Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1561
Article Thumbnail

Video Editing Performance with Intel Xeon W-3200 Series Processors

Written on August 21, 2019 by Matt Bach
Share:

Introduction

In the PC industry, there are typically three different classes of hardware: Consumer, Workstation, and Server. There is often a lot of crossover between these categories, but the current breakdown of Intel's more popular processor lines is as follows:

  • Intel 9th Gen = Consumer
  • Intel X-series = Consumer / Workstation (unofficial)
  • Intel Xeon W = Workstation
  • Intel Xeon SP = Server

Of course, this doesn't mean you can't use an Intel 9th Gen processor in a workstation (we often do for lightly-threaded workloads), but as far as the intent of the product line goes, this is a fairly reasonable - if simplified - summary.

One of the biggest overlaps tends to be between the Intel X-series and Xeon W product families. The X-series CPUs are much less expensive, but the new Xeon W-3200 processors are available in even higher core counts (28 vs 18 cores), support Registered ECC memory, and have 64 PCI-E lanes which often makes them a much better choice for quad-GPU workstations.

While looking at straight performance is not really a fair comparison due to these additional features, here is how the Xeon W-3200 processors we will be testing in this article compare in terms of basic specs to their X-series counterparts:

Xeon X-series
Intel Xeon W-3225 8 Core
3.7GHz (4.4GHz Turbo)
MSRP: $1,199
Intel Core i7 9800X 8 Core
3.8GHz (4.5GHz Turbo)
MSRP: $589
Intel Xeon W-3245 16 Core
3.3GHz (4.6GHz Turbo)
MSRP: $1,999
Intel Core i9 9960X 16 Core
3.1GHz (4.5GHz Turbo)
MSRP: $1,684
Intel Xeon W-3265 24 Core
2.7GHz (4.6GHz Turbo)
MSRP: $3,349
No Equivalent

As you can see, there is certainly a price premium for the Xeon CPUs, although there is an argument to be made that the Xeon W-3245 is actually priced in line with the X-series processors. Compared to the i9 9960X 16 Core, the W-3245 has a slightly higher maximum Turbo Boost frequency as well as a higher base clock which generally means a higher all-core Turbo frequency. The price is actually the same(ish) as the Core i9 9980XE, which essentially makes it a higher frequency, lower core count version of that model.

Given the fact that the vast majority of apps do not scale well enough to really take advantage of this high of a core count, this may actually make the W-3245 better than the X-series processors from a price-to-performance standpoint in many applications.

Intel Xeon W-3200 series processors for video editing

While the choice between these two product lines is often decided by more than straight-up performance, it is still useful to know exactly how the performance compares. To that end, we decided to benchmark the Intel X-series, Intel Xeon W-3200 series, as well as the AMD Threadripper processors in a range of applications including Premiere Pro, After Effects, Photoshop, and DaVinci Resolve.

Note that this article shares testing data with our recent "CPU Roundup: AMD Ryzen 3, AMD Threadripper 2, Intel 9th Gen, Intel X-series" articles for Premiere Pro, After Effects, DaVinci Resolve, and Photoshop. So if you wish to know how these Xeon W-3200 processors compare against any of those product lines, the scores can be directly compared.

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

Looking for a Video Editing 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:

AMD Threadripper Test Platform
CPU AMD TR 2990WX
AMD TR 2970WX
AMD TR 2950X
AMD TR 2920X
CPU Cooler Corsair Hydro Series H80i v2
Motherboard Gigabyte X399 AORUS Xtreme
RAM 8x DDR4-2666 16GB (128GB total)
Shared Specs
Video Card NVIDIA Titan RTX 24GB or
NVIDIA GeForce RTX 2080 Ti 11GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit (version 1903)

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

The benchmarks we will be using are the latest release of our public benchmarks, plus NeatBench 5. Full details on what is tested as well as links to download and run them yourself are available at:

Raw Benchmark Data

While our benchmarks present 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, examining the raw results for that task is going to be much more applicable than any of the overall scores. Feel free to skip to the next sections for our analysis of these results if you rather get a wider view of how each CPU performs.

Premiere Pro Benchmark Results

Intel Xeon W-3200 Series Processors Premiere Pro Performance Benchmark

Starting off with Premiere Pro, the Intel Xeon W-3200 series processors perform exactly how you would expect them to given their specs. Xeon processors are of course more expensive than their Core counterparts due to the additional features they support, but it is good to see that from a performance standpoint, they are right on the money.

In fact, if you are looking for the absolute best performance in Premiere Pro, the Xeon W-3265 is about 5% faster than the Core i9 9980XE. You will be paying a hefty premium for it since the MSRP for the W-3265 is about $1,400 higher than the i9 9980XE ($3,349 vs $1,979), but if money is no object then the W-3265 is currently the fastest CPU we have tested for Premiere Pro.

Neat Video NeatBench 5 Benchmark Results

Intel Xeon W-3200 Series Processors Neat Video NeatBench 5 Performance Benchmark

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 a combination of both, but since using the GPU often overshadows the CPU we decided to just look at the "Max CPU Only" result.

Here, the Intel Xeon W-3200 series processors do extremely well. We expected this from the W-3265 24 Core processor due to the fact that Neat Video scales well with higher core counts, but a 20% improvement over the Core i9 9980XE is more than we anticipated. The W-3245 16 Core also did better than we expected, coming in at ~6% faster than the Core i9 9960X 16 Core. In fact, the W-3245 even holds up on a pure price-to-performance standpoint as it costs about the same as the Core i9 9980XE, yet is slightly faster.

Overall, this is a great win for the Intel Xeon W-3200 series processors. You typically expect to give up a certain amount of "performance-per-dollar" value in order to gain ECC memory and 64 PCI-E lane support, but at least for this benchmark, that actually isn't a major concern.

After Effects Benchmark Results

Intel Xeon W-3200 Series Processors After Effects Performance Benchmark

Unlike Premiere Pro and Neat Video, in most cases After Effects is not terribly great at using even a moderate number of CPU cores. Theoretically this shouldn't be a problem for the Xeon W-3200 processors since they have similar core counts and Turbo Boost frequencies as the X-series CPUs, but oddly enough, the Xeon W-3200 processors actually did very poorly in our After Effects benchmark.

We don't really have an explanation for why the performance was so poor at the moment, but the results were consistent across multiple runs of the benchmark. After Effects can be sensitive to memory performance, so it may be due to an issue between AE and the 6-channel memory, but it will require further investigation to find out for sure.

However, one thing we will point out is that we saw a similar performance issue with our DaVinci Resolve Fusion benchmark in the next section. Fusion utilizes hardware in a very similar way to After Effects, so whatever the issue is, it appears to not be isolated to just After Effects.

Intel Xeon W-3200 Series Processors After Effects Render Node Performance

One way to get After Effects to take advantage of higher core counts when rendering is to spawn multiple render instances with aerender using either a custom script or with plug-ins like RenderGarden. Our AE Render Node benchmark is designed to mimic the more popular plugins available for this, and allows the Xeon W-3200 processors to catch up with the X-series processors.

One thing that doesn't show up in the benchmark results is the fact that the Xeon W-3220 processors were slower than the X-series with only one or two render threads, but the performance gap got smaller and smaller as more threads were spawned. With higher core count CPUs, there is a certain point where you will no longer see a performance gain when you add more cores, resulting in a "soft cap" for performance. So, while these results are in many ways the complete opposite of the "standard" After Effects benchmark, the results actually make perfect sense when you factor in the fact that there is often a "soft" performance wall.

DaVinci Resolve Studio Benchmark Results

DaVinci Resolve is a very GPU-focused application, but there is still plenty of need for a powerful CPU. On the whole, the results are a bit mixed for these tests. For our 4K media test, the Intel Xeon W-3200 processors are right in line with the Intel X-series. The additional cores on the W-3265 24 Core even allow it to surpass the Core i9 9980XE by a small margin making it the fastest CPU we have tested to date.

However, the W-3265 in particular seems to hit a wall in our 8K test. We have no idea what happened (all these tests run at the same time so there is no chance we accidentally changed a BIOS setting or anything like that), but the W-3265 was almost exactly on par with the W-3245 in the 8K media test. The most likely cause is that we are simply GPU-limited in this test, and adding a second Titan RTX card would allow the W-3265 to really stretch its legs.

Fusion is another area where the Xeon W-3200 CPUs fare poorly. Fusion utilizes hardware in a very similar to After Effects, and it appears that there may currently be some issues with either single-threaded performance or thread scheduling with these Xeon W processors that results in poor performance in both After Effects and Fusion.

Photoshop Benchmark Results

Intel Xeon W-3200 Series Processors Photoshop Performance Benchmark

Photoshop may not be a true video editing application (unless you are one of the rare people that uses the video timeline features), but it is very commonly used alongside Premiere Pro, After Effects, and similar apps so we decided to throw it into the mix.

Photoshop doesn't scale particularly well with higher core counts, so really the performance isn't anything terrific with the Xeon W processors, but not all that bad either. The W-3265 is really the only one that performed a bit worse than you may expect, but since it runs at a lower frequency than the other models, it makes sense that it won't be great in a lightly threaded application like Photoshop.

How well do the Intel Xeon W-3200 processors handle video editing?

Outside of a few odd performance results in After Effects and Fusion that we want to investigate further, the Intel Xeon W-3200 series processors performed largely on par with similarly specced Intel X-series CPUs. This is actually a bigger win for Intel than it may appear since previous Xeon W CPUs often had problems with applications that did not have terrific core scaling. This time around, however, the high Turbo Boost speeds on these new Xeon W-3200 series processors largely removes that concern.

From a price-to-performance standpoint, the Intel X-series processors are still going to be the winner in most cases, although you can often argue that the W-3245 and W-3265 keep up even with that metric. However, Intel's Xeon line has never been about getting the best performance for your dollar since you gain features like Registered ECC memory support and 64 PCI-E lanes in compensation for the higher price.

Overall, it is great to see that you shouldn't need to worry about giving up performance by going with a Xeon W-3200 processor over one of the more "budget-friendly" Intel X-series processors. Motion Graphics and VFX artists may want to hold off for a bit until we figure out what is going on with AE and Fusion, but for editors, the choice of which CPU product line to use is largely going to come down to whether the higher cost of the W-3200 series processors is worth the Xeon-exclusive features like Registered ECC memory support or 64 PCI-E lanes.

Looking for a Video Editing 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: Video Editing, Premiere Pro, After Effects, DaVinci Resolve, Photoshop, Intel vs AMD, Xeon W, W-3225, W-3245, W-3265
Misha Engel

Which camera are you using to get 8K H.265 100 mbps and what is the chroma sub sampling?
What is the compression ratio of the 8K RED material?

Wouldn't it be an idea to ad BRAW with the Davinci Resolve test since you get a free Davinci Resolve Studio copy with a BMD camera that shoot resolutions of 4k or higher. BMD has sold more BMPCC 4K than all other RAW capable camera's together since it's introduction. The same is currently happening with the BMPCC 6K.

Posted on 2019-08-21 20:42:33

The H.265 footage is transcoded from RED (as noted on the benchmark pages) and is using just the default AME/Pr HEVC 8K UHD preset, but I'm not sure what the chroma subsample rate is. Both 8K H.265 footage native from a camera and BRAW are something we want to add in the future, we just need to either get our hands on a camera that supports those or find someone who is willing to record some test clips for us. If you know of anyone who would be interested in helping us out with that, definitely let us know!

Posted on 2019-08-21 20:49:47
Misha Engel

Test Media (59.94 FPS)

4K CinemaRaw Light
4K H.264 150mbps 8-bit
4K ProRes 422 16-bit This does not exist according to Apple.
4K ProRes 4444 16-bit This does not exist according to Apple
4K RED What is the compression ratio? 3:1 Good quality for 4k RAW all the way to crap quality 22:1
8K H.265 100mbps Only a complete idiot converts 8K RED to 8k h.265 100 mbps and than starts editing (8k RAW will only give you a perfect demosaiced image to 6k rgb at best).
H.265 is a long GOP delivery format and when it comes from a camera at 422(decent camera standard) no help from the GPU and it is most of the time converted to ProRes, DNX or cineform codec to edit it when you have no super computer.

8K RED What is the compression ratio?12:1 only when needed for 8k all the way to crap quality 22:1(linus youtube stuff).

The film standard is 24 fps at the highest quality (true for indies and studio's for movies, docs, commercials and other no youtube vlogger audience content).
These people tend to use ARRIRAW(uncompressed), RED(lowest compression possible), Sony(lowest compression possible, etc...
Your DR test where often quoted on many forums and I liked them a lot up to and including Davinci Resolve 15 (nice mix of practical codecs and resolutions, not perfect but pretty good).
Above mentioned frame rate and codecs(of which some don't exist and one can only be reached by converting an other format) gives me the impression the Puget Systems is only focussing on youtubers/vloggers who don't give about image quality.

Just go to redusers.net and you will find many people who shoot RED camera's on a daily basis and they tend to shoot at the highest resolution possible, with the lowest compression ratio(compression kills the image) and when ever possible never above 30 fps (24 fps is the standard).
The biggest question for a reduser when it comes to post processing is, what kind of CPU do I need to decompress 8k 24 fps 5:1 in realtime (answer i9-9960X or 7960X running 4.4..4.5 GHz all-core) on Davinci Resolve (Still waiting on an NLE to support the GPU decoding).

When you need BRAW material you can download it form the blackmagic website or search 5 seconds on youtube/google and you can download BRAW files ranging from Full-HD all the way up to 6k with frame rates from 0 to 300 fps at compression ratio's Q0, Q5, 3:1, 5:1, 8:1 and 12:1.

One more question, what kind of effects did you throw at the 8k material that you needed 20 GByte of VRAM (we often rent heliums and do really heavy noise reduction and never ran out of memory with our 16GB GPU's)?

Posted on 2019-08-22 01:46:32
Emilio Gonzalez

Hi Matt Bach, you can download 6k and 4k Braw files from the official product page in the blackmagic design website. As always, thank you for the great articles.

Posted on 2019-08-22 15:52:30

We've seen those and even done some testing with them as well. The problem we have is that we need to be able to distribute the test media we use since we have our benchmarks available for public download. We've reached out to Blackmagic asking for permission, but haven't received a reply. We have a similar situation for ARRI footage as well.

Usually we can find someone who is willing to record us a few test clips, but so far haven't had any luck for BRAW. If you know of someone who could help us out, definitely toss me an email at mattbach@pugetsystems.com ! Worst case, we'll just rent a camera, but we simply don't have the bandwidth to take that on at the moment. Definitely something we will do in the future, however!

Posted on 2019-08-27 22:50:46
Emilio Gonzalez

I have a ursa mini pro 4.6k and I can definitely provide some footage to use in the tests. I will try to record something next week and then get in touch by email.

Posted on 2019-08-28 12:38:37

That would be perfect, I really appreciate it! Ideally what we are looking for are clips that are as close to ~14 seconds long as you can (we only need 12 seconds, but that gives us a bit of extra without making the file too large). A zooming/panning/moving shot of anything with a decent amount of color should work well - we mostly just need things constantly changing in frame and decent amount of different colors so that we can do more with it in terms of grading in Resolve.

We need UHD (or true 4K, either way) clips for both 29.97 and 59.94 FPS. I'm honestly not sure what compression ratio is the most common to use for BRAW, so I will bow to your judgement on that. I imagine it will be either 3:1 or 5:1, but again, your call. I would prefer to avoid uncompressed for now, however, simply due to the file sizes we would have to deal with.

This is an example of the RED footage we are using: https://www.youtube.com/wat... or the H.264 we are now using: https://www.youtube.com/wat...

If you have any questions, feel free to reach out to me via email (just to avoid cluttering these comments). mattbach@pugetsystems.com

Posted on 2019-08-28 17:14:47
Zi Ga

Nice review, i was waiting for that, can you tell me please which motherbaord did you use? Do you think if you would using Render Garden on W-3265 cpu for After effects the resoults would be better for xeon cpu?

best

Posted on 2019-08-22 12:18:34

Motherboard and other specs are in the test setup section. Render Garden should be almost identical to our AE render node test. It may end up slightly different since you need to have AE running to launch Render Garden (which will eat up some RAM/VRAM) and I don't think it exports to .PSD, but for all practical purposes it should be very close.

Posted on 2019-08-22 13:24:55
Diede

As Resolve does multiple things, I would like to see a different chart what does for editing compared to grading (you are already doing Fusion). It is unclear what these results mean in that perspective. I'm particularly interested in the playback on the editing part of resolve...

Posted on 2019-08-23 14:23:29
Joe Sherman

Couple Questions: do you think Foundry nuke behaves like After effects? Seems these programs like single core speed. How do you explain the Zeon W 3225 speed when it has the highest clock speed. How come come there are so few Zeon W 3200 series compatible motherboards out there. The Asus board in your article is not even listed on Asus site. Super Micro has only one compatible board available! best joe

Posted on 2019-08-27 22:37:58

From what I understand, Nuke should be relatively similar to After Effects for live playback performance. Exporting may be better threaded to take advantage of more CPU cores, but that isn't something we have tested (yet).

The 3225 does pretty good, but it is really only in applications that are very lightly threaded that it keeps up with the other Xeon W processors. On those apps, the CPUs should be running at close to the maximum Turbo Boost, where the 3225 is only .2GHz slower. There is also the fact that the Turbo frequencies can change based on power draw and thermal limits, so a lower core count CPU with otherwise identical specs can at times end up running at a higher frequency than a higher core count CPU.

As for the available motherboards, we hear you on that - we wish there were more boards available as well! The Asus board is listed on their site though, you just need to make sure you are looking at their commercial products: https://www.asus.com/Commer...

Posted on 2019-08-27 22:47:17
Joe Sherman

This is a great discussion Matt! If you look at almost every movie being made today seems Nuke is an integral tool. I use it as a digital blackboard and it works well with digital camera's such as my Arri Mini. Filmlight makes a plug-in for Nuke that is amazing but once you attach a Baselight color node to some footage your lucky to get 2fps performance out of a good system. Using this combo is the only way to work in the Filmlight ecosystems on a windows machine. Nuke gives you a reel time window of computer resources being used https://uploads.disquscdn.c...

See the lower right corner of the attached photo for resources being used...the bottleneck is all CPU. Seems like I spend my life waiting on the CPU! Any suggestions? best joe

Posted on 2019-08-27 23:26:54