H.264 footage. I'm not sure if there is a quality loss for that like their is for exporting, but I've never heard anyone mention it, so there must be no (or art least minimal) quality difference for decoding.

Posted on 2019-08-12 17:47:11
leonard

The overall score is " masking " Adobe's random performance, for example, in your individual charts in 4k h264 10 bit 59.94 fps multicam live playback half, the ryzen 3600 scores 21.06 frames and 3900x scores 13.85 frames, so ryzen 3600 is faster, strange.
Then, 8k h.265 59.94 fps Lumetri color Live playback Half, ryzen 3600 10.65 frames, 3800x 57.25 frames, 3900x 24.42 frames :) is this a joke , and 9900k 6.06 frames.
I'm betting you won't see this kind of variations in Davinci resolve.
In my opinion Premiere pro can't be used as a benchmark anymore, when other NLE's stabilizes or tracks objects/faces in seconds and Premiere pro does it in MINUTES, then the weakest link is not the hardware, it's the software.

Posted on 2019-07-23 07:11:36

Premiere Pro is definitely really hard to benchmark, for exactly the reasons you listed. Variation between runs can be huge - especially with H.264 media. We're still working on trying to get it more consistent, but it isn't easy at all.

One thing we may end up having to do is to run the benchmark multiple times and take the "best" result. That still doesn't completely solve the issue since we may get the "best" result on the first run, or it might take 10 runs. If we did that, we would probably also have to drop down the number of codecs we test so that it doesn't take a stupid long time to do a single benchmark run.

At least for right now, we are doing enough tests that things seem to be averaging out pretty well. There are a few that don't at all (8K H.265 59.94FPS comes to mind), however, which is why we left that one out of our analysis.

If there was another Premiere Pro benchmark out there that successfully tested live playback, we would definitely use that. But since there isn't, all we can do is to do our best (any info is better than no info), and continue to improve the benchmark as we go.

Posted on 2019-07-23 17:31:51
leonard

Exactly, very hard to benchmark.
I edit all my footage with Premiere pro 12.1.2 ( 2018 ) and was curios if the new version (13.1.3 ) runs better ( DJI drone files and some heavy lumetri ),
my god, it dropped frames even playing back a simple 4k DJI drone file with NOTHING applied, how can newer versions of Premiere pro run worse and worse, i really need to move to another NLE ASAP.

Posted on 2019-07-24 11:08:57
Jozsef Weigert

I cannot see the detailed benchmark results as the resolution of the image is too low here and at the After Effects review, too. (The Photoshop benchmark is fine).

Posted on 2019-07-23 07:32:55
Hwgeek

open the image in new window and remove the "&width=1200&height=800" limit.
for example:
https://www.pugetsystems.co...

Posted on 2019-07-23 07:35:11
Jozsef Weigert

Thanks, it worked!

Posted on 2019-07-23 07:36:50

Sorry about that! We recently moved to a new gallery view for our charts, but clicking on it opens up the old gallery rather than the full sized image. This is getting fixed in our current website programming "sprint", so it should be fixed in the next week or so.

Posted on 2019-07-23 17:32:49
Jozsef Weigert

Great, thanks for the reply!

Posted on 2019-07-24 07:21:19
Hwgeek

WOW, looks like the 3950X will be a monster!
Also pls test the Ryzen 3000 with Noctua cooler in future reviews, Ryzen 3000 Boosts like GPU depending on temp {Best under 60c} so will get little better results, Pro users will not keep that cooler with 3900X, even on my 2700X it was noisy when spin-up to 3000+RPM.
Great article like always.
P.S: if we can judge by 3900X and see that dual chiplet design with I/O chiplet doesn't hurt performance vs single chiplet, then you can imagine how good the new ThreadRippers with 128MB/256MB cache with 16~64 cores will perform.

Posted on 2019-07-23 07:33:06
leonard

Unless we are talking about some plugin that has nothing to do with Adobe like neat video, or some other NLE like Davinci Resolve, the new threadrippers will probably be at the bottom of the charts, underutilized, full of unexplained bugs that make threadripper lag behind consumer Intel chips, reviewers will say the usual " Adobe likes single core performance" so 64 cores and you will track objects with 1-2 cores in minutes while others use everything you've got and does it in seconds.

Posted on 2019-07-23 07:57:28
Mefist

I work with BRAW from Blackmagic. Does he look like RED? Who will be faster? 9900K or 3900X?

Posted on 2019-07-23 09:20:35

BRAW isn't natively supported by Premiere Pro yet, so we really have no idea. I know there is a plugin to enable it, but that probably isn't indicative of the performance once it is officially supported.

That said, I don't believe BRAW uses the GPU for debayering/decoding, so the performance won't be like RED. If I had to guess, it is probably more like CinemaDNG than anything, but RAW codecs especially can perform very differently from each other since it depends heavily on how the debayering is processed (single threaded, multi threaded, GPU accelerated, etc.)

Posted on 2019-07-23 17:36:56
Mefist

Thank, who is faster in working with proress?

Posted on 2019-07-23 17:41:27

Intel and AMD are pretty close at similar pricepoints. AMD Ryzen has a slight lead over the Intel 9th Gen, but it isn't be a whole lot and won't be very noticeable unless you work with 59.94FPS media.

Posted on 2019-07-23 17:52:53
Mefist

thanks, I work 50fps

Posted on 2019-07-23 20:51:38
Mart

What about "Ryzen 3950x" ?

Posted on 2019-07-23 14:13:46

That CPU isn't out yet. When it does release, we'll probably do a smaller article with just 4-5 CPUs to see how it performs.

Posted on 2019-07-23 17:37:28
Mart

Okay, thanks

Posted on 2019-07-24 08:02:07
Joseph Croasdaile

Thanks for this test, but man do I still have a ton of questions. Also I should state I still edit in 1080p on a 5 year old Intel cpu... not doing anything in 4k.

3900x and more threads
I was so excited for the 3900x, but the tests so far are mixed to put it lightly. Some things I was excited to see like the handbrake encoding times I would love to see a benchmarks in regards to Adobe Media Encoder / proxy rendering times in premiere as that impacts my workflow. The plugin test also makes me wonder the impact if I am using a lot of Red plugins / filters. I did also see the AE benchmarks which again were mixed... but I need to read up more on them as I need to see if 3900x is better for doing 3D / plugins within After Effects

9900k and the benefit of quicksync
I never really considered editing strictly in h.264/h.265 as my current setup can't handle it. So the 9900k quicksync decoding sounds interesting as it would speed up the timeline and possibly remove the need for proxy clips. THIS WOULD BE MASSIVE!!! The big question I have is I have seen countless people trash quicksync in regards to encoding quality. I assume you could get the decoding benefits while working the premiere timeline and disable quicksync on export. Now if you tell me that the 9900k can tear thru 1080p h264 / h265 video like butter in Premiere's timeline without dropping any frames (even when highspeed scrubbing) then I know which CPU I am getting...

Posted on 2019-07-23 18:10:31

Performance in AME should be pretty much the same as our export tests. We don't test straight transcoding, but the test sequence with just Lumetri Color will be really close to the same as a straight transcode. And I definitely agree with your thoughts on plug-ins, we really have no idea how something like RED Giant will perform - that is one of the reasons that adding separate tests for popular plug-ins is something we really want to add.

As for quality with Quicksync, there is a TON of bad information out there. For decoding (playback), I don't think quality is a concern. Exporting is where it is, but you can switch between Hardware Accelerated (Quick Sync) and Software Only in the export settings. The exact quality difference is something we really want to do an in-depth analysis of, but in general it should be fine if there isn't a ton of motion or patterns and you are uploading to Youtube. You may want to use a slightly higher bitrate to compensate for the quality difference, but it is probably fine. This is really something that everyone is going to have a different opinion of, however, so unfortunately it is kind of a "try it and see" situation.

If you are delivering to a client outside of streaming services, you probably want to stick with Software Only just to get the best quality at whatever bitrate you use. Something else to be aware of is that Hardware Accelerated encoding just straight breaks in Adobe apps if you set the bitrate to higher than ~60mbps. No errors report, but it fails over to use a really low bitrate instead of the target you set. I think that is part of the "Quick Sync is terrible" talk - they didn't check to make sure the file size at the end was right and are comparing a 40mbps video to a 5mbps one. This is something we've even been guilty of ourselves.

Posted on 2019-07-23 18:43:50
Mark Harris

Interesting comment about the 60mbps. I need to test that now. Thanks!

Posted on 2019-08-01 18:03:51
Mark Harris

There people trashing QuickSync are basically talking nonsense. You have to do like 300% zoom to see artifacts vs CPU and mostly in some patters or specific areas so its basically a moot point for anyone that has actually compared both results.

Posted on 2019-08-01 18:02:03
Behrouz Sedigh

For Neat Video Benchmark , I think L3 cache plays important role.3900x does have 64mb L3 While TR 2920x = 32mb , even Ryzen 3600 which is equal to Tr2920 , does have 32mb (2x16mb) cache.this only applies to AMD Ryzen.

Posted on 2019-07-23 18:37:11
MacherTV_Peter

Great article. Thank you a lot!

Posted on 2019-07-23 19:07:15
Tom

*All the latest drivers, OS updates, BIOS, and firmware applied as of July 2nd, 2019

Does that included all security patches for Intel CPUs?

Posted on 2019-07-23 19:41:01

If you mean the meltdown/specter stuff, yes, all those patches are applied.

Posted on 2019-07-23 19:54:50
Tom

Cool thanks

Posted on 2019-07-23 20:03:07
Richard Staller

Have you seen any impact in performance from applying those patches?

Thanks in advance and I love these articles.

Posted on 2019-07-27 20:24:31

Those patches have been live for so long that I can't really recall exactly to be honest. I remember doing a bit of testing when they first came out and not really noticing much of a difference, however, which is why we didn't really worry about it very much.

Posted on 2019-07-29 02:40:17
Richard Staller

Okay, thanks!

Posted on 2019-07-30 22:09:37
dfsaqwe

first time reading this site's reviews and i have a lot of questions.

your benchmark includes "gpu" tests but which gpu are you using? the test bench says it has GTX 2080s but your review makes mention of intel quick sync - so which encoder are you using? intel or nvidia?

and then further to that point why would you include gpu test when its well known these encoders cannot achieve the quality of software encoding?

and even further to that point, would it not make more sense to do software only test in order to compare these cpus against each other at highest quality settings?

Posted on 2019-07-23 23:30:10

The discrete GPU used is the NVIDIA RTX 2080 Ti, but on the CPUs that support it (the Intel 9th Gen), we left onboard graphics enabled in order to allow Premiere Pro to utilize Quick Sync for hardware accelerated encoding/decoding of H.264/H.265 media. Premiere Pro doesn't use the GPU for encoding unless you use a third party plugin, it is just used for any effects that are GPU-accelerated.

Hardware accelerated decoding really doesn't result in any difference in quality that we've seen, but you are definitely right that there is a difference with encoding. However, I think there are a couple things that people don't realize about it. First, I suspect that many of the claims about quality are due to the fact that if you set the bitrate too high (above ~60mbps), Premiere Pro and AME simply break and flip over to using a super low bitrate instead. So I think some people end up comparing a 80mbps video to a ~5mbps (or whatever) video without being aware of it. We've been guilty of this ourselves since the "High Bitrate" preset it above this bitrate limit. Also, the quality difference (and performance gains for that matter) are highly dependent on how much motion there is in the original clip as well as any complex patters.

I did some quick testing with our sample clips and really didn't notice a major quality difference at 40mbps, which is why we went ahead and used hardware acceleration when supported. We actually used to go the other route and only did software only, but got a ton of feedback that we should do our testing with hardware acceleration. We could do both, but that would honestly be very confusing for most of our readers, so I would like to avoid that if we can.

This is really one of those situations where there isn't a perfect answer, so we are (currently) going with what tends to be the most requested by both our readers and customers. One thing I do plan on doing is a more in-depth analysis of when hardware acceleration affects quality, and what the performance gains could be - but I think that is going to end up being a major project since the source clip seems to make such a big impact on both of those.

Posted on 2019-07-23 23:43:50
Rasmus Glud
Premiere Pro doesn't use the GPU for encoding unless you use a third party plugin, it is just used for any effects that are GPU-accelerated.


How about AME? Aren't the CUDA and OpenGL options in AME considered GPU-accelerated encoding? I stricly use AME when outputting since I'm able to use CUDA with my 1080 Ti for faster encoding. I guess you are bencmarking the export option within Premiere Pro, right? In that case it would be really interesting to see how 3rd gen Ryzen performes along CUDA and OpenGL in AME. Isn't that how people export their PP-projects these days? Or have I completely misunderstood the advantage of using AME (CUDA / OpenGL) as the "export engine"? Man, if I've been doing something wrong with my exports for half a decade... damn.

Posted on 2019-07-28 12:14:17

Premiere Pro and AME are pretty much identical in terms of how they work - I'm pretty sure they share a lot of the same code base for exporting. We actually used to do our Premiere Pro export testing through AME before we decided to go ahead and make our benchmark into a full-fledged plug-in and could easily export through Premiere Pro directly.

The main thing exporting though AME gets you is the ability to export while still being able to do things in Premiere Pro since exporting straight from Premiere Pro means you can't do anything until the export finishes. Performance, quality, and everything else like that is identical between the two.

Posted on 2019-07-29 02:39:28
Streetguru

Should make an interactive graph where people can just select between 2 CPUs to compare.

Still gotta get into testing Sony Vegas sometime as well given how much cheaper it is than premiere when it pops up on the humble bundle store.

Hopefully AMD bothers spending some dev money sometime at adobe to make better use of polaris/vega/navi GPUs to make up for the lack of quick sync.

Posted on 2019-07-26 03:00:00

Interactive graphs is actually something we have on our list of programming projects. It is part of a much larger (and cooler) project, however, so it may be a little bit.

As far as Sony Vegas, we simply don't see much of a demand for high-end workstations for Vegas. Every once in a while we get a customer looking for a system, but it is rare. That isn't to say there aren't a lot of Vegas users out there, but we have to tailor our testing towards our customers first and foremost since that is primarily who our testing is intended to benefit.

Vegas is on the list of potential software to target at some point, but it is behind a few other that we will likely take on first like Avid.

Posted on 2019-07-29 02:30:11
leonard

Downloaded media files used in benchmark, i'm sorry to say Matt Bach but they have nothing to do with h.264 and h.265 from Panasonic GH5 and Sony A7s, A7III and the usual most used cameras for videography.
I've put the H264_100Mbps_10bit_4K_2997.mp4 on the timeline and a complete slideshow, it just didn't play, then i put GH5 h.264 10 bit file, played perfectly fine on a ryzen 1800x.
Your h.264 and h.265 files are completely wrong, nobody works with those kind of files.
If you don't have a GH5 or Sony a7 III, get some files from this guy - https://go.whoismatt.com/gh5/ , they are 10 bit 24 frames 4k files, play perfectly smooth.
For me your test is completely invalid, doesn't reflect reality at all, you used broken H.264 and h.265 files that Premiere pro is not optimized for.
If someone doesn't believe me, just download Premiere Pro Benchmark Files from Puget and just put let's say H264_100Mbps_10bit_4K_2997 or H264_100Mbps_10bit_4K_5994 on premiere pro timeline with nothing applied and hit play, it will get stuck in seconds, then use 4k files from your camera like GH5, Sony A7 III, Fuji X-t3, Nikon z6.......etc. it will play perfectly fine.

Posted on 2019-07-26 07:56:03

H.264/265 is really funky since how often you have I-frames is completely hidden by camera manufacturers. For ease of testing, we created the H.264/265 and ProRes media straight our RED footage in Premiere Pro. That said, I did do some 1:1 comparisons to footage we have on file from a GH5, and the performance was roughly the same. Are you sure the footage from your camera was 100mbps and that it was recorded in LongGOP (not All-Intra)?

Posted on 2019-07-29 02:34:46
leonard

On my ryzen 1800x i can put any combination of MOV or MP4 from GH5 with any bitrate and resolution and it won't lag at all in Premiere Pro unless i do some heavy color grading with lumetri or apply some denoiser.
If we were to take your benchmark to be genuine it would mean there is no computer in the world that can handle 4k 10 bit h.264 in multicam or with some lumetri applied.
Well, that is not true at all, plenty of people cutting GH5 10 bit files with much slower computers.
Even LongGOP files from DJI drones which are the worst files to work in any NLE and they don't drop frames, they stutter sometimes because Premiere process in advance but nowhere near what your files do.
I'm sorry to say this but i will never trust internet reviews again, at least now i know how you got those "Intel favored results"
Again, if anyone is curios just download those files ( if puget don't take them down ) and just put them on timeline and hit play.

Posted on 2019-07-29 12:48:35
Jim Leonard

While this "leonard" person could do a better job of communicating, he's right: I did the same test, and found that the files provided by Puget in the benchmark have difficulty playing. I have an i7-8700k with 64G RAM and two 1080tis, and I found that footage from my own camera was decoded via the Intel quicksync extensions and played fine. But when I tried the h264_100Mbps files, even the 29.97 file didn't play smoothly because something about it wasn't getting hardware decoded, it was all in software. As a test, I re-encoded it with identical parameters using Adobe Media Encoder, and THAT file played fine and was hardware-decoded.

I actually think it's the Long GOP settings you used, since when it is trying to play via software, I only see 10% CPU usage. Premiere is having a great deal of trouble trying to interpret those files. So I don't think they're representative of actual files found in the wild. What encoding settings did you use?

As a further note, slowly horizontal panning of a still scene isn't the best test of an h.264/h.265 decoder. Most of the work is done in such a scene is done with the motion vectors (and in h.265's case, global motion compensation). Something should be moving in the scene to be a better test.

Posted on 2019-08-14 02:04:10
Stephen Jones

Isn’t quicksync just GPU acceleration? Wouldn’t that advantage be negated by using the dgpu for acceleration?

Posted on 2019-07-30 13:32:13
Flanders

I don't quite catch that either. Is QuickSync exclusively supported?

Posted on 2019-07-31 14:30:51
Mark Harris

Not exactly. Quicksync will give you much better performance than using CUDA or similar for Premiere. That is why the 9900k can outperform the 3900x when using the iGPU which is a great feature and makes the 9900k for someone like me that does, Photoshop, Premiere and gaming, the best CPU for my needs.

Posted on 2019-08-01 17:59:33
Stephen Jones

That’s the great thing about competition, more people’s needs get met. From what I’ve seen, the the timeline performance on 3900x is superior though so I guess it would boil down to weather timeline performance or exporting speed is more important to someone

Posted on 2019-08-02 00:11:32
Mark Harris

Really? I have seen the opposite so where did you see that?

Posted on 2019-08-03 21:25:15
Stephen Jones

Quicksync only helps with rendering. The higher computing power of the 3900x comes into play in actual editing workflow. the touch on it on this one and other reviews as well.
https://www.youtube.com/wat...

Posted on 2019-08-06 15:39:05

I have never heard of Coreteks, but there are some things in that review that threw me off a bit. The biggest thing I noticed is that they did an export test using H.264 media (where the Intel 9900K with Quicksync was easily best), but when they talked about scrubbing the timeline he mentioned that they are using "uncompressed 4K footage". So they switched the media used between the two tests from one that the 9900K is better at (H.264) to one that Ryzen is better at (uncompressed in some format). I'm all for testing different codecs (that should be obvious from our tests), but doing one test with one codec, then switching to a completely different codec for a second test seems misleading to me.

They also only showed the scrubbing with the 3900X - there is no video showing how good or bad the 9900K and other CPUs did. In fact, while he talked about how good the 3900X is, he never once mentioned how any of the other CPUs performed for that.

It is a bit odd that they didn't actually do a comparison between the different CPUs for scrubbing. Also odd that they switched to uncompressed 4K for the scrubbing "test". I honestly don't doubt that Ryzen 3900X is better for that with uncompressed footage (our own testing indicates that it should be), but if they did the same timeline test with something like 4K H.264 10-bit, I guarantee the results would be in favor of the i9 9900K with Quick Sync.

Posted on 2019-08-06 16:48:43
Stephen Jones

I never denied that quicksync acceleration is faster for rendering than pure cpu. Either way I’m just glad to see real competition. It was so boring when intel was definitively better at everything. Can’t wait to see what oems do with these new chips.
I don’t think it was misleading at all, they went from testing export to testing timeline performance and were very clear about using uncompressed footage.

Posted on 2019-08-06 18:12:48
yezhacker

This is the benchmark I was expecting most, but didn't want to comment until the issue with the tables were fixed. It's really odd how h.264 100mbps behaves. As someone commented before, anyone could tell that no consumer PC is able to manage such codec. Apparently, even 8K RED RAW is easier to work with than h.264 which is crazy. The higher framerates you worked with this time also makes the performance looks even worse, at least for the 9900k, although I understand you are getting every processor to its limits. By the way, the 3900x is really good.

Posted on 2019-08-02 01:20:12

Keep in mind that the 100mbps H.264 media is also 10-bit, not 8-bit like the 50mbps media. We did a bunch of testing to figure out exactly what bitrate we wanted to include in our benchmark, and when we tested 100mbps 8-bit H.264, performance was just fine so it isn't the bitrate that is the problem, it is the color depth. Not everyone uses 10-bit, but we decided to go with it so we would have an "easy" H.264 test (50mbps 8-bit) and a "hard" H.264 test (100mbps 10-bit).

I'm not sure Premiere Pro in particular seems to absolutely hate 10-bit H.264/H.265 media. It doesn't seem to be much of a problem in other apps like DaVinci Resolve, but Premiere Pro simply struggles with it. If I had to guess, it is more of a bug than an intrinsic issue, however, so hopefully it gets fixed in a future update.

Posted on 2019-08-05 20:43:49
yezhacker

You used to list all the codecs and bitrates you work with, before. It would be good to do it again on your next benchmarks, for clarity. Not only hardware specs are important.

Take this as constructive criticism, keep doing your great work, I learned a lot since I discovered your website. I'll be expecting your next benchmarks. Thanks =)

Posted on 2019-08-06 19:02:12

We list them on the benchmark page www.pugetsystems.com/go/PRb... and in the raw benchmark results, but yea, we did remove them from the test setup section. Mostly because most readers completely skip over that section, so to keep things a bit more trim we are starting to move all of the information about the benchmark itself to individual benchmark pages. Also, now that we have many of our benchmarks available for people to download and use themselves, it just makes sense to go into the fine details like that in one central location.

Posted on 2019-08-06 19:05:48
yezhacker

That makes sense

Posted on 2019-08-06 19:09:29
Ian Firmani

Please, explain why I've better results using the benchmark with my Ryzen 5 3600 than yours, even better than the 2700X.
https://uploads.disquscdn.c...

Posted on 2019-08-20 05:51:42

There are a ton of things that could explain the ~10% difference performance. First, you are using a different GPU, but it could also be the speed of your RAM. Or, it could be your motherboard. Some boards have hidden auto-overclocking enabled by default that you need to disable to get true stock speeds. We've also seen plenty of boards that auto-overclock without any way to disable it (often by turboing all cores to the max turbo instead of scaling the frequency based ont he number of cores used).

Did you actually do the performance testing with those Intel CPUs or the iMac Pro? They look to be identical to our scores, which is fine to use (although you should really attribute the source for any testing you didn't personally do), but that throws in a huge number of variables. Not only the could hardware ones I mentioned, but also Windows version, Premiere Pro version, drivers, BIOS, etc.

Posted on 2019-08-20 16:48:01
Ian Firmani

Yes, actually i'm using diffetent hardware, but i'm using a "low-end" hardware (talking on light workstation uses). My Mobo is a MPG X570 gaming plus, that is the msi entry-level x570. My ram is a 2x HyperX fury 16GB 3200mhz, I know that ryzen is very sensitive to memory frequency, but my timings are 18-21-21 and yours are 15-16-16 and it's even a bit more important than the frequency itself. The RX 5700 is equivalent to a RTX 2060 super, not a 2080, so there's not performance boost because the GPU. For software i'm using the lastest updates on windows 10 pro, bios, chipset and gpu drivers. With premiere i'm using the same version as you.

On config of the bios, I disable all auto oc stuff because I'm using ryzen master for testing the different configuratios: stock, precision boost overdrive and oc @4.2Ghz on all cores (and ryzen have a minimum oc headroom). I'm searching for someone in chile with the exact build that you're using to compare and validate my benchmarks.

On the other hand, I notice that you make these benchmarks on July 2. So I think that you used the first press release bios and drivers, and everyone knows about prerelease perfomance troubles. If this is the case I think that a 10% difference is not a small difference.

I'm doing these graph's because I run the benchmark and found this numbers, so I did it again using the others profiles and graph it. And I want to bring information for the people of my country in our language (is a mess that not much people speak english here) so they don't search for your reviews or others english speakers websites. Just spanish.

I'll attribute all the information that is not created by me, so don't worry about that. Thank for the fast response and have a good night! (I know that in some parts I use engrish instead of english, sorry for that.)

Posted on 2019-08-21 01:43:01
Samrat kundu

Which one would be good for Adobe premiere Pro, light room and durable / longer run..
i7 8700 non k Or Ryzen 5 3600 (I won't OC) ?
With 4x8GB 3000mhz c16 Ram..

Posted on 2019-08-20 16:08:53

Will be any Lightroom benchmark with the new Ryzen 3000 CPUs?

Posted on 2019-08-30 09:59:24
Dash

AMD Ryzen 9 with PBO enabled cooled by custom loop, on Asus WS X570-ACE motherboard, 64gb ram.

I did read this review and Puget nearly convinced me to not go to AMD due to the lack of Quicksync potentially slowing down H264 footage on timeline.

PLEASE let me assure everyone, this was the best upgrade EVER! Puget I'm afraid your review is flawed based on your custom compressed H264 footage which simply does not represent the footage consumer/prosumer grade cameras capture, both 1080P and 4K.

I went from i9-9900K to Ryzen 9 3900X, because of the extra grunt many YouTube reviewers were claiming was amazing for workstation scenarios.

My Adobe Premiere timeline is BETTER than before, with multi-layered effects, both Adobe internal and 3rd party, many of which use GPU as well as CPU.

Optical flow conversion (that is effectively Adobe's alternative to Twixtor) is amazingly fast! Render times are immensely rapid compared to before and the whole PC just generally really does feel a lot quicker. On top of this, I did not have to re-install my PC. Simply had to go through and remove intel drivers where possible and then installed the AMD drivers from their website. Then simply updated Adobe Premiere to latest update, everything is beautiful, no crashes, and am beginning to enjoy using Premiere again.

Though quite a few colleagues have moved over to DaVinci and Nuke to get away from Adobe platform due to poor utilisation of hardware ie cpu cores etc., and I am now also learning DaVinci, Nuke soon, and will be switching over completely in a couple of months. If Adobe ever decides to improve their aging application designs and bring their Creative Cloud platform fully into the multicore future then indeed like all things, I will go where the speed is, no matter what the branding, a learning curve is part and parcel of being in the fastest most adaptable workflow.

Well done AMD, you now have a customer who left you way back in the Athlon days, if you can continue to give us MORE at sensible pricing, then I will stay with you.
Also well done ASUS for making the RGB-free X570 workstation motherboard, absolutely love this board, some were complaining about the onboard chipset fan, but honestly, it rarely comes on and when it does, you don't hear it! Plus I get to have a x8 PCIE card running at X8 alongside my x16 GPU running at x16.

Sorry for the long winded nature of the above. Bottom line, AMD Ryzen 9 IS the REAL deal!

ps. my son inherited the 9900K, he is a gamer and likes a few extra frames.

https://uploads.disquscdn.c...
https://uploads.disquscdn.c...

Posted on 2019-09-01 13:59:07