Review Apple Mac Studio (Early 2025) Review: Renewed vigor with M4 Max and M3 Ultra

[Admin]

The Mac Studio gets a performance boost with M4 Max and M3 Ultra power.

Apple Mac Studio (Early 2025) Review: Renewed vigor with M4 Max and M3 Ultra : Read more

 

[JamesJones44]

Thanks for the review! It's nice to seem some benchmark numbers outside of Geekbench. I thought maybe other benchmarks would show the Ultra would typically out perform the Max for highly parallelized applications, but Geekbench's multi core scaling was masking if that was true or not.

 

[m3city]

Wouldnt amd 7700 be a better comparison? Similar core qty, newer, same single core perf as apple newones.

 

[bit_user]

Thanks for testing. I'd never buy one, myself, but it's still nice to watch what's happening on the other side of the fence.

The single-threaded performance of the M4s seems very impressive.

Lastly, can you clarify whether higher or lower is really better, on the xcode benchmark? I'm pretty sure lower is better, although it's weird that the M2 Ultra came in below the M3 Ultra:

​

 

[bit_user]

I thought maybe other benchmarks would show the Ultra would typically out perform the Max for highly parallelized applications, but Geekbench's multi core scaling was masking if that was true or not.

Geekbench's multithreaded benchmark is pretty useless. It tries to be something in between single-threaded and a fully MT benchmark. But, because we really don't know much about its intrinsic scalability, that ends up making it virtually impossible to divine much meaning from those numbers or use them to make predictions about specific workloads.

The best way to go is measure ST performance of each core type, as a way of characterizing the core. Then, measure fully MT performance. If I want to run something that's lightly threaded, I will know which end to extrapolate from, based on how many threads it has, relative to the CPU.

 

[mc0e134ac]

A flawed review, in many ways:
1. The competing CPUs (Threadripper 5975WX) are last gen. A Threadripper 7980x would be the closest CPU to compare to, as the PRO variants are in a different league. Who would do professional work on 512GB of RAM without ECC?
2. Speaking of "epic performance", yeah, why not add the Threadripper 7995WX PRO too to the comparison?
3. The comparison is also flawed because it is comparing GPU accelerated workloads (like Handbrake) on the M3 Ultra with CPU only benchmarks on the previous gen CPUs.
4. Entire comparisons have been made between Macs only, like the File Transfer test, which shows a very weak 2.7GBs / result, which is weak even by PCIE gen 4 standards. A professional workstation can easily get over 20GB/s
5. The competing GPUs were not disclosed so it can be assumed they were not used. That shows bad faith, because a middle range current gen NVIDIA GPU can easily walk over M3Ultra, not to mention the very large professional oriented Quadros.
6. Weak selection of benchmark which have little resemblance to real world apps.

 

[mikegtx]

Wouldnt amd 7700 be a better comparison? Similar core qty, newer, same single core perf as apple newones.

You are correct. The author chose to compare the new Mac Studio against a 3 year old AMD chip and a 2 year old Intel one. Intel and especially AMD have made great progress since then; their current chips could well meet or exceed the performance of Apple's M3 Ultra. The prices for the AMD and Intel chips cited in the article are also very out-of-date. For these reasons, this review is unfortunately misleading and is disappointingly not up to Tom's Hardware's usual journalistic standards.

The new Mac Studio is an interesting product that deserves a more honest and comprehensive review; hopefully someone at Tom's Hardware will give us this in the near future.