Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1719
Article Thumbnail

Premiere Pro performance: PC Workstation vs Mac Pro (2019)

Written on April 9, 2020 by Matt Bach
Share:

TL;DR: Mac Pro (2019) vs a PC Workstation for Premiere Pro

Compared to a $20,000 Mac Pro, it only takes about $4,000 with a PC based around the AMD Ryzen 3900X CPU in order to get similar performance in Premiere Pro. And if you opt instead for an even more powerful PC using an AMD Threadripper 3970X, you can get up to 50% higher performance compared to the Mac Pro for roughly a quarter the cost!

We understand that there is a lot of benefit to staying in the Apple ecosystem if you also have an iPhone, MacBook, etc., but that is an absolutely massive price and performance advantage for going with a PC workstation over a Mac Pro.

Introduction

Apple's new Mac Pro is a solidly engineered system, but its hefty price tag meant that it took us a while to get our hands on one. Thanks to our friends at Linus Media Group (creators of Linus Tech Tips), however, we are finally able to see how well it performs in Adobe Premiere Pro!

Anytime we do Mac vs PC testing, things can easily become a bit heated in our comments section, so we are going to largely side-step the question of whether you should use a Mac or a PC. Here at Puget Systems, we rarely try to outright convince anyone to move from Mac to PC, but rather take the standpoint of offering as much information as we can so that you can make an informed decision for your situation. If your workflow is better on a Mac, that is great! But if you are ready and able to take advantage of the higher performance, lower cost, and greater customization of a PC and want to work with a company that is intimately familiar with the transition from Mac to PC, we are here to help!

Mac Pro vs PC workstation for Adobe Premiere Pro

In this article, we will be examining the performance of a new 2019 Mac Pro in Premiere Pro compared to a range of PC workstation configurations using AMD Threadripper, AMD Ryzen, Intel X-series, and Intel 9th Gen processors. In addition, we will be including an iMac Pro as well as a previous generation Mac Pro as additional reference points. If you are interested in how the Mac Pro compares in other applications, we also have other articles for Photoshop and After Effects available on our article listing page.

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

Looking for a Premiere Pro Workstation?

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

Configure a System!

Labs Consultation Service

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

Find Out More!

Test Setup

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

AMD Ryzen Test Platform
CPU

AMD Ryzen 9 3950X ($749)
AMD Ryzen 9 3900X ($499)

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

Intel Core i9 9900K ($499)

CPU Cooler Noctua NH-U12S
Motherboard Gigabyte Z390 Designare
RAM 4x DDR4-2666 16GB (64GB total)
AMD Threadripper 3rd Gen Test Platform
CPU AMD TR 3990X ($3,990)
AMD TR 3970X ($1,999)
AMD TR 3960X ($1,399)
CPU Cooler Noctua NH-U14S TR4-SP3
Motherboard Gigabyte TRX40 AORUS PRO WIFI
RAM 4x DDR4-2933 16GB (64GB total)
Intel X-10000 Series Test Platform
CPU Intel Core i9 10980XE ($979)
Intel Core i9 10940X ($784)
Intel Core i9 10920X ($689)
Intel Core i9 10900X ($590)
CPU Cooler Noctua NH-U12DX i4
Motherboard Gigabyte X299 Designare EX
RAM 4x DDR4-2933 16GB (64GB total)
Shared PC Hardware/Software
Video Card NVIDIA GeForce RTX 2080 Ti 11GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit (version 1909)
Mac Pro (2019) iMac Pro Mac Pro (2013)
CPU 28‑core Intel Xeon W 2.5GHz 14-core Intel Xeon W
2.5GHz
12-core
2.7GHz
RAM 96GB DDR4 ECC 64GB 2666MHz DDR4 ECC 64GB 1866MHz DDR3 ECC
Video Card Radeon Pro Vega II Duo 2x32GB Radeon Pro Vega 64 16GB Dual AMD FirePro D700 6GB
Hard Drive 1TB SSD storage 1TB SSD 1TB PCIe-based SSD
OS Mac OS X (10.15.4)
Current Price $19,599 $7,549 $3,549
(used from OWC)

*All the latest drivers, OS updates, BIOS, and firmware applied as of March 30th, 2020

In order to see how each of these configurations performs in Premiere Pro, we will be using our PugetBench for Premiere Pro V0.9 benchmark and Premiere Pro 2020 (14.0.4). This benchmark version includes the ability to upload the results to our online database, so if you want to know how your own system compares, you can download and run the benchmark yourself.

We went back and forth about including pricing, but since that is one of the biggest downsides to using a Mac, we opted to go ahead and include the pricing for the Mac systems we are using in this article. The Mac Pro (2019) and iMac Pro pricing is pulled directly from Apple.com at the date of this post, but since the Mac Pro (2013) is no longer sold, we went with what OWC is currently selling a used unit for with the same specs.

However, keep in mind that some of the configuration options for the Mac Pro we have may not do a ton for Premiere Pro. For example, while Premiere Pro technically can utilize multiple GPUs, we typically don't see much of a benefit unless the GPUs are relatively low-end. If you want to actually get a Mac Pro for Premiere Pro, you may get similar performance with a single GPU rather the dual GPUs we have in this Mac Pro and save $2,800.

But no matter how you slice it, the Mac Pro is not a cheap system. To put it into context, here is the pricing for a handful of PC configurations we will be comparing it to:

Of course, you can certainly increase the price by adding things like additional storage, but for a semi-direct comparison to the Mac it should be in the ballpark. What it comes down to is that compared to the Mac Pro we are using, even the most expensive PC configuration we will be testing is less than half the cost. If you went with the most basic configuration you can get for the Mac Pro, the price (~$6k) is going to be roughly the same as the AMD Threadripper 3970X 32-core system.

If you are looking at just pure price-to-performance (which admittedly breaks down as workstations get more and more expensive), the Mac Pro better be pretty amazing to justify its hefty price tag.

Benchmark Results

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

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

Premiere Pro Benchmark Analysis

Premiere Pro is fairly decent at leveraging the power of higher core count CPUs, which is why AMD's Threadripper processors with 24, 32 and even 64 cores are at the top of the chart. Even so, the 28 Core Mac Pro does not give particularly great performance in Premiere Pro.

Overall, the new 2019 Mac Pro only performs on par with the AMD Ryzen 9 3900X 12 Core configuration which is about 1/5 the cost. To be completely honest, we expected much better performance out of the Mac Pro and are a bit surprised to see how relatively poor it performed given it's specs on paper. It is certainly faster than the 14 Core iMac Pro (although only by about 12%), but we expected the Mac Pro with twice the cores and dual GPUs to fare much better.

If you are looking for a Premiere Pro workstation, it almost seems ridiculous to claim that a PC with an AMD Threadripper 3970X CPU and an NVIDIA GeForce 2080 Ti will be 50% faster than a Mac Pro at 1/3 the cost, but here we are. To be fair, the difference for live playback - which most people value over export speed - is only about 12% faster with a 3970X-based system, but the much lower export performance with the Mac Pro even though it costs 3x more is... not great.

Mac Pro (2019) vs a PC Workstation for Premiere Pro

Since there are so many reasons why either a Mac or a PC may be right for you, we generally try to focus on the straight performance results and not tell you which you should purchase. But in this case, the Mac Pro is so underwhelming that it is hard to not simply say "Don't buy a Mac Pro for Premiere Pro".

This isn't like our Photoshop testing where the Mac Pro was only a hair slower than a PC, or our After Effects testing where a PC can easily be 20% faster at a much lower cost. This time, we are talking a PC being up to 50% faster on average for 1/3 the cost. We understand that there is a lot of benefit to staying in the Apple ecosystem if you also have an iPhone, MacBook, etc., but that is a huge amount of performance and cost savings you will be giving up to get a Mac Pro.

By skipping the Mac Pro and going with a PC, you could easily save $14,000 which could be used for a host of other things to improve your workflow. Maybe you can finally upgrade your reference monitor to a really nice Eizo or Flanders Scientific model. Or use it as an opportunity to move to a central NAS storage unit from LumaForge. Or just take a couple months off to recharge. And this isn't taking into account the amount of money you might be able to earn due to the higher performance of a PC.

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

Looking for a Premiere Pro Workstation?

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

Configure a System!

Labs Consultation Service

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

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

Did you load windows on the Mac Pros to see if it is the hardware or the OS that is so doggy?

Posted on 2020-04-11 03:13:46

Wow, I knew Apple's stuff is overpriced, but this is beyond crazy.

Posted on 2020-04-11 12:11:47
Dennis Eustace

Yeah Apple is expensive but super reliable. I've had a laptop for ten years with out any issues. I agree PCs are more bang for the buck, but how long will they last? need to get repaired? A lot of the performance with Adobe is the Nvidia card. I wish I could use one in a Mac! I just went back to PC from Mac for a computer for After Effects and right out of the box it had issues and needed to be sent back. Now I 'm wondering if I should have then repair it or get my money back and buy a less powerful Mac that I know will last trouble free for many years.

Posted on 2020-08-05 21:50:28

Oh, when it comes to laptops, sure, I can see Macbooks being a better option. I'm more of a desktop person though, I build my own PCs, and I rarely have any problems. But for sure, Apple offers well build devices. I just think in terms of the raw performance, they could be a bit better. It's a shame that Apple doesn't offer Nvidia GPUs anymore.

Posted on 2020-08-06 13:17:35
Jamie Grant

Hello im considering upgrading from a 7900x to a 10920x or a 10940x. I use premiere pro, after effects an cinema 4d alot. is the performance of the 10940x worth the extra $150 from the 10920x. or is it best to save the extra dollars and go with the 10920x.

Posted on 2020-04-13 12:22:52

Going from the 1092X to the 10940X is usually a pretty modest upgrade (5% or so), but really only you can make that call. $150 for 5% more performance may not be worth it for some, but for others having 5% higher performance will pay for itself in a matter of weeks. Either is a solid choice, it just comes down to your budget and how important that extra bump in performance is.

Posted on 2020-04-13 16:46:17
David Farkas

Nice work as always, Matt. Wondered if you guys would be getting around to this comparison (and please do one for Lightroom too!).

Just curious if the Mac Pro was running Metal or OpenCL for GPU acceleration in Premiere. I've seen other comparisons showing huge gains with Metal even on the MacBook Pro. Personally, I'm using a TR3960X with RTX2080 Super for Premiere, and expected performance to be on par or slightly ahead of the Mac Pro 28-core. I honestly didn't expect to see up to 50% better performance. So, just wondering if OpenCL might have been holding the Mac Pro back here. If not, well.... glad I went with Threadripper and Nvidia for 1/4 the price!

Posted on 2020-04-13 18:01:48

We really hope to have our Lightroom Classic benchmark working on MacOS "soon", but it may take a bit. There is a lot of testing that we have to use external keyboard/mouse automation for, which isn't ideal to do oin MacOS due to all the additional security they have.

On the Mac systems, we used Metal exclusively since OpenCL is now considered depreciated. We have actually seen slightly better performance with OpenCL in the past, but Adobe has done a lot of work with Metal recently. But since OpenCL is legacy now, we have no plans of testing with anything other than Metal on Mac.

Something else that I am working on a post for right now is that Premiere Pro Beta has added GPU accelerated encoding for H.264/HEVC on Windows. I'm sure they will add it on MacOS at some point, but I'm seeing anywhere from a 2-4x improvement in export performance with it. So at least in the short term, it looks like a PC will be actually end up even faster than we shows in this post (at least for exporting).

Posted on 2020-04-13 18:31:32
David Farkas

Thanks for clarifying.

I'd really like to see Adobe further optimize Premiere Pro to fully load all cores like Blender does. Even Lightroom now can fully load all cores to 100% during export, which has been notoriously poor for multi-core support.

Coming from a 10-core 6950X where all cores were 100% loaded, I'm maxing out around 80% on 24-cores with Threadripper during 4K H264 export from ProRes 422 using Lumetri adjustments with LUT. And the GPU is running around the same. So, rendering is now not CPU limited nor GPU limited as each has room to run. Even without dedicated encoding support from the GPU, there is still room for improvement for sure.

Will the hardware GPU acceleration (NVEC?) impact image quality as it does for streaming or when using QuickSync?

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

Posted on 2020-04-13 19:27:53

Honestly, I'm not sure that focus on multi-threading performance is the right approach anymore. Some things like CPU-based rendering, sure, but more and more it is about utilizing both the CPU and GPU together. GPUs are often a much better choice for tasks that can run in parallel since GPUs have so many more cores than a CPU (thousands vs dozens). So for the tasks that can be efficiently multi-threaded, it is way better for everyone for the developer to port it to the GPU rather than the CPU. But what that leaves are all the single or lightly-threaded tasks to be run on the CPU. The end result is an application that can only utilize a handful of cores effectively, but since it is using both the CPU and GPU, it could be several times faster than if the developer solely focused on CPU multi-threading. You may not see 100% load on either in task manager (which isn't completely accurate anyway), but that is often due to the fact that things are having to move back and forth from the CPU to the GPU.

This doesn't always work of course - sometimes you need more memory than what GPUs are currently capable of. Or sometimes the tasks can be scaled efficiently up to a certain number of cores, but nowhere near the number of cores that a GPU has. Or probably a whole host of things I'm not even familiar with.

Specialized hardware like the NVENC encoding is a totally different matter, but it is along the same lines of moving tasks to the hardware that is best suited to run it. It will make it look like Premiere Pro got worse if you just look at the CPU load, but the reality is a massive speedup.

As for quality loss with NVENC, I'm going to upload the exports for people to be able to look for themselves, but even zooming in to 400%, I had a hard time deciding whether it was actually lower quality or not. Different, yes, but I don't know if I could pick the "better" one in a blind test. But, this is only with a handful of clips - more motion, patterns, etc. could show a larger difference. I'm sure bitrate will be a big impact as well since software only should be better at lower bitrates.

Posted on 2020-04-13 20:45:40
Miroslav Goshev

Awesome, are you planning to make similar
test for Davinci Resolve?

Posted on 2020-04-14 06:27:34

Yes, but it is going to be at least a couple of weeks. I have the data from the Mac Pro already, but we usually swap the GPUs in our test beds to the Titan RTX so that we can run the 8K tests (you need a lot of VRAM to work with 8K timelines in Resolve) which means that it takes a lot longer to run those benchmarks than the Adobe suite.

Posted on 2020-04-14 17:50:35
GA Video

If it was not for the last release, I would abandon Resolve. I do not think it makes sense to run benchmarks on something with such poorly written OS/hardware interface. I would like to see someone creating a profile of that application to see how bad it is in using computer hardware. I am running Windows 10 on a Intel i9 9980XE 3.0GHz Eighteen Core with 2xAsus GeForce RTX 2080 TI 11GB and was amused to see how much disk I/O and little CPU/GPU use that program had in the past. I should say that release 16.2 is much better and hope this will continue. I suspect that my computer is still an overkill for that application. Previous releases could probably work as well on a basic, original Pentium based computers as on a supercomputer. Having someone to profile Resolve performance would force its developers to up their game.

Posted on 2020-04-17 14:00:26

I am really, really, really looking forward to 8K Davinci Resolve benchmarks. It's kinda ironic that Apple cares so little about us 8K Resolve users that they cannot be bothered to make a $50,000 workstation that's worth a crumb, but it's kinda sad that none of the 3rd-party hardware vendors have taken up the cause by benchmarking what WE care about to help us find the solutions we are looking for (that Apple cannot be bothered to address).

Posted on 2020-04-15 13:17:28
GA Video

I do think any workstaion above $5,000 is an overkill because Resolve cannot take full advantage of hardware it is running on. I have adequate performance, editing simple music videos in HD at 60fps, running Resolve 16.2. I cannot imagine editing 8K with that application and believe a more powerful computer would not make any difference.

Posted on 2020-04-17 15:02:29
Technophobe01

My sense is that I'd like to see noise factored into the performance score.

Why? I have both a PC (Windows Xeon Processor 10 Pro setup) and a Mac Pro (24 Core setup)

I have to say that for me the deciding factor for use is noise.

The Mac Pro is totally silent under load, the PC setup is noisy under load. I know this might sound strange but it makes a huge difference. I hate fan noise.

The Puget Systems are excellent, the deciding factor wasn't price alone, it was the ability to be able to sit and think in silence. I use MacOS, Windows, Ubuntu at the Office, and have access to Ubuntu, SLES, and Redhat regularly via AWS, Azure.

For data analytics (CUDA) work I boot the MacPro into Windows and have an internal NVidia Titan card setup, actual production code is on AWS and Azure. The MacPro runs Windows10 with NVidia cards, the NVidia cards however cannot be used with MacOS, the drivers are not available.

So yes, you can get better performance but the combination of MacOSX, and silence works for me.

So, my sense is to ask how noisy is the PC benchmark system compared to a Mac Pro? I work with illustrator, photoshop etc and Premiere Pro as well as Maya.

I'd love to see the noise profile.

Posted on 2020-04-15 14:13:18
David Farkas

Good point. Noise normalized results would certainly be interesting. I think with proper fan curves, and with either Noctua Quiet PWM or BeQuiet Silent Wing 3 fans you can get near silence on a PC build. I'm running my Noctua fans at 800 RPM and staying under 80C during Premiere Export, with near silence, on a 24-core Threadripper. (see screenshot above) It's possible.

Posted on 2020-04-15 21:01:30

Your response highlights how hard this would be, David: there is sooo much variability in the realm of PC hardware! You could build the same performance in a system that sounded like a leafblower (maybe some exaggeration for emphasis) or an almost imperceptible background hum.

More practically, the areas we do our testing for these articles (our Puget Labs office and a server room) both have a lot of background noise from other sources which would make it impossible to isolate the sound of an individual workstation.

So as cool as such a comparison would be, we cannot do it at this time - and even if we could, folks would probably accuse us of either cherry-picking the quietest PC components unfairly / unrealistically... or else of *not* picking the quietest options, and resulting in a less-than-optimal result. Suffice it to say, however, that if what you are looking for is a very quiet PC we can certainly provide one - and I'm sure some other system builders can as well, along with the option to build it yourself :)

Posted on 2020-04-15 21:13:21
Technophobe01

William,

My sense would be to trust PugetSound to play fair, you have a reputation as a provider of quality hardware.

My interest was to look at the cost point of performance PC / GPU setup with a water cooled system with low velocity fans. My feeling (though I could be wrong) is that pushes the price up on the PC side, still interesting to look at. In my case I cheat, the PC is in separate office space - long cables are my friend.

Take care and stay safe - your work here is appreciated.

Posted on 2020-04-16 16:27:02

Thank you fort those kind words! :)

You are right that a full liquid-cooling setup would increase the cost of a PC, but personally that isn't the direction I would go for quiet operation. Upgrading to quiet fans and careful selection of other hardware (like hard drives, video cards, and the chassis) can turn most configurations into extremely quiet workstations without the cost, complexity, and leak risks that come with full liquid-cooling... and without adding another source of noise, in the form of the pump. In fact, most of the comparison points which Matt made in his article already include some of those quiet elements - and could be decked out the rest of the way for probably $100 or less (mostly just upgrading the chassis fans).

Posted on 2020-04-16 17:35:31
Zé Cotinha

It doesn't make any sense what you're talking about.
The noise of the PC fans is easily controlled by the BIOS / FIRMWARE of the motherboard, just go to it and adjust to the smallest possible.
There are also very quiet fans on the market that you can buy.

My criticism of Apple is that it does make silent computers, but at what cost? They are totally dependent on air conditioning in the room set at 18C. And I got tired of seeing imacs pro and macpro that simply lock or burn for hours at 100% usage, just because they prioritize the silence of the fans.

Posted on 2020-05-03 14:42:02
Technophobe01

Zé,

Thanks, I agree it may not make sense from your perspective, however we do not hold the same perspective. Ambient computer noise is for many an important selection criteria and we are happy to pay for silence. Apple has identified this as a market attribute and profited handsomely from the discovery. Chuckle.

You are not wrong but consider that the majority of people do not want to jump into setting up the BIOS / Firmware. That is exactly why we can recommend PugetSound Systems, we can rely upon them to do that, test the results and provide great support throughout the process.

From a technical perspective I could do as you suggest, I just do not want to invest my time in that direction. Why? Rather than invest the time in that particular expertise my approach is to outsource and rely upon the deeper expertise of PugetSound Systems, and Apple. Both offer me advantage they give me my time back which is a truly finite resource. I hoard my time.

Why? Consider, a life can be visualized as N rows and 52 columns, and average life will have 60 to 70 rows or 3,120 to 3,640 weeks. How we choose to allocate that time is important. My choice is to invest in others expertise, where I cannot afford to do that I invest in becoming expert enough to navigate the problem domain myself.

Stay safe, good karma flying your way.

Posted on 2020-05-03 16:06:50
Zé Cotinha

I understand that there are certain people who prefer to buy a computer already pre-configured with silent fans. This is great that there is a PugetSystem that sells silent workstations.
But PugetSystem computers, I can change the power of the fans.
In the IMacPro, the ventilation system is too precarious, too few fans. MacPro, on the other hand, is much more like a Customized PC with ideal ventilation.

I got tired of seeing iMacs crashing when rendering because of the temperature, on MacPro it doesn't.
Never recommend the iMacPro with i9 9900k to render videos, that is a time bomb and Apple will say that it was not made to stay hours at 100% usage and will not give you a guarantee.

That's why Hackintosh is growing, people can customize the fans and lower the CPU / GPU temperature.

Posted on 2020-05-04 13:54:36
GA Video

I have both Mac Pro (2013) and PC (Intel Core i9 9980XE 3.0GHz Eighteen Core, 2xAsus GeForce RTX 2080 TI 11GB). My PC is almost as quiet, under load, as Mac Pro (6 core Intel). The noisiest parts of my computer were disk drives during RAID 5 initialization. Right now the noisiest item in my office is the air conditioning coming through the floor air ducts. The level of computer noise was very important to me. After running my PC for few months, I am planning to convert my work to PC and get rid of Mac. The fact that DaVinci people may be getting their quality control act together may speed the whole process.

Posted on 2020-05-03 15:40:45
R.J. Leong

Sometimes, the choice of a GPU and the speed of RAM can make a HUGE difference in the benchmark scores. Just a few minutes ago I saw a benchmark result score with an AMD Ryzen 9 3950X, which is normally a good choice for Premiere Pro. But that user crippled that system by using only a Pascal GPU (specifically, a Quadro P4000) and running the system's RAM at only DDR4-2133 speed. The end result is a downright abysmal overall Standard preset score of 499. That's slower than a 14-core iMac Pro - and barely any faster than an old 2013 12-core Mac Pro!

In other words, if you're going to buy such a high-performance CPU for a custom Windows build, you'd better make sure all of the other components (RAM and GPU, in particular) are also well matched to the CPU in terms of relative performance. In the case of that aforementioned 3950X, the outdated GPU and the slow RAM had severely bottlenecked that system.

Posted on 2020-05-04 14:59:40
R.J. Leong

Someone recently tested low- and mid-range mainstream systems with mainstream CPUs over a period of a few days. The roundup is divided into two groups:

Low-end CPUs with 16 GB of DDR4-3200 RAM and a GeForce RTX 2080 SUPER:
AMD Ryzen 3 3100
AMD Ryzen 3 3300X
AMD Ryzen 5 1600
AMD Ryzen 5 3600 (initial test)

Lower-midlevel CPUs with 16 GB of DDR4-3600 RAM and a GeForce RTX 2080 Ti:
AMD Ryzen 5 3600
AMD Ryzen 7 3700X
AMD Ryzen 9 3900X
Intel i5-9600KF
Intel i7-9700K
Intel i7-9900K

All systems were running Premiere Pro 14.0.4. Not that it mattered much, but the retested Ryzen 5 3600, the Ryzen 7 3700X and the Ryzen 9 3900X were running Windows 10 Home while the others were running Windows 10 Pro (the AMD systems were running the latest 1909 update while the Intel systems were still running the 1903 version). The Intel systems either lack QuickSync or have the feature disabled.

Some of the results were surprising. The Intel systems were choked (and yes, that was me flashing a choke sign there, wrapping my two hands around my neck) by the disk storage subsystem used in those systems (presumably the same as the ones used in the AMD systems) as well as the low amount of system RAM. The HUGE surprise? The abnormally high overall Standard preset score of the Ryzen 5 3600 system: It beat every one of the other systems - both low-end and midrange. Now I figured out why: The abnormally high GPU score for the installation of the RTX 2080 SUPER that's in that PC. That skewed the overall Live Playback score higher as well. I threw out the 3600's overall score, declaring it an anomaly

And as I mentioned, the three Intel systems, as choked by the low amount of system RAM as they were, did not perform sufficiently higher than the two quad-core AMD low-end systems to justify their price tags. In fact, the i5-9600KF system actually underperformed the low-end AMD Ryzen 3 3300X in this roundup, while the i7-9700K barely scored higher than a three-year-old AMD Ryzen 5 1600 system. Even the i9-9900K underperformed the AMD Ryzen 3700X in this comparison. Still, I don't understand why the Ryzen 7 3700X performed nearly as well with only 16 GB of RAM as the Ryzen 7 3800X did in Puget Systems' own in-lab testing with that very same GPU but with 64 GB of RAM instead of just 16 GB.

Since then, the Ryzen 5 3600 was retested with the same components as the higher-end systems in this roundup, along with a new result from a Ryzen 9 3900X with the same RAM and GPU configuration. Not surprisingly, the 3600 falls a bit shy of that of the 9900K while although the 3900X performed quite well, it is clearly obvious that that CPU is bottlenecked by the low amount of RAM. The Intel CPUs were not only choked by such a low amount of RAM, but they were strangled by this one-rank-per-channel memory configuration: The 8 GB DIMMs, by this time, were only single-ranked (early models were double-ranked), and Intel's memory controller really needed two memory ranks per channel in order for the company's CPUs to perform like they should. It's as if Intel's memory controller has only a 32-bit path to each memory rank, with the two ranks in each DIMM slot connected in series instead of parallel at memory controller level, instead of the 64-bit path that it should have had. AMD's memory controller is much less impacted by the number of memory ranks per channel.

The moral of this story: While one can run Premiere Pro adequately with only 16 GB of RAM, don't settle for less than 32 GB of RAM if you want a pleasant editing experience with that program. With 32 GB, the Intel i9-9900K would have come out ahead of the 3700X with that same amount of RAM. And at the same time, the Ryzen 3 3300X with 32 GB of RAM would have edged out the old Ryzen 5 1600.

And Intel will be shipping a successor CPU platform to the Coffee Lake lineup over the coming months. However, the new CPUs are both electrically and physically incompatible with existing LGA 1151 motherboards: These new 10th-Generation mainstream desktop CPUs use the new LGA 1200 socket - and older CPUs are completely incompatible with it. Intel is sticking to its two-CPU-generation life cycle for this socket All of the i3, i5 and i7 CPUs for LGA 1200 will have hyperthreading, while the i9 for this socket will have 10 cores and 20 threads. This means that all 10th-Generation desktop i5 CPUs will have 6 cores and 12 threads, while all 10th-Generation mainstream desktop i7 CPUs will have 8 cores and 16 threads. Even the i3 for this socket will have 4 cores and 8 threads. This means that the new i7-10700K should perform very much like the current i9-9900K in these tests - at a price of around $374 (if Intel's pricing is to be believed). Even an i5-10600K should perform comparably to an i7-8700K; however, at a cost of $262, it is a bit overpriced for what it offers. In other words, Intel's initial pricing for the entire Comet Lake lineup, although lower than the Coffee Lake CPUs that those Comet Lake CPUs are replacing, is still a bit too high for the performance that they offer.

Posted on 2020-05-06 14:38:56
Marc Wielage

I'd love to see identical tests done on Resolve 16.

Posted on 2020-09-18 04:51:08

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

Posted on 2020-09-18 07:19:42
Marc Wielage

That doesn't test the 2019 Mac Pro with Resolve.

Posted on 2020-09-18 08:51:01

Ah, true - our Resolve benchmark doesn't yet support MacOS... but that is something Matt is working on :)

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

Posted on 2020-09-18 15:38:52

Yea, we are getting closer to a Mac supported version. Probably would be out right now if it wasn't for all the NVIDIA RTX 30-series launches going on right now. The Mac Pro we were borrowing for this test is long returned, but I'm hoping we can leverage our public benchmark database ( https://www.pugetsystems.co... ) and "crowd source" the Mac side of the tests.

Posted on 2020-09-18 17:22:40
Jose Santos

I would love to see a similar article but regarding Davinci Resolve! Are you guys planning on doing something of the sort?

Posted on 2020-09-27 16:06:02

We are getting closer to a Mac supported version of our Resolve benchmark, at which point we can do some Mac vs PC testing for it. Probably would be out right now if it wasn't for all the NVIDIA RTX 30-series launches going on right now.

The Mac Pro we were borrowing for this test is long returned, but I'm hoping we can leverage our public benchmark database ( https://www.pugetsystems.co... ) and "crowd source" the Mac side of the tests.

Posted on 2020-09-28 16:52:53