First Mac Studio M3 Ultra benchmarks significantly outpace the M2 Ultra

AppleInsider

Unsurprisingly, the performance of the new Mac Studio with M3 Ultra appears to be markedly faster than its predecessor, at least according to initial benchmarks.

Apple's Mac Studio

Following the first benchmarks of the M4 Max in the new Mac Studio, initial figures are starting to appear for the M3 Ultra version. Compared to the previous Mac Studio's M2 Ultra, Geekbench shows improvements that will be significantly noticeable in the most demanding uses.

There will be more sets of figures for the Mac Studio's M3 Ultra once the machine is users' hands. However for now, the headline is that it is between 16% and 30% faster than its predecessor.

In November 2024, AppleInsider compared how Apple Silicon had improved since the M1. At that point, the then-current M2 Ultra Mac Studio scored 2777 single core and 21351 multi-core, in Geekbench testing.

By comparison, the first Geekbench figures for the M3 Ultra show a single-core score of 3221, which is approximately 16% faster. The M3 Ultra multi core score is 27749, or approximately 30% faster than the M2 Ultra.

In Metal GPU benchmarking, the M2 Ultra scored 221824, compared to the M3 Ultra's 259668. That makes the new machine faster by 17%.

Compared to the M4 Max

The Mac Studio is also available with an M4 Max processor, and benchmarks for that chip have been established in the 16-inch MacBook Pro. Once again, the M3 Ultra beats the M4 Max -- but only in the multi-core test.

For that, the M3 Ultra's 27749 score is around 7.8% faster than the M4 Max, with its 25735. That's a decent improvement, even if it may not be noticed by most users.

We've seen this before on benchmarks with the M2 Ultra. There not being a huge separation in multi-core benchmarking is mostly an artifact of how the benchmarks work, as the real-world gains for the M2 Ultra were clear.

It's only with the single core that the M3 Ultra disappoints, though. For that, the M4 Max scores 3930 compared to the M3 Ultra's 3221 -- meaning that the M4 Max is some 22% faster for nearly every "casual" computing job.

It does seen to prove, though, that most users not involved in deep computation will be better served with the less expensive M4 Max version of the Mac Studio.

Compared to the M3 Max

It's not a surprise that the M3 Ultra processor is faster than its predecessor. Nor is it a shock that what Apple says is the fastest Mac they've ever made should also beat the M3 Max processor in the MacBook Pro.

But the margin of victory is noteworthy. For the GPU, where the M3 Ultra's Metal score is 259668, the M3 Max MacBook Pro managed 155991.

First Geekbench scores for the M3 Ultra in the Mac Studio

That makes the new Mac Studio a fraction over 66% faster than the M3 Max MacBook Pro. This is the most significant difference, though, with other scores coming in closer.

For instance, the multi-core score for the M3 Ultra is 33.5% faster than that of the M3 Max. The scores are 27749 for the M3 Ultra, and 20785 for the M3 Max.

That's still appreciably faster, but users will need to be on very time-pressured work to notice the single core difference. In that case, the M3 Ultra scores 3221, but the M3 Max is on 2971 for a difference of just over 8%.

With only one example showing in the Geekbench listings so far, any figures have to be considered preliminary at best. But the earliest figures comparing the benchmarks of the M4 Pro versus the M2 Ultra proved to be broadly correct.



Read on AppleInsider

cpsro

Where does 2971 single-core score come from for the M3 Max? Geekbench.com indicates over 3100 for the various Mac platforms with M3 Max... and that's an average of a lot of results that includes some people who didn't run a careful benchmark. My own M3 Max results are over 3200.
As for multi-core, Geekbench.com reports ~20,900 average, while my careful test results were over 21,600.

spencer314

So, only pay the M3 Ultra if you need a lot of GPU performance, or a lot more RAM than you can attach to an M4 Max (or for some reason need double the displays). The Ultra does so far seem to be most underwhelming part of the Apple Silicon family. 

blastdoor

Spencer314 said:

So, only pay the M3 Ultra if you need a lot of GPU performance, or a lot more RAM than you can attach to an M4 Max (or for some reason need double the displays). The Ultra does so far seem to be most underwhelming part of the Apple Silicon family. 

It seems that way when looking at Geekbench 6, but I hope most people who are seriously considering the Ultra would realize that Geekbench 6 might not be the right benchmark to look at. 

It's not that there's anything wrong with GB6, it's just that it is focused on the performance of a single program using multiple cores. But if you are seriously considering paying the premium for a 32 core processor, there's a good chance that your workloads involve multiple programs running at the same time and/or 'embarrassingly parallel' workloads, in many simultaneous programs (or threads) run independently of each other. 

Put another way, GB6 is really more appropriate for mainstream consumer or prosumer users, it's not necessarily the best benchmark for high-end workstation or server type use cases. 

Interestingly, GB5 might actually be the better benchmark for people who can really benefit from a bunch of cores. That's because the multicore GB5 just runs multiple copies of the single thread version of the benchmark concurrently, so it is a good test of an embarrassingly parallel workload. 

If you look at the multithread GB5, you'll see that the M2 Ultra is faster than the M4 Max, which is not what you see with GB6. 

john-useless

I've been a Mac user since the original 1984 model … but benchmarks in recent years confuse me, given the nature of multi-core machines these days, not to mention performance vs. efficiency cores, etc. Consider these two new Mac models:

• The Mac Studio with the base M4 Max processor has a 14-core CPU with 10 performance cores and 4 efficiency cores, plus a 32-core GPU.
• The Mac Studio with the base M3 Ultra processor has a 28-core CPU with 20 performance cores and 8 efficiency cores, plus a 60-core GPU.

I understand the idea that a lot of software basically gets its work done with the CPU and that only some software is written to get its work done with the GPU. I also understand the idea that each generation of processor does its work faster — thus, M4 processors will have higher single-core scores than comparable M3 processors.

But unless those M3 processors are far, far slower than M4 processors (which isn't the case — we're not talking M1 versus M4 here), wouldn't the model with the M3 Ultra outperform the model with the M4 Max every time because the M3 Ultra has twice as many cores? I thought, perhaps mistakenly, that macOS more or less hides the number of cores from software — that is, an app sends instructions to the CPU once, and macOS takes care of giving that work to all of the cores available to it on a given machine.

I have this image in my mind of horses pulling two wagon trains full of cargo (equal amounts in each train) across the plains. One wagon train has 14 horses, and they are younger and stronger. The other wagon train has 28 horses. They're a bit weaker and more tired … but even so, they're not that much weaker, and there are twice as many of them! Wouldn't the 28-horse team (the M3 Ultra) beat the 14-horse team (the M4 Max) every time? (I suppose it's not as simple as that.)

My use case: I do a lot of editing in Final Cut Pro, mostly HD but some 4K, and some of the projects are 30 minutes long. Is it worth it for me to buy a Mac Studio with M3 Ultra? Twice as many horses which aren't that much weaker…

edited March 7

entropys

A strange choice to release an M3 ultra instead of an M4 ultra at this point.  But , no doubt reasons that make sense to Apple.

given it is modern Apple, I reckon prime reason is the profit margin. The M3 is still faster than an M4 max, but importantly parts bin and thus its margin is better than an M4 ultra would be.  

edited March 7

cpsro

For general purpose computing, the M4 Max is barely faster than the M3 Max, and the M3 Ultra is looking like it's only a hair faster than the M4 Max. The next significant speed bump will likely arrive with the M5 family which should be implemented in the next generation ASML process.

nubus

john-useless said:

My use case: I do a lot of editing in Final Cut Pro, mostly HD but some 4K, and some of the projects are 30 minutes long. Is it worth it for me to buy a Mac Studio with M3 Ultra? Twice as many horses which aren't that much weaker…

I would say a M4 Max with 64 GB is a beast of a computer. Apple mainly talks 8K in the press release for M3 Ultra and you're not doing 6K or 8K. The 14 M4 horses are faster and smarter (AI improved from M3 to M4). They also make a lot less noise than 28 M3 horses.

The return policy from Apple will give you 14 days to evaluate the Studio M4 Max 64 GB. Can't imagine that you're going to be disappointed. The video editors at work have long switched to using FCP on laptops.

aderutter

The M3 range had a better jump up from the M2 range than the M4 had from the M3 for my type of workload, including 3d hardware raytracing - so it feels frustrating that the new Ultra is M3 based and thereby not as good as we hoped it would be, but I think the Ultra will be better than people expect, or rather the M4 Ultra would not have been that much better than the M3 Ultra.

For my 3D work the GPU isn’t really important as I primarily use ZBrush nowadays and that is entirely CPU based. Even the Redshift renderer ZBrush comes with is CPU based (but I could choose to subscribe to the GPU version if I felt the need). More RAM and CPU is more important to me than more GPU for my 3D work, as I spend 99% of my time not rendering. 

We’ll have to wait and see what the review tests show to get a better idea of real-world performance but anyone doing significant video work, especially those making use of the specific hardware encoders/decoders could see a big benefit from the Ultra - but they might not notice it much except for shorter rendering times if they are not doing massive projects in 8k.

My other half is a graphic designer and she recently moved from a 2019 intel iMac to a 2024 M4 Pro Mac Mini with Studio Display (both setups cost about the same) and she doesn’t feel that it is that much better except the heat/fans and of course being able to run the latest versions of certain programs. 

The horse train analogy reminds me of something I read about ZBrush cores from some time ago, whereby to determine when more cores is better is to multiple the core frequency with the number of cores and the bigger number wins…

M3 Ultra = (4.05 x 20) + (2.75 x 8) = 103
M4 Max = (4.5 x 10) + (2.95 x 4) = 56.8

an M4 Ultra would have been: (4.5 x 20) + (2.95 x 8) = 113.6 

… I just searched for the source and found this: “This means that a 6 core CPU at 2.5 GHz would have a score of 15 while a 4 core CPU at 3 GHz would only have a score of 12. The former edges out the latter even though each core is slower.” from https://www.reddit.com/r/ZBrush

foregoneconclusion

Max chips have significantly higher memory throughput than the Pro or standard versions of M series chips. Which means that Ultra will have DOUBLE the RAM data processing advantage of the Max.

applepoor

Back in the day of the IIfx in the early 90s which was a two core CPU, Photoshop was about the only game in town claiming to use multi processors.  I did not see much difference when running the same job on my Iici. Adobe admitted they had not mastered muti-core when I called.

So, I wonder in a M3 Ultra Mac Studio if there is a way to verify all 80 GPU cores are actually doing something let alone all 32 of the CPU cores?

The power supply is rated 480 watts so if that load is in process, could a sandwich be warmed on top of the case? 

edited March 8

nubus

foregoneconclusion said:

Max chips have significantly higher memory throughput than the Pro or standard versions of M series chips. Which means that Ultra will have DOUBLE the RAM data processing advantage of the Max.

M3 Ultra has 800 GB/s memory bandwidth while M4 Max is doing 546 GB/s. That is +46.5% and far from "DOUBLE".

9secondkox2

M2 Ultra is ancient news. Who cares how it compares to that? 

M3 ultra is basically old news too. 

In light of the m4 max and the potential it held for an ultra tier, this is so sad. 

Apple should compare to AMD/Nvidia/Intel as well to give a true picture. 

Perhaps the silver lining is that we get an m5 ultra sooner than later. 

Until then, this is kind of maddening. The cpu multi score is barely faster than m4 max. 

Whereas an m4 ultra would put apple competitive with nividias best, this is just kind of ok in the grand scheme. 

Bought an m3 air last year to hold me over until some serious GPU power comes in. 

Been waiting on a new large iMac/iMac Pro with real horsepower. But kinda losing hope. May have to break down and get a Mac Studio with the pro XDR when they update it with better refresh. 27” is too small these days. 

Can’t wait for a year from now. 

danox

The true picture will be known by the end of next week Friday, Saturday at the latest, and Apple is probably using them in the back room for some time for Apple Intelligence right now it is good news that they are releasing and developing something like everyone else I was hoping they would go a little bit faster, but when they announced last year that they were using M2 Studios as servers in the background for Apple Intelligence that meant that they were finally reluctantly getting serious, I’m more interested in is what type of network software is Apple developing in the background to allow many Mac computers to operate together as one in some type of cluster it has always seemed that Apple Is more behind (not working all that hard) in that aspect of computing? Due to profit motive.

Apple over the years, seems to be more reluctant (like target video display mode same reason profit) to support multiple Macs being used together as one to accomplish some computer task in comparison to PC or Linux computers like gaming I don’t think it’s because they’re incapable. I also hope with the increasing mind share in the AI area that some developers will start to write software to support increased networking between Macs, someone’s going to do it I hope.

edited March 8

foregoneconclusion

nubus said:

foregoneconclusion said:

Max chips have significantly higher memory throughput than the Pro or standard versions of M series chips. Which means that Ultra will have DOUBLE the RAM data processing advantage of the Max.

M3 Ultra has 800 GB/s memory bandwidth while M4 Max is doing 546 GB/s. That is +46.5% and far from "DOUBLE".

M3 Ultra is double the M3 Max. M4 Max can't be configured as an Ultra. 

stabitha_christie

9secondkox2 said:

M2 Ultra is ancient news. Who cares how it compares to that? 

M3 ultra is basically old news too. 

In light of the m4 max and the potential it held for an ultra tier, this is so sad. 

Apple should compare to AMD/Nvidia/Intel as well to give a true picture. 

Perhaps the silver lining is that we get an m5 ultra sooner than later. 

Until then, this is kind of maddening. The cpu multi score is barely faster than m4 max. 

Whereas an m4 ultra would put apple competitive with nividias best, this is just kind of ok in the grand scheme. 

Bought an m3 air last year to hold me over until some serious GPU power comes in. 

Been waiting on a new large iMac/iMac Pro with real horsepower. But kinda losing hope. May have to break down and get a Mac Studio with the pro XDR when they update it with better refresh. 27” is too small these days. 

Can’t wait for a year from now. 

Wait, a MacBook Air meets your needs but you are whining about the Ultra, a process you don't use nor need. You have absolutely out dumbed yourself, good job. 

rogue01

$2,000 more for a slower Mac for the majority of tasks.  Great job Apple.  Same stupidity as the $3,000 price hike for a Mac Pro that has very few cards that work in it.  

mattinoz

Rogue01 said:

$2,000 more for a slower Mac for the majority of tasks.  Great job Apple.  Same stupidity as the $3,000 price hike for a Mac Pro that has very few cards that work in it.  

If you have a single thread bound workflow why you even buying a studio over a mini?
or a macbookpro. 

These machines are for people whose workflows aren’t constrained like that. That is why they buy machines like this or rack a dozen minis into a small cluster or both. 

programmer

john-useless said:

I've been a Mac user since the original 1984 model … but benchmarks in recent years confuse me, given the nature of multi-core machines these days, not to mention performance vs. efficiency cores, etc. Consider these two new Mac models:

• The Mac Studio with the base M4 Max processor has a 14-core CPU with 10 performance cores and 4 efficiency cores, plus a 32-core GPU.
• The Mac Studio with the base M3 Ultra processor has a 28-core CPU with 20 performance cores and 8 efficiency cores, plus a 60-core GPU.

I understand the idea that a lot of software basically gets its work done with the CPU and that only some software is written to get its work done with the GPU. I also understand the idea that each generation of processor does its work faster — thus, M4 processors will have higher single-core scores than comparable M3 processors.

But unless those M3 processors are far, far slower than M4 processors (which isn't the case — we're not talking M1 versus M4 here), wouldn't the model with the M3 Ultra outperform the model with the M4 Max every time because the M3 Ultra has twice as many cores? I thought, perhaps mistakenly, that macOS more or less hides the number of cores from software — that is, an app sends instructions to the CPU once, and macOS takes care of giving that work to all of the cores available to it on a given machine.

I have this image in my mind of horses pulling two wagon trains full of cargo (equal amounts in each train) across the plains. One wagon train has 14 horses, and they are younger and stronger. The other wagon train has 28 horses. They're a bit weaker and more tired … but even so, they're not that much weaker, and there are twice as many of them! Wouldn't the 28-horse team (the M3 Ultra) beat the 14-horse team (the M4 Max) every time? (I suppose it's not as simple as that.)

My use case: I do a lot of editing in Final Cut Pro, mostly HD but some 4K, and some of the projects are 30 minutes long. Is it worth it for me to buy a Mac Studio with M3 Ultra? Twice as many horses which aren't that much weaker…

Excellent questions.

The short answer is "its complicated".

A slightly longer answer includes some of the following factors:

• No software is 100% parallel.  There are always some components which run serially (i.e. the first result is required before the second can be computed, and so on).  Amdahl's Law (https://en.wikipedia.org/wiki/Amdahl's_law) basically says that parallel hardware can only speed up the parallel portion of a workload, so scaling to an infinite number of cores would only get you to the speed of the non-parallel portion.
• Parallel cores aren't entirely unrelated.  They must communicate (which is often the serial portion of the algorithm), and that communication introduces some slow downs.  Even if they aren't explicitly communicating, they are sharing resources (e.g. the connection to memory) and thus run into contention there which slows them down a little.
• Signals crossing between chips (even Apple's ultra function connector) tend to be slower than on-chip signals.  This means that those communication overheads get a little worse when crossing from one Max to the other, and you can't always avoid that crossing (indeed, Apple's OS makes it mostly invisible to the software... but nobody would likely try to optimize for that anyhow).
• Horse analogy:  one horse by itself doesn't contend with anything put pulling on its load and pushing on the ground.  Two horses have to deal with the connection between them, jostling from the other, etc.  28 horses would have a whole lot of tugging and jostling, and who knows, maybe some of them don't like each other so there's kicking and biting happening too.  The digital equivalent of that does happen.  
• The bottleneck in a computation might not be how fast the instructions execute.  It might be memory latency or bandwidth, I/O latency or bandwidth, or use of some special function hardware (encoders/decoders, neural units, etc).
• GPUs are very very parallel, but each parallel thread of work they can do is less general and less performant than in a full-fledge CPU.  So they aren't great for all tasks, and the software running on them has to pretty much be written specifically for them.
• CPUs vary greatly, and the M-series chips have 2 kinds -- efficiency vs performance.  The former are slower, and the OS needs to figure out where to run what.  It doesn't always get that right, at least not right away.
• CPUs these days get a lot of their performance by executing multiple instructions at the same time from one sequence of instructions.  A lot of those instructions have to execute in the right order, and that limits how many can be done at once.  At an extremely detailed level this depends on the software being run.  Some software is carefully crafted to run as many non-intertwined instructions in parallel as possible, and then having a CPU that has a very "wide" dispatch and SIMD instructions can go very fast.  Most software is nowhere near that carefully crafted (sometimes its just not possible, sometimes its just not worth the effort, and sometimes there hasn't been the time or expertise available), so the in-core parallelism is only lightly utilized even though the CPUs are actively trying to re-order the instructions to go as fast as possible.
• The slowest thing in most modern machines is the memory (well, the I/O is slower, but inside the computer...).  To deal with that, a hierarchy of memory caches are built into the chip.  These are (relatively) small high speed memories that hold copies of data that has already been read from or written to the main memory.  Since it is very common to re-access a given piece of data that has been accessed recently, keeping it in a high speed cache close to the processor can help a lot with performance.  But its not magic, and there are always tradeoffs.  Caches work on bunches of data, and they are divided into levels (usually called L1, L2, L3) of varying size and speed and different amounts of sharing between cores.  This means they're not working with just what the program needs next, they're doing extra work here and there.  The sharing between cores means they're competing for this resource.  Plus a mix of software runs on the same cores (e.g. your workload, the UI, the file system, the networking, the browser you leave running in the background, etc), and it wants different data in the same caches.  Optimizing for cache use is extremely challenging.
• ... and so on.  And on.  And on.  It really is very complicated.  

netrox

programmer said:

john-useless said:

My use case: I do a lot of editing in Final Cut Pro, mostly HD but some 4K, and some of the projects are 30 minutes long. Is it worth it for me to buy a Mac Studio with M3 Ultra? Twice as many horses which aren't that much weaker

I'd say that in your case, no if you own a Mac purchased since the M1 or last generation of Intel with discrete cards (AMD). 

Video and image editing does not require "deep computation" - M4 Max is blazing fast - it does the job very well. You don't even need 64GB of RAM - it makes no difference at all.

Mac Studio with M3 Ultra seems targeted toward users who use AI LLMs and scientific computing where loading massive amount of datasets into RAM can make a huge difference. LLMs are often around 4GB at a minimum (and produces laughable outputs) to 128GB with better outputs. For AI to work efficiently, they need to be loaded into RAM. It may be cheaper to just use the AI servers than to buy Mac Ultra 3 but if a person truly wants everything "offline" then Ultra 3 is suitable. You can tell that Apple is targeting them by offering massive 256GB or 512GB as an option and advertised about running the LLM in RAM. 


 

programmer

netrox said:
Mac Studio with M3 Ultra seems targeted toward users who use AI LLMs and scientific computing where loading massive amount of datasets into RAM can make a huge difference. LLMs are often around 4GB at a minimum (and produces laughable outputs) to 128GB with better outputs. For AI to work efficiently, they need to be loaded into RAM. It may be cheaper to just use the AI servers than to buy Mac Ultra 3 but if a person truly wants everything "offline" then Ultra 3 is suitable. You can tell that Apple is targeting them by offering massive 256GB or 512GB as an option and advertised about running the LLM in RAM. 

Yes.  I don’t imagine that Apple expects to sell many of those models.  And their margins on them are undoubtedly…. generous.  Important for Apple to have these machines in the market though.