Putting Apple’s iMac Pro Through the Paces

At the end of December, Apple made good on the release of the new iMac Pro and started selling and shipping the new workstations. While this could be characterized as a stop-gap effort until the next generation of Mac Pro is produced, that doesn’t detract from the usefulness and power of this design in its own right. After all, the iMac line is the direct descendant in spirit and design of the original Macintosh. Underneath the sexy, all-in-one, space grey enclosure, the iMac Pro offers serious workstation performance.

I work mostly these days with a production company that produces and posts commercials, corporate videos, and entertainment programming. Our editing set-up consists of seven workstations, plus an auxiliary machine connected to a common QNAP shared storage network. These edit stations consisted of a mix of old and new Mac Pros and iMacs (connected via 10GigE), with a Mac Mini for the auxiliary (1GigE). It was time to upgrade the oldest machines, which led us to consider the iMac Pros. The company picked up three of them – replacing two Mac Pro towers and an older iMac. The new configuration is a mix of three, one-year-old Retina 5K iMacs (late 2015 model), a 2013 “trash can” Mac Pro, and three 2017 iMac Pros.

There are plenty of videos and articles on the web about how these machines perform; but, the testers often use artificial benchmarks or only Final Cut Pro X. This shop has a mix of NLEs (Adobe, Apple, Avid, Blackmagic Design), but our primary tool is Adobe Premiere Pro CC 2018. This gave me a chance to compare how these machines stacked up against each other in the kind of work we actually do. This comparison isn’t truly apples-to-apples, since the specs of the three different products are somewhat different from each other. Nevertheless, I feel that it’s a valid real-world assessment of the iMac Pros in a typical, modern post environment.

Why buy iMac Pros at all?

The question to address is why should someone purchase these machines? Let me say right off the bat, that if your main focus is 3D animation or heavy compositing using After Effects or other applications – and speed and performance are the most important factor – then don’t buy an Apple computer. Period. There are plenty of examples of Dell and HP workstations, along with high-end gaming PCs, that outperform any of the Macs. This is largely due to the availability of advanced NVidia GPUs for the PC, which simply aren’t an option for current Macs.

On the other hand, if you need a machine that’s solid and robust across a wide range of postproduction tasks – and you prefer the Mac operating ecosystem – then the iMac Pros are a good choice. Yes, the machine is pricy and you can buy cheaper gaming PCs and DIY workstations, but if you stick to the name brands, like Dell and HP, then the iMac Pros are competitively priced. In our case, a shift to PC would have also meant changing out all of the machines and not just three – therefore, even more expensive.

Naturally, the next thing is to compare price against the current 5K iMacs and 2013 Mac Pros. Apple’s base configuration of the iMac Pro uses an 8-core 3.2GHz Xeon W CPU, 32GB RAM, 1TB SSD, and the Radeon Pro Vega 56 GPU (8GB memory) for $4,999. A comparably configured 2013 Mac Pro is $5,207 (with mouse and keyboard), but no display. Of course, it also has the dual D-700 GPUs. The 5K iMac in a similar configuration is $3,729. Note that we require 10GigE connectivity, which is built into the iMac Pros. Therefore, in a direct comparison, you would need to bump up the iMac and Mac Pro prices by about $500 for a Thunderbolt2-to-10GigE converter.

Comparing these numbers for similar machines, you’d spend more for the Mac Pro and less for the iMac. Yet, the iMac Pro uses newer processors and faster RAM, so it could be argued that it’s already better out of the gate in the base configuration than Apple’s former top-of-the-line product. It has more horsepower than the tricked-out iMac, so then it becomes a question of whether the cost difference is important to you for what you are getting.

Build quality

Needless to say, Apple has a focus on the quality and fit-and-finish of its products. The iMac Pro is no exception. Except for the space grey color, it looks like the regular 27” iMacs and just as nicely built. However, let me quibble a bit with a few things. First, the edges of the case and foot tend to be a bit sharp. It’s not a huge issue, but compared with an iPhone, iPad, or 2013 Mac Pro, the edges just not as smooth and rounded. Secondly, you get a wireless mouse and extended keyboard. Both have to be plugged in to charge. In the case of the mouse, the cable plugs in at the bottom, rendering it useless during charging. Truly a bad design. The wireless keyboard is the newer, flatter style, so you lose two USB ports that were on the previous plug-in extended keyboard. Personally, I prefer the features and feel of the previous keyboard, not to mention any scroll wheel mouse over the Magic Mouse. Of course, those are strictly items of personal taste.

With the iMac Pro, Apple is transitioning its workstations to Thunderbolt 3, using USB-C connectors. Previous Thunderbolt 2 ports have been problematic, because the cables easily disconnect. In fact, on our existing iMacs, it’s very easy to disconnect the Thunderbolt 2 cable that connects us to the shared storage network, simply by moving the iMac around to get to the ports on the back. The USB-C connectors feel more snug, so hopefully we will find that to be an improvement. If you need to get to the back of the iMac or iMac Pro frequently, in order to plug in drives, dongles, etc., then I would highly recommend one of the docks from CalDigit or OWC as a valuable accessory.

5K screen

Apple spends a lot of marketing hype on promoting their 5K Retina screens. The 27” screens have a raw pixel resolution of 5120×2880 pixels, but that’s not what you see in terms of image and user interface dimensions. To start with, the 5K iMacs and iMac Pros use the same screen resolution and the default display setting (middle scaled option) is 2560×1440 pixels. The top choice is 3200×1800. Of course, if you use that setting, everything becomes extremely small on screen.  Conversely, our 2013 Mac Pro is connected to a 27” Apple LED Cinema Display (non Retina). It’s top scaled resolution is also 2560×1440 pixels. Therefore, at the most useable settings, all of our workstations are set to the same resolution. Even if you scale the resolution up (images and UI get smaller), you are going to end up adjusting the size of the application interface and viewer window. While you might see different viewer size percentage numbers between the machines, the effective size on screen will be the same.

Retina is Apple’s marketing name for high pixel density. This is the equivalent of DPI (dots per inch) in print resolutions. According to a Macworld article, iPhones from 4 to 5s had a pixel density of 326ppi (pixels per inch), while iMacs have 218ppi. Apple converts a device’s display to Retina by doubling the horizontal and vertical pixel count. More pixels are applied to any given area on the screen, resulting in smoother text, smoother diagonal lines, and so on. That’s assuming an application’s interface is optimized for it. At the distance that the editors sit from a 27” display, there is simply little or no difference between the look of the 27” LED display and the 27” iMac Retina screens.

Upgradeability

Future-proofing and upgrades are the biggest negatives thrown at all-in-ones, particularly the iMac Pros. While the user can upgrade RAM in the standard iMacs, that’s not the case with iMac Pros. You can upgrade RAM in the future, but that must be done at a service facility, such as the Apple Store’s Genius service. This means that in three years, when you want the latest, greatest CPU, GPU, storage, etc., you won’t be able to swap out components. But is this really an issue? I’m sure Apple has user research numbers to justify their decisions. Plus, the thermal design of the iMac would make user upgrades difficult, unlike older mac Pro towers.

In my own experience on personal machines, as well as clients’ machines that I’ve helped maintain, I have upgraded storage, GPU cards, and RAM, but never the CPU. Although I do know others who have upgraded Xeon models on their Mac Pro towers. Part of the dichotomy is buying what you can afford now and upgrading later, versus stretching a bit up front and then not needing to upgrade later. My gut feeling is that Apple is pushing the latter approach.

If I tally up the cost of the upgrades that I’ve made after about three years, I would already be part of the way towards a newer, better machine anyway. Plus, if you are cutting HD and even 4K today, then just about any advanced machine will do the trick, making it less likely that you’ll need to do that upgrade within the foreseeable life of the machine. An argument can be made for either approach, but I really think that the vast majority of users – even professional users – never actually upgrade any of the internal hardware from that of the configuration as originally purchased.

Performance testing

We ultimately purchased machines that were the 10-core bump-up from the base configuration, feeling that this is the sweet spot (and is currently available) within the iMac Pro product line.

The new machine specs within the facility now look like this:

2013 Mac Pro – 3GHz 8-core Xeon/64GB RAM/dual D-500 GPUs/1TB SSD (Sierra)

2015 iMac – 4GHz 4-core Core i7/32GB RAM/AMD R9/3TB Fusion drive (Sierra)

2017 iMac Pro – 3GHz 10-core Xeon W/64GB RAM/Radeon Vega 64/1TB SSD (High Sierra)

As you can see, the tech specs of the new iMac Pros more closely match the 2013 Mac Pro than the year-old 5K iMacs. Of course, it’s not a perfect match for optimal benchmark testing, but close enough for a good read on how well the iMac Pro delivers in a real working environment.

Test 1 – BruceX

The BruceX test uses a 5K Final Cut Pro X timeline made up only of built-in titles and generators. The timeline is then rendered out to a ProRes file. This tests the pure application without any media and codec variables. It’s a bit of an artificial test and only applicable to FCPX performance, but still useful. The faster the export time, the better. (I have bolded the best results.)

2013 Mac Pro – 26.8 sec.

2015 iMac – 28.3 sec.

2017 iMac Pro – 14.4 sec.

Test 2 – media encoding

In my next test, I took a 4½-minute-long 1080p ProRes file and rendered it to a 4K/UHD (3840×2160) H.264 (1-pass CBR 20Mbps) file. Not only was it being encoded, but also scaled up to 4K in this process. I rendered from and to the desktop, to eliminate any variables from the QNAP system. Finally, I conducted the test using both Adobe Media Encoder (using OpenCL processing) and Apple Compressor.

Two noteworthy issues. The Compressor test was surprisingly slow on the Mac Pro. (I actually ran the Compressor test twice, just to be certain about the slowness of the Mac Pro.) The AME version kicked in the fans on the iMac.

Adobe Media Encoder

2013 Mac Pro – 6:13 min.

2015 iMac – 7:14 min.

2017 iMac Pro – 4:48 min.

 Compressor

2013 Mac Pro – 11:02 min.

2015 iMac – 2:20 min.

2017 iMac Pro – 2:19 min.

 Test 3 – editing timeline playback – multi-layered sequence

This was a difficult test designed to break during unrendered playback. The 40-second 1080p/23.98 sequence include six layers of resized 4K source media.

Layer 1 – DJI clips with dissolves between the clips

Layers 2-5 – 2D PIP ARRI Alexa clips (no LUTs); layer 5 had a Gaussian blur effect added

Layer 6 – native REDCODE RAW with minor color correction

The sequence was created in both Final Cut Pro X and Premiere Pro. Playback was tested with the media located on the QNAP volumes, as well as from the desktop (this should provide the best possible playback).

Playing back this sequence in Final Cut Pro X from the QNAP resulted is the video output largely choking on all of the machines. Playing it back in Premiere Pro from the QNAP was slightly better than in FCPX, with the 2017 iMac Pro performing best of all. It played, but was still choppy.

When I tested playback from the desktop, all three machines performed reasonably well using both Final Cut Pro X (“best performance”) and Premiere Pro (“1/2 resolution”). There were some frames dropped, although the iMac Pro played back more smoothly than the other two. In fact, in Premiere Pro, I was able to set the sequence to “full resolution” and get visually smooth playback, although the indicator light still noted dropped frames. Typically, as each staggered layer kicked in, performance tended to hiccup.

Test 4 – editing timeline playback – single-layer sequence

 This was a simpler test using a standard workflow. The 30-second 1080p/23.98 sequence included three Alexa clips (no LUTs) with dissolves between the clips. Each source file was 4K/UHD and had a “punch-in” and reposition within the HD frame. Each also included a slight, basic color correction. Playback was tested in Final Cut Pro X and Premiere Pro, as well as from the QNAP system and the desktop. Quality settings were increased to “best quality” in FCPX and “full resolution” in Premiere Pro.

My complex timeline in Test 3 appeared to perform better in Premiere Pro. In Test 4, the edge was with Final Cut Pro X. No frames were dropped with any of the three machines playing back either from the QNAP or the desktop, when testing in FCPX. In Premiere Pro, the 2017 iMac Pro was solid in both situations. The 2015 iMac was mostly smooth at “full” and completely smooth at “1/2”. Unfortunately, the 2013 Mac Pro seemed to be the worst of the three, dropping frames even at “1/2 resolution” at each dissolve within the timeline.

Test 5 – timeline renders (multi-layered sequence)

In this test, I took the complex sequence from Test 3 and exported it to a ProRes master file. I used the QNAP-connected versions of the Premiere Pro and Final Cut Pro X timelines and rendered the exports to the desktop. In FCPX, I used its default Share function. In Premiere Pro, I queued the export to Adobe Media Encoder set to process in OpenCL. This was one of the few tests in which the 2013 Mac Pro put in a faster time, although the iMac Pro was very close.

Rendering to ProRes – Premiere Pro (via Adobe Media Encoder)

2013 Mac Pro – 1:29 min.

2015 iMac – 2:29 min.

2017 iMac Pro – 1:45 min.

Rendering to ProRes – Final Cut Pro X

2013 Mac Pro – 1:21 min.

2015 iMac – 2:29 min.

2017 iMac Pro – 1:22 min.

Test 6 – Adobe After Effects – rendering composition

My final test was to see how well the iMac Pro performed in rendering out compositions from After Effects. This was a 1080p/23.98 15-second composition. The bottom layer was a JPEG still with a Color Finesse correction. On top of that were five 1080p ProResLT video clips that had been slomo’ed to fill the composition length. Each was scaled, cropped, and repositioned. Each was beveled with a layer style and had a stylized effect added to it. The topmost layer was a camera layer with all other layers set to 3D, so the clips could be repositioned in z-space. Using the camera, I added a slight rotation/perspective change over the life of the composition.

Rendering to ProRes – After Effects

2013 Mac Pro – 2:37 min.

2015 iMac – 2:15 min.

2017 iMac Pro – 2:03 min.

Conclusion

After all of this testing, one is left with the answer “it depends”. The 2013 Mac Pro has two GPUs, but not every application takes advantage of that. Some apps tax all the available cores, so more, but slower, cores are better. Others go for the maximum speed on fewer cores. All things considered, the iMac Pro performed at the top of these three machines. It was either the best or close/equal to the best.

There is no way to really quantify actual editing playback performance and resolution by any numerical factor. However, it is interesting to look at the aggregate of the six tests that could be quantified. When you compare the cumulative totals of just the iMac Pro and the iMac, the Pro came out 48% faster. Compared to the 2013 Mac Pro, it was 85% faster. The iMac Pro’s performance against the totals of the slowest machines (either iMac or Mac Pro depending on the test), showed it being a whopping 113% faster – more than twice as fast. But it only bested the fastest set by 20%. Naturally, such comparisons are more curiosity than anything else. Some of these numbers will be meaningful and others won’t, depending on the apps used and a user’s storage situation.

I will say that installing these three machines was the easiest I’ve ever done, including connecting them to the 10GigE storage network. The majority of our apps come from Adobe Create Cloud, the Mac App Store, or FxFactory (for plug-ins). Except for a few other installers, there was largely no need to track down installers, activation information, etc. for a zillion small apps and plug-ins. This made it a breeze and is certainly part of the attraction of the Mac ecosystem. The iMac Pro’s all-in-one design limits the required peripherals, which also contributes to a faster installation. Naturally, I can’t tell anyone if this is the right machine for them, but so far, the investment does look like the correct choice for this shop’s needs.

(Updated 6/22/18)

Here are two additional impressions by working editors: Thomas Grove Carter and Ben Balser. Also a very comprehensive review from AppleInsider.

©2018 Oliver Peters

Advertisements