DPR

Apple M1 Ultra vs M1 Max: Head-to-head shootout with photo and video editing apps

0
These two Mac Studio computers look identical from the outside, but one of them has twice the CPU cores, twice the GPU cores, and twice the RAM…

All product photography by DL Cade

The release of Apple’s new Mac Studio wasn’t just exciting because of the form factor, or the fact that creatives have been asking Apple for a mid-range desktop tower for years. It was also exciting because it came hand-in-hand with the release of the M1 Ultra: the most powerful Apple Silicon system on a chip (SOC) yet.

The M1 Ultra is essentially two M1 Max chips fused together using a new “ultra-fusion” interconnect technology, delivering twice the CPU cores, twice the GPU cores, twice the Neural Engine cores, twice the dedicated media engines, and twice the RAM with twice the bandwidth. That’s up to 20 CPU cores (16 performance, four efficiency), 64 GPU cores, 32 neural engine cores, four video encoding engines, four ProRes encode and decode engines, and up to 128GB of LPDDR5 unified memory with 800GB/s of memory bandwidth.

The M1 Ultra is essentially two M1 Max chips fused together using a new “ultra-fusion” interconnect technology.

And since the ultra-fusion interconnect presents these two M1 Max chips as a single, unified SOC in software, there’s good reason to hope that the M1 Ultra can actually offer twice the performance in apps that need, and are optimized to use, all of these resources.

The M1 Ultra is essentially two M1 Max chips stacked end to end. Image: Apple

This is all amazing in theory—truly an engineering marvel, no matter how much you resent Apple for locking it inside a box with no upgradable components—but how does the M1 Ultra actually compare to the M1 Max in real-world photo and video-editing tasks? Are any of the apps currently used by most creatives actually going to benefit from all this power, or are you better off saving the $3,000 it costs to upgrade to Apple’s most powerful Mac Studio configuration?

Today, we find out.

Shortly after we published our Mac Studio review using a configuration with M1 Max, Apple made good on their promise and sent us another unit with M1 Ultra inside. It is, quite literally, twice as powerful as our M1 Max Mac Studio in every single way, and we ran both of these computers through all of the same benchmarks to see just how much of an impact all that hardware can really have on your photo and video editing workflow.

Meet the contestants

Today, we’re essentially comparing the fully loaded M1 Max version of the Mac Studio against the fully loaded M1 Ultra version. The two machines we have in the office come with different amounts of storage (2TB on the M1 Max and 4TB on the M1 Ultra), but if we control for storage, these are the specs and price points of the computers that we’ll be testing:

M1 Ultra Mac StudioM1 Max Mac Studio
CPU

M1 Ultra

20-core CPU

M1 Max

10-core CPU

GPU

M1 Ultra

64-core GPU

M1 Max

32-core GPU

RAM128GB Unified Memory64GB Unified Memory
Storage2TB Integrated NVMe SSD2TB Integrated NVMe SSD
Price$6,200$3,200

In reality, the extra 2TB of storage included in the M1 Ultra version of the Mac Studio that Apple sent over costs an additional $600, bringing the total price of this loaner to $6,800, but as far as we can tell there is no performance difference between the storage included in our 2TB M1 Max Mac Studio vs the 4TB M1 Ultra Mac Studio.

In other words: any performance difference you see (or don’t see) below is the result of $3,000 worth of relevant upgrades.

Back to top


Performance test

The M1 Ultra is literally twice as powerful as the M1 Max, but not all apps are able to turn that additional hardware into real-world performance.

To test the performance of the M1 Ultra against the M1 Max in the most controlled way possible, we ran all of the benchmarks below while each computer was connected to the same 5K Studio Display using the same scaling options in the Display settings (default for the display). This might seem like an odd thing to mention, but we’ve seen performance in certain applications (i.e. Lightroom Classic) measurably change as you change the output resolution and even the scaling of the display that you’re using.

Beyond the test setup, we ran the same benchmarks we always do, so these results are directly comparable against any other PC or Mac we’ve tested and reviewed previously on DPR. We tested Lightroom Classic, Capture One Pro, Photoshop, Premiere Pro, and Final Cut Pro.

Each individual test was run a minimum of three consecutive times to ensure that we’ve reached thermal equilibrium, and the average of the last three runs is what we’re reporting below.

Adobe Lightroom Classic

To test Lightroom performance, we use 100 copies of the studio scene photo from four different cameras: the 20MP Canon EOS R6, the 47MP Nikon Z7 II, the 61MP Sony a7R IV, and the 100MP Fujifilm GFX 100. Each batch of 100 RAW files is first imported and 1:1 previews are created (that’s our “Import” benchmark), after which we apply a custom preset full of global edits and export those same files as full-resolution, 100% quality JPEGs (that’s our “Export” benchmark).

The cache is cleared and the program is restarted between each test.

Generally speaking, Lightroom import and preview generation is a test of CPU performance while export is a test of CPU + RAM. In both tests, you would hope that the M1 Ultra with 2x the cores and 2x the RAM would be about… well… 2x faster. Incredibly, you’d be right.

Canon EOS R6 ImportNikon Z7 II ImportSony a7R IV ImportFujifilm GFX 100 Import
M1 Max1:262:192:246:02
M1 Ultra00:481:081:132:29

This is a much bigger leap in performance than we were expecting […] the M1 Ultra is legitimately twice as fast as the M1 Max in Lightroom Classic.

This is a much bigger leap in performance than we were expecting given that Lightroom is notorious for being poorly optimized. But you have to give Adobe credit, Lightroom Classic is taking full advantage of the extra CPU cores and extra RAM available in the M1 Ultra.

This pattern continues to play out in exports:

Canon EOS R6 ExportNikon Z7 II ExportSony a7R IV ExportFujifilm GFX 100 Export
M1 Max2:285:186:4511:16
M1 Ultra1:202:563:426:04

This is legitimately a 2x improvement in performance, and it had us very hopeful for all the rest of the benchmarks. Could the M1 Ultra possibly be 2x faster across the board? Unfortunately, the answer is no. From here on out, the performance improvements are a bit smaller, and closer to what you would expect from a generational upgrade.

Capture One Pro 22

We use the same benchmarks in Capture One Pro that we use in Lightroom Classic. We import each batch of 100 RAW files, generate previews at default resolution (there is no 1:1 option), apply the same custom preset/style, and then export each set as full-sized 100% quality JPEGs.

Generally, there’s a big performance difference between Lightroom and Capture One because C1 uses GPU acceleration to speed things up. Export, in particular, is a ton faster thanks to GPU acceleration, and since the M1 Ultra has 2x the GPU cores as the M1 Max, we were hoping for a repeat of the Lightroom Classic performance gains. Unfortunately, whatever Capture One is doing to accelerate their app, they’ve not yet optimized to take advantage of the M1 Ultra’s extra cores.

The latest version of Capture One Pro 22 (15.2) is 2x faster at import and preview generation compared to the previous release, but there’s almost no difference between the performance you’ll get with the M1 Max and the M1 Ultra.

The latest version of Capture One Pro 22 (15.2) is 2x faster at import and preview generation compared to the previous release, but there is almost no difference between the performance you’ll get with the M1 Max and the M1 Ultra. At import, the two computers are practically identical:

Canon EOS R6 ImportNikon Z7 II ImportSony a7R IV ImportFujifilm GFX 100 Import
M1 Max00:2600:3300:3800:55
M1 Ultra00:2500:3200:3700:52

At export, things are a bit faster. The M1 Ultra puts up export times that are up to 21% faster than the M1 Max. That’s certainly better than the meager difference we saw in our import tests, but it’s still nowhere near Adobe Lightroom Classic, which is a full 100% faster on M1 Ultra than M1 Max.

Canon EOS R6 ExportNikon Z7 II ExportSony a7R IV ExportFujifilm GFX 100 Export
M1 Max00:592:062:344:29
M1 Ultra00:491:502:093:42

Photoshop

Moving on to Photoshop, we ran Puget Systems‘ PugetBench for Photoshop benchmark to see if the extra RAM, GPU cores, and CPU cores will speed up any heavy-duty post-processing you might be doing. One note: the scores below come from Version 0.8 of PugetBench for Photoshop, because this was the last version to include the PhotoMerge test and because it runs as a script, not a plugin, so it’s fully compatible with Apple Silicon Macs.

While the M1 Ultra Mac Studio does unseat the M1 Max version as the highest scoring computer we’ve ever tested in PugetBench, it only beats the M1 Max by about 1.2% overall.

We had high hopes on this one, since the M1 Max Mac Studio put up some pretty impressive numbers in our full review. But while the M1 Ultra does unseat the M1 Max as the highest scoring computer we’ve ever tested in PugetBench, it only beats the M1 Max by a thin margin, and actually scores lower in a couple of the category scores:

OverallGeneralGPUFilterPhotoMerge
M1 Max1287.2129.2117.5111.5162.2
M1 Ultra1303.0132.6117.2110.1166.1

Premiere Pro

Like Capture One Pro, Premiere Pro typically does a great job of using the GPU horsepower at its disposal to speed up tasks like rendering, applying certain video effects, and transcoding/exporting videos into various formats. To test this performance, we use a 4K project based on 10-bit 8K ProRes 4:2:0 footage from the Sony a1.

This footage is arranged in a timeline full of effects, LUTs, and titles, which we then render and export in a couple of different formats. To wrap things up, we also see how long it takes to apply Warp Stabilize to a wobbly 15-second clip from the same shoot, since this is technically one of Adobe Premiere’s GPU accelerated effects.

Here’s the video that we’re using, in case you’re curious:

The story in Adobe Premiere Pro is similar to Photoshop: there is a small performance difference, but nothing drastic.

The story in Premiere is similar to Photoshop: there is a small performance difference, but nothing drastic. The M1 Ultra is about 9% faster at rendering and 5% faster at exporting H.264 and HEVC files.

Render AllExport H.264Export HEVCWarp Stabilize
M1 Max1:471:411:402:13
M1 Ultra1:381:321:352:07

Given how much a heavy-duty NVIDIA GPU can speed things up when using Premiere Pro on the PC, I’m tempted to believe that this is an optimization issue. For now, we’ll keep our fingers crossed that an upcoming update to Premiere Pro will better use the extra cores and media engines that the M1 Ultra brings to the table.

Final Cut Pro

Finally, if there’s one piece of software that you’d expect to benefit hugely from Apple’s most powerful processor, it would be Apple’s own video editor Final Cut Pro. We use the same exact video and source footage mentioned above to test Final Cut Pro performance, only this time we render the footage, export a master file, export H.264, export HEVC, and test Final Cut Stabilize instead of Warp Stabilize.

Note: For these tests, Apple provided an unreleased version of Final Cut Pro (10.6.2) that has explicitly been optimized to take advantage of the extra horsepower in the M1 Max and M1 Ultra. Apple tells us this version will be out for everyone sometime in April, but they sent over an early build so we could test performance on software that’s actually tuned to take advantage of the latest hardware.

Unfortunately, even using an optimized version of Final Cut, the M1 Ultra wasn’t able to provide the 2x performance bump we were hoping to see.

Unfortunately, even using an optimized version of Final Cut, the M1 Ultra wasn’t able to provide the 2x performance bump we were hoping to see. It does provide a much bigger boost than Premiere Pro—rendering was 26% faster, and exports were between 7% and 8% faster on the M1 Ultra compared to the M1 Max—but it’s still a bit slower than Premiere when it comes to both rendering and transcoding this particular project to H.264.

Render AllExport Master FileExport H.264Export HEVCFinal Cut Stabilize
M1 Max2:4000:432:201:0500:28
M1 Ultra2:0700:372:111:0000:28

Back to top


Performance Takeaways

As we ran these benchmarks and saw the range of different results that were coming in, the same question came up again and again: is this an issue of optimization, or utilization? In other words: is the problem that Capture One Pro, Adobe Premiere Pro, and Adobe Photoshop aren’t yet optimized to take advantage of the new hardware, or are our benchmarks simply not “beefy” enough to put the M1 Ultra to good use?

Lightroom Classic is already tuned to take advantage of multiple cores whenever they’re available, and the M1 Max with its 10-core CPU and 64GB of RAM obviously hadn’t hit the ceiling of what Adobe is capable of using. But apps like Capture One, Photoshop, Premiere Pro, and even Apple’s own Final Cut Pro either need further optimization, or we may need to throw a lot more work at them before we see a 2x performance difference between the Max and the Ultra.

Capture One, for example may benefit from further optimization. The latest update made import and preview generation about 2x faster on both the M1 Max and M1 Ultra, but even if you throw a ton of high-resolution photos at the program, there’s barely any difference in performance between these two machines.

Out of curiosity, we re-ran our import and export benchmarks using 1,000 Fujifilm GFX 100 RAW files (over 210GB in total) instead of the usual 100 files. The results: import was 4% faster, while export was 18% faster. It doesn’t appear that throwing more work at the program makes much of a difference:

In the end, the creators who will benefit most from an upgrade from the M1 Max to M1 Ultra are photographers who rely heavily on Lightroom Classic.

Final Cut Pro, on the other hand, is apparently already optimized for both M1 Max and M1 Ultra. In that case, the performance gains we saw in the benchmarks above could potentially increase as you throw larger projects at the program. Unfortunately, we didn’t have anything bigger to test in FCP—the Sodo Moto library with its 8K Sony a1 footage is the most graphically intensive project we have lying around—but the performance of the M1 Ultra over the M1 Max may improve as project size and complexity increases.

In the end, it appears that the creators who will benefit most from an upgrade from the M1 Max to M1 Ultra are photographers who rely heavily on Lightroom Classic. This is not the result we expected, but it’s the result we got. If you’re still using Adobe’s classic DAM and RAW photo editor, you’ll cut your wait time in half during long imports and exports. But if you’re using any of the other creative programs we tested above, you may want to think twice before upgrading.

Back to top


Conclusion – 2x price doesn’t mean 2x performance

The M1 Ultra is definitely the fastest Apple Silicon chip on the market. But for many creators the extra performance isn’t worth spending twice as much money compared to the M1 Max.

When the M1 Ultra arrived at the DPReview offices, we weren’t sure what to expect. Twice as much hardware rarely translates into 2x the performance, but it’s also rare that you get a APU that’s literally just two of the previous generation APU fused together.

My guess going into these tests was that video editing would see the biggest bump, photo editing wouldn’t benefit much, and neither would benefit by 2x because the kinds of creative workflows we test just don’t need 128GB of RAM and 64 GPU cores. I was wrong. Of all the use cases we tested, it seems that photo editors who rely on LRC are going to see the biggest performance improvement if they upgrade from the M1 Max to the M1 Ultra, while video editors using either Final Cut Pro or Premiere Pro will have to settle for a smaller performance bump.

It seems that photo editors who rely on Lightroom are going to see the biggest performance improvement if they upgrade from the M1 Max to the M1 Ultra, while video editors using either Final Cut Pro or Premiere Pro will have to settle for a smaller performance bump.

Only you know your workflow and how much time a 10% or 20% speed boost will buy you, but for most users it’s probably not going to be worth spending the extra $3,000 required to get double the CPU, GPU, and RAM.

I can’t emphasize this enough: picking up a fully loaded M1 Ultra Mac Studio is literally more than twice as expensive as the fully loaded M1 Max variant. So unless you’re a die-hard Lightroom Classic user or you’re willing to bet that your creative application of choice is eventually going to optimize its way to 2x performance, the M1 Max is almost certainly the smarter purchase for most photo and video editors.

Back to top


Notice: Trying to access array offset on value of type bool in /var/www/vhosts/worldgames.gr/httpdocs/wp-content/themes/the-next-mag/inc/libs/tnm_core.php on line 746

Notice: Trying to access array offset on value of type bool in /var/www/vhosts/worldgames.gr/httpdocs/wp-content/themes/the-next-mag/inc/libs/tnm_core.php on line 746

Video: Steve Perry goes hands-on with the new Nikkor Z 800mm F6.3 VR S lens in Florida

Previous article

Leica unveils $4,495 limited-edition matte black Summicron-M 28mm F2 ASPH lens

Next article

You may also like

More in DPR