Mac Pro vs. iMac: Video Compression

Posted on January 3, 2014 by Larry

[ Please see my disclosure statement on product reviews. ]

There are a variety of excellent performance reviews of the new Mac Pro on a variety of sites, so I decided to compare the Mac Pro with an iMac from a different perspective: video compression. What I learned surprised me, as you’ll see in this article.

EXECUTIVE SUMMARY

The purpose of this test was to judge compression speed, not image quality, in an effort to compare these two systems; though compressed image quality seemed comparable between the two systems.

When running Apple Compressor 4.1 as a bench-mark, the new Mac Pro is faster for some compression tasks and significantly slower for others when compared to a recent model iMac. If video compression is your primary use for a new computer, you may be better off buying a top of the line iMac.

Take a look at the table below. Different compression tasks yield significantly different completion speeds. Select the system that meets the needs of the compression tasks you need to accomplish.

A NOTE ON HARDWARE ACCELERATION

One of the speed advantages of the iMac is that it uses an Intel technology called “QuickSync.” This is a special processor “engine” inside many consumer-grade Intel CPUs that accelerates H.264 compression for certain encoding settings; for example, when compressing for Apple devices, QuickTime or MPEG-4 movies using the H.264 codec. The Mac Pro Xeon CPU is considered “workstation-grade,” and doesn’t provide this hardware acceleration. This explains why the iMac is faster when encoding in single-pass mode, which enables hardware acceleration, but slower in multi-pass mode, which disables hardware acceleration.

Hardware acceleration is a two-edged sword. It is MUCH faster than software encoding. However, it only yields image quality and file sizes equal to single-pass encoding. This will often be fine for movies that don’t contain a lot of movement, such as screen captures; or movies where getting it done fast is better than image quality, such as news or digital dailies. However, hardware compression is generally not the best choice for movies with lots of movement between frames or where you need the highest image quality with the smallest file size.

WHAT I DID

I ran a series of 21 compression tasks on both a current model iMac and new Mac Pro, noting how long the compression took and the difference in file sizes created. I used four test files:

A ten-minute audio and video episode from “2 Reel Guys” which included a reasonable amount of movement between frames.
A two-and-a-half-minute audio and video talking head video from my YouTube channel.
A twelve-minute audio and video screen capture from my current Compressor 4.1 training.
A one-hour audio-only podcast from the Digital Production Buzz.

All videos were 720p ProRes 422 or ProRes 4444 files with uncompressed audio. The audio podcast was in uncompressed WAV format.

I created nine compression test settings:

The default YouTube 720p setting to create a QuickTime movie
The default Apple devices HD Best quality to create an MPEG-4 movie
The default DVD compression setting for audio and video, which scaled the HD video to SD
A custom QuickTime setting running in single-pass mode to test hardware acceleration
A custom QuickTime setting running in multi-pass mode, which turns hardware acceleration off.
A custom MPEG-4 setting running in single-pass mode to test hardware acceleration
A custom MPEG-4 setting running in multi-pass mode, which turns hardware acceleration off.
The default MP3 audio compression setting
The default AAC audio compression setting

All settings matched between the two computers. Both Compressor and Mavericks were running the latest version. The Mac Pro had its latest firmware update installed.

Here are the settings I used for the custom QuickTime setting: H.264 codec, 2000 kbps data rate, frame reordering on, keyframes every 90 frames.

Here are the settings I used for the custom MPEG-4 setting.

WHAT I LEARNED

Compression speeds varied depending upon the length and complexity of the source files, though compressed file sizes were essentially the same between the two computers (which I would expect). All files were stored and saved to the desktop.

NOTE: As measured by the Blackmagic Design Speed Test (BMD), the Mac Pro was roughly 5 times faster at reading and writing to the desktop than the iMac. This speed differential does not seem to be significant in compression.

The BAD NEWS

When compressing audio, the Mac Pro was about 15% slower than the iMac.
When using hardware acceleration, the Mac Pro was, on average, 29% slower than the iMac. (Speeds ranged from 2.8% faster to 73% slower.) This is due to the fact that the Mac Pro does not support hardware acceleration for compression.

The GOOD NEWS

When compressing using multi-pass settings, the Mac Pro was, on average, 45% faster than the iMac. (Speeds ranged from 33% to 60% faster.)
When compressing for YouTube, the Mac Pro was 37% faster.
When compressing for Apple devices, the Mac Pro was, on average, 24% faster. (Speeds ranged from 12% to 35% faster.)
The Mac Pro compressed MPEG-4 videos faster than QuickTime videos, when compared to the iMac.

Click the table to see a PDF of all my results.

Column definitions:

The left column describes the specific test
iMac 2013 are the timing results from the iMac computer
File size is the size of the compressed files from that test, created by the iMac
Mac Pro are the timing results from the Mac Pro computer
File size is the size of the compressed files from that test, created by the Mac Pro
Mac Pro Speed Difference is the difference in speed between the iMac and Mac Pro. Negative numbers indicate where the iMac is faster.
HW Accel. iMac Speed Diff. is the difference in speed when hardware acceleration is enabled. This column compares differences in speed with the iMac between accelerated and non-accelerated compression for the QuickTime and MPEG-4 custom setitngs. Not all compression settings support hardware acceleration.

NOTE: To compare the differences in hardware acceleration between the iMac and Mac Pro, look in the Mac Pro Speed Difference column. In all but one case, the Mac Pro is slower when hardware acceleration is turned on than the iMac.

Mac Pro Single vs Multi Speed Diff. compares the difference in speed between single pass and multi-pass for QuickTime and MPEG-4 custom settings; as the Mac Pro does not support hardware acceleration.
Notes indicates that the QuickTime compression setting created uncompressed audio, while the MPEG-4 setting created compressed audio. As this setting did not change between machines, I am ignoring audio compression when comparing results.

NOTES ON THE TEST

iMac
3.4 GHz i7
16 GB RAM

I am working with a new Mac Pro, which is on loan from Apple. (You can read my first review of it here.)

New Mac Pro
2.7 GHz 12-Core Xeon E5
32 GB RAM

I used the same compression settings on both computers. Timings were measured by Compressor and displayed in the Completed tab. One job was fully complete before the next job started. Two jobs never ran at the same time.

Compressor was run in single instance mode, which is its default setting. Though I didn’t test for this specifically, I discovered that for short movies, single instance mode is about 20% faster than multiple instance mode. This difference disappears as the duration of the source media increases.

NOTE: Running Compressor in multiple instance mode does not guarantee faster performance. In general, I recommend leaving Compressor in its default setting with multiple instances are turned off.

Here’s an article that explains the difference between single-instance and multiple-instance mode and when to use which.

With the exception of compressing for DVD no files were resized and no filters were applied. All source files were copied to the desktop of the computer, and all compressed files were also stored to the desktop of the test computer. No network drives, or direct attached drives, were used for any part of this test.

Audio file sample rates were converted from 48 kHz to 44.1 kHz.

The same compression settings were used between the three video tests. The only difference was in the source media.

The only difference I made between the single-pass and multi-pass compression settings was checking, or unchecking, the multi-pass check box.

FINAL THOUGHTS

I was totally surprised by these findings. Until we start to see applications optimized to take advantage of the power of the Mac Pro, if video compression is your key task, a high-end iMac is your best choice.

As always, let me know what you think.

Bookmark the permalink.

54 Responses to Mac Pro vs. iMac: Video Compression

Newer Comments →

Benji says:

January 3, 2014 at 6:16 pm

You can only test what you got, but I think that it’s premature to recommend the iMac as best choice over the Mac Pro for video encoding. You can only speak for the performance of the 12 core version. All the other enhancements of the Mac Pro can’t make up for clock speed. As an avid FCPX user and all around video guy, I’m surprised Apple sent you the 12-core version as the 8-core is what they said is the “sweet spot” for video work because it’s a better balance of number of cores vs. clock speed (even with ALL cores fired up it can still turbo boost to 3.4ghz).

So these benchmarks are not representative of what a Mac Pro configured for video encoding would achieve. You are using the MacPro that is optimized for cores, not clock speed. The nature of the Xeons is choosing the amount of cores you think you’ll take advantage of and sacrificing CPU speed with the more cores you choose. Someone who renders using the 12 core in Maya, C4D, 3DS Max etc. would see a huge increase in speed vs. the iMac, but ONLY when it came time to render as that task utilises all the cores. Most other daily tasks would be slower than the iMac.

A more equal comparison would be to test ANY of the Mac Pros EXCEPT the 12core as all of them will turbo boost up to 3.9ghz and are more likely to SUSTAIN those speeds due to the increased thermal design power of the Xeons (130w compared to 84w i7 in the iMac). I think the story would be different here and you would see improvements across the board in Compressor compared to the iMac. How much I don’t know, but the 12-core’s top turbo boost speed is 3.5ghz, and that is only if ONE core is being used (it decreases as more cores are fired up).

It’s more likely that in testing all the Mac Pro’s processor options, the end result would be advising the purchase of a Mac Pro with “X” configuration if you want to optimize for video encoding (who knows, perhaps the quad core since it has the highest single threaded clock speed @3.7ghz, but obviously not the 12 core).

I’m also wondering if hardware acceleration takes advantage of multiple cores. You mentioned “disabling instances”, I’m not sure if that is for distributed rendering or it also gives Compressor permission to use multiple cores. If the latter, than the 12 core Mac Pro surely never got to shine.

Reply
- Larry says:
  
  January 3, 2014 at 8:13 pm
  
  Benji:
  
  Good comments, though I disagree with several.
  
  The 8-core sweet spot is balancing cost vs performance. For flat-out top speed in FCP X, the 12-core is superior.
  
  I think these tests point out something I would not have expected: that the Mac Pro would lose ANY speed race to an iMac. CPU clock speed is important, but not all encompassing, otherwise the MacPro would lose by the same percentage in all tests. Some tests is won, others it lost.
  
  During the course of this test, I learned that the Mac Pro does not support hardware acceleration, that clock speed is only part of the equation – with or without turbo boost – and that different codecs yield different results on both systems. It will be very hard for any Mac Pro to beat the hardware acceleration in an iMac.
  
  Also, instances have no relation to cores. Here’s an article that explains this in more detail: http://www.larryjordan.biz/compressor-4-1-a-tip-to-increase-speed/
  
  Larry
  
  Reply
  - Benji says:
    
    January 3, 2014 at 8:54 pm
    
    Thanks Larry, I checked out the Compressor article and it’s very helpful to know the difference between instances and cores. Though interestingly enough I have a 2011 iMac w/ sandy bridge i7 and 32GB RAM and compressor will only show 1 instance available in the drop down box. Not sure what’s going on there :/
    
    I’m still not sure about top FCPX speed being the 12 core. Are you referring to export (which I’m guessing would utilize all the cores and definitely be faster than the other processors)? The rest of the performance increases of FCPX 10.1 on the Mac Pro is due mostly to the FirePro cards, correct? If I had a 6 or 8 core Mac Pro with D700s, do you think I would notice the speed difference while editing (not export) in FCPX compared to the 12 core?
    
    I guess another thing that comes into play is background rendering, which can happen constantly in FCPX (if you choose) and the 12 core would have a lot of extra threads to throw at it, which would contribute to everything feeling a lot snappier than with the other processors.
    
    Reply
    - Larry Jordan says:
      
      January 3, 2014 at 9:15 pm
      
      Benji:
      
      Keep in mind that background rendering only happens when you are not doing something in the foreground. And… Apple is making a big deal that the new Mac Pro is so fast that you can turn background rendering off and let the Mac Pro play everything in real-time without rendering until you get to final export.
      
      Just on a more practical note, if the 12-core were not the best performing Mac Pro, Apple would not be sending it to reviewers – they would be sending the Mac Pro that features the best performance.
      
      Larry
      
      Reply
      - Benji says:
        
        January 4, 2014 at 12:42 am
        
        That’s the thing, Apple isn’t only sending out the 12 core version for benchmarks/reviews. You and Anantech received the 12 core version with 32GB RAM while “The Verge”, “PC Mag”, and “The New York Times” received the 8-core version w/ 64GB RAM.
        
        After reading Anantech’s review it seems the 8-core may be the fastests for my uses (3D rendering, after effects compositing, and some HD video editing). But it also seems that this time with turbo boost in play more than ever and the optimization of software for the FirePro cards up in the air (except for FCPX), it really depends on what you’re using the Mac Pro for as to the speed you’ll get (cores vs. clock speed). If money weren’t an issue, perhaps the 12 core may not necessarily be the configuration of choice as in certain instances the top of line iMac will out perform it due to its lower clock speed.
        
        I think with every other apple device you can check off the biggest numbers all the way down the list on the Apple store configuration page and know you’ll be getting the fastest version, but with the Mac Pro this time around it seems more complicated/confusing.
        
        In thinking about video encoding, I wonder if the FCPX “Share” options that don’t send the timeline to compressor for encoding would leverage the GPU for export and if so how fast this is compared to a similar preset in Compressor. In other words, are the GPUs mainly for live playback or can we use them them to improve export/encoding speeds?
- Tim says:
  
  August 17, 2016 at 11:49 am
  
  I would like to see an update to this. I wonder how much has changed in 2yrs. I have a 12 core I use for editing.
  
  Reply
  - Tori Hoefke says:
    
    August 17, 2016 at 7:34 pm
    
    Tim:
    
    Surprisingly little has changed. First, because the Mac Pro hasn’t been updated in three years. While these tests were done on an 8-core system, your 12-core will be a bit faster, but not by much, because the CPU speed of the 12-core is slower than the CPU speed of the 8-core. Apple Compressor does not use all 12 cores when compressing single files, due to limitations in the H.264 codec. You can improve this by running multiple instances, but that will only really help when compressing multiple files at the same time.
    
    ProRes will compress faster on an 12-core system than an 8-core, but neither will be as fast as an iMac, because the current version of the Mac Pro does not support hardware acceleration, which is supported in the iMac.
    
    In other words, while the specific numbers will be different, the overall trend and results will be the same.
    
    Larry
    
    Reply
Benji says:

January 3, 2014 at 8:03 pm

A little unrelated to this particular post, but while you’ve got the Mac Pro can you see if the iMac works as a Target Display for it? According to the official Target Display documentation the iMac should in theory work as a display for the Mac Pro (requirements are to have a desktop mac with a thunderbolt port and an iMac that supports target display mode), but I can’t find any information if this actually works or not. I’m trying to figure out if I should sell my iMac and purchase a new display for my new Mac Pro or keep the iMac and use it as a display and general purpose stuff while the Mac Pro is rendering.

See here:
http://support.apple.com/kb/HT3924?viewlocale=en_US#4

Reply
- Larry says:
  
  January 3, 2014 at 8:13 pm
  
  Benji:
  
  I’ll see if I can test this.
  
  Larry
  
  Reply
ryan says:

January 3, 2014 at 10:16 pm

In your Mac Pro summary you stated: “If you are doing video compression, the Mac Pro wins hands down over any other system. The hardware acceleration and GPUs will save you months of time.” Now that you’ve had some time to look at it more in-depth vs. an iMac, does that deserve a re-visit?

Reply
- Larry says:
  
  January 4, 2014 at 12:24 am
  
  Ryan:
  
  Good catch. I’ve adjusted the ending of the Mac Pro summary: http://www.larryjordan.biz/the-new-mac-pro-ridiculously-fast/
  
  Larry
  
  Reply
Charles Jones says:

January 6, 2014 at 8:47 am

I think there are more factors to the choice of a iMac over a Mac pro. For example, for the very few of us such as myself who waited for the Mac Pro to update the simple Mac Pro upgrade/replacement is a Perfect Purchase. I waited a patient 5 years, (maybe too long) LOL. Not all of us purchases every mac that comes out and there was about 3 years when they did nothing to the Mac pro guys, that little upgrade was silly to buy since we knew a new one was on its way. SO i say that depending on how old your current MACPRO is that is a HUGE factor on just replacing it with a new one because you are gonna get a Speed boost in the user experience regardless and Not all of us are all in one users we are workstation pro users where we like dual monitors and we want them to match, or maybe I’m just a desk diva and appearance of my set up matters. EITHER WAY All that matters is that if your system is old, really then a new mac pro 6 or eight core will be faster than what you have currently using so it may be a great choice.
Speed is in the eye of the beholder.

Reply
John Lamont says:

January 6, 2014 at 9:32 am

If you consider adding PCI flash to a top-end current iMac it will fly even faster. Not convinced Apple has wrung the most it can out of the workstation class processor and i/o structure. Think Mac Pro late 2014 might be the one to buy with truly new graphics cards, unconstrained PCI busses and well engineered retina support, time will tell. Good catch on hardware h.264 rendering missing from Xenon’s.
A great decision by Apple to make all parts replaceable, will help residuals over time.

Reply
Gregory Minton says:

January 6, 2014 at 10:54 am

A big thanks for this article Larry. I work at a college and have started encoding lots of H.264 video. Last August I asked for a 5k budget line for a better video workstation. I’ve been holding out waiting, and waiting for the new Mac Pro. My dilemma was, with a base model new Mac Pro and a thunderbolt raid storage unit, I didn’t even have enough money left over for a mouse and keyboard! Not to mention No monitor, No SD card reader, No optical drive.

After reading this, I’ve changed the PO to an iMac. Now my 5k is now getting me:

• iMac 3.5GHz Quad-core Intel Core i7, Turbo Boost up to 3.9GHz
• 27” Monitor
• 32GB Ram
• 512GB Flash System Drive
• NVIDIA GeForce GTX 780M 4GB GDDR5
• SDXC Card Slot
• USB Super Drive
• Magic Mouse
• Wireless Keyboard
• PROMISE Pegasus2 R4 8TB (4 by 2TB) Thunderbolt 2 RAID System

If I had a little more money to spend, I would have loved to have joined the Mac Pro club but it just wasn’t realistic for my needs right now.

Reply
- Larry Jordan says:
  
  January 6, 2014 at 1:56 pm
  
  Gregory:
  
  You are welcome… and correct. There are times where the Mac Pro makes sense. There are also times where an iMac is a better decision. That’s one of the reasons I wrote the article.
  
  larry
  
  Reply
Sjoerd de Vries says:

January 6, 2014 at 3:12 pm

Larry and others,

I have in order a 6 core 32 GB 1 TB PCI flash D700.
But I want to now if i could better order a 8 core (btw I can order a 8 c but not a 12c (financial))

I use the machine only for video with FCP-X 10.1 Motion and compressor

Reply
- Larry Jordan says:
  
  January 6, 2014 at 3:21 pm
  
  Sjoerd:
  
  What you ordered should be fine for both FCP X and Motion. The gating factor on performance is your storage system, not the computer.
  
  You would get slightly better performance with the 8-core, but, frankly, I don’t think its worth taking out a second mortgage solely for the processors alone.
  
  You put the emphasis in the right spot – graphics cards and RAM.
  
  Larry
  
  Reply
  - Sjoerd de Vries says:
    
    January 7, 2014 at 5:02 am
    
    Larry,
    
    Perhaps forget to tell you that I have now a 12 core MacPro 2.66 GHz 16 GB Radeon 5870
    
    Reply
Sjoerd de Vries says:

January 6, 2014 at 3:52 pm

Thanks Larry, I did make a good choice than

Reply
David says:

January 7, 2014 at 8:33 am

Good article, Larry, really interesting stuff. I didn’t know that the Xeon’s didn’t have the hardware acceleration, so that makes me feel even better about the 2013 iMac that I’m getting this week!

Is there any chance you could do similar tests with Adobe Media Encoder? I don’t think GPU acceleration is disabled with multipass encoding, but I think GPU acceleration also works differently in AME vs Compressor. When you send a sequence from Pr to AME then all the GPU accelerated features of Premiere are utilized. However, if you do what you did, and just drop a file into AME then only certain things are GPU accelerated like scaling, deinterlacing, some other things (details here: http://helpx.adobe.com/media-encoder/using/whats-new-media-encoder-7-1.html#gpu-acceleration). So I would be really interested to know how AME stacks up vs Compressor here, because when I did a straight test of the AME vs Compressor (without changing any compressor preferences), AME was much faster; but I’m almost positive that multipass encoding was selected for both programs.

Reply
- Larry Jordan says:
  
  January 7, 2014 at 11:30 am
  
  David:
  
  Interesting idea – let me see what i can do.
  
  Larry
  
  Reply
  - John Putch says:
    
    January 7, 2014 at 10:58 pm
    
    Et al-
    
    6 TB ports and the size factor, plus the access make this pig a wonder. I plan to use mine to color my film using speed-grade and finish up post on ROUTE 30, THREE!.
    
    Thanks for all the reporting on the dream machine, Larry.
    
    JP
    
    Reply
David Arbor says:

January 8, 2014 at 5:34 am

@John Putch,

John, are you planning to grade your project in SpeedGrade through Direct Link or Native Mode? If you’ll be using Direct Link then GPU acceleration is not enabled with OpenCL on the Mac. See bullet 2 for more details here: http://blogs.adobe.com/movingcolors/2013/12/12/speedgrade-cc-7-2-update/

The Xeons are fast, but I don’t think they will be fast enough for a perfectly smooth workflow without GPU acceleration. On the other hand, if you’re grading say, a finished ProRes file, then you can do that natively with scene detection and happily have GPU acceleration.

Reply
- putchfilms says:
  
  January 8, 2014 at 3:22 pm
  
  HI David,
  
  Thanks for the info. But I will be grading ProRes HQ EDL. Full rez of the movie itself. So I guess I won’t have any trouble.
  
  John
  
  Reply