Apple Compressor: Faster Video Compression

Over the last several weeks, I’ve been testing Apple Compressor 4.2 (the current version) for video compression speed. During this time, I’ve learned:

• The iMac is faster at video compression than a Mac Pro, when compressing for the web
• The latest version of Compressor is the fastest yet
• There is a lot of variation in compression speed between different software

However, there’s a feature in Compressor that I haven’t tested: running multiple instances of the software. This allows Compressor to “clone” itself to, theoretically, process files faster. The iMac supports one additional instance, while the Mac Pro supports up to three instances. That’s what I decided to test this week.

To help you catch up, here are other recent articles I’ve written on video compression:

• This article compares the compression speed between Compressor 4.1.2 and Compressor 4.2.
• This article examines image quality in the latest version of Compressor 4.2 when compressing files using the H.264 codec.
• This article compares the speed between a new Mac Pro, a 21″ iMac and a 15″ MacBook Pro using Apple Compressor 4.2.
• This article compares the compression speed of six different video compression software

EXECUTIVE SUMMARY

Using four of the default YouTube compression settings included with Compressor 4.2 for this test, I learned several things:

1. If you are compressing using an iMac, never use multiple instances.
2. If you are compressing with a MacPro, always use multiple instances.
3. An iMac is faster than a MacPro when compressing video for the web. Period.
4. For fastest speed on a Mac Pro, apply multiple settings to a single clip, rather than duplicate that clip multiple times and apply one setting to each clip iteration. (How you apply settings does not make any difference to an iMac.)

The latest version of Compressor (4.2) hooks into the hardware-acceleration offered by both iMac and MacBook Pro computers to accelerate H.264 compression. (H.264 is the current standard delivery codec for all web-based video files.) The MacPro does not currently support hardware-accelerated video compression; a fact made very clear by these tests.

(Click here to download a PDF with details of all my test findings.)

TEST SYSTEMS

Late 2013 Mac Pro
Yosemite 10.10.3
3.0 GHz, 8-core Xeon processor
32 GB RAM
AMD D700 GPU

Late 2013 21″ iMac
Yosemite 10.10.3
3.1 GHz Intel Core i7
16 GB RAM
NVIDA GeForce GT750M GPU with 1024 VRAM

NOTES

• Past tests also included a 15″ MacBook Pro. As it uses the same hardware acceleration as the iMac, compression times will be similar.
• A 27″ iMac will have speeds comparable to the 21″ iMac, based upon CPU speed.
• The brand-new MacBook Pro will have speeds similar to the iMac. The new laptop has blazing-fast hard disk speeds, which are only of limited use in video compression.
• This test used two different codecs: XDCAM and ProRes 422 HQ. Because the durations of the test videos are different, we can’t compare the compression times of the different codecs.

ANALYSIS: CREATING THE BATCH

For this test, I compared the compression speed of two different movies, each using a different codec. What I did was apply four different default YouTube settings to each of two different videos (the same four settings were used for all tests), then timed how long each set of videos took to compress. (Timings were reported by the application and by reading Compressor’s log files.)

I tested this four different ways:

• Apply all four settings to a single clip; which resulted in two clips in the Batch. Time compression with multiple instances turned off, then time it again with multiple instances turned on.
• Import each clip four times and apply only one setting to each iteration of the clip; which resulted in eight clips in the Batch. Time compression with multiple instances turned off, then time it again with multiple instances turned on.
• Whenever multiple instances were turned on, I used the maximum number supported by the application: 1 for the iMac and 3 for the Mac Pro.

I wanted to see if the number of files in the batch changed the speed of compression when multiple instances were enabled.

The short answer is that it didn’t make any difference to an iMac whether this was a single clip with multiple settings or a multiple clips with one setting; compression times were within a few seconds of each other.  However, applying multiple settings to a single clip made the Mac Pro compress files about 25% faster. As the table above illustrates, the way you apply compression settings to clips significantly altered the compression speed.

ANALYSIS: MULTIPLE INSTANCES

The MacPro does not support hardware acceleration for video compression. So, while it has massive horsepower, it compresses video totally in software. When it comes to video compression speed, this is a serious liability.

Using all default YouTube compression settings:

• The iMac was almost three times faster than the MacPro when compressing XDCAM footage with multiple instances turned off.
• The iMac was more than twice as fast as the MacPro when compressing ProRes 422 HQ footage with multiple instances turned off.
• The difference narrowed when multiple instances were enabled for the Mac Pro, but at no time was the MacPro even close to the iMac in compression speed.
• Enabling multiple instances for the iMac slows compression to a crawl, making video compression take 3 times longer than when multiple instances are not enabled.

COMPRESSION OVERLAP

In the PDF, you’ll see a line titled: “Overlapping Compression Savings.” This is the amount of time saved by having the computer compress multiple files at the same time (which is what multiple instances does).

I’m not yet sure how meaningful this number is. For example, on the iMac, overall performance was so slow that any time savings caused by compressing multiple files at once were completely lost by the overall slowness of the system caused by compressing multiple files at once.

On the Mac Pro, however, overall compression speed was faster due to the multiple iterations, with a few additional minutes saved due to having multiple files compress simultaneously.

NOTE: I am somewhat suspicious of the 26 minute time saving in the last column of the Mac Pro. I think I messed up a calculation. The rest of the numbers are log readings. Whether or not this number is in error, it doesn’t change any of my conclusions.

SUMMARY

• If video compression speed is important to you, buy an iMac. Faster CPUs will yield better results.
• Storing source files an internal SSD or fast external RAID improves compression speeds (I learned this in earlier tests).
• Never turn on multiple instances when compressing using an iMac.
• If you are using the MacPro, enable multiple instances. However, multiple instances will slow down Final Cut Pro X if you are also editing on the same system. For fastest results, enable multiple instances and don’t do any other work when using a MacPro for video compression.
• If you need to do editing and compression on a Mac Pro, don’t enable multiple instances.
• Remember that compression is a background activity. This means that if you are doing CPU intensive activities in the foreground, compression will take longer.

This has been an interesting series of tests and I’ve learned a lot. As with all things, your specific situation may yield different results. Always test your workflow to see what works best for you.

(Click here to download a PDF with details of all my test findings.)