Contribute
Register

Apple Event announced for October 30th: 'Scary Fast'

Assuming those stats are bumping the rev limiter, the 12950HX is pulling 55W and the M3 Max pulls 30W. So that's 2x per W to AppleSi.

I couldn't quickly find stats for low power efficiency, so I have only a wild guess that AppleSi might mean the diff between a day of use on battery vs. half-day.

Microsoft never talks up Windows features for power but the PC industry has a track record of being heavy-handed re power consumption— benchmarking as fast is always more important than a balance of traits to PC. Intel cares very much, but their mobile is downsized desktop, and MSFT abandoned mobile long ago.

OTOH mobile efficiency is the baseline from which the AppleSi has grown, and if Apple doesn't do power well it spoils the heart of their business. So it's remarkable that AppleSi is holding its own in raw CPU against its far better established architectural rival.

I believe that if a detailed story on total system balance were to be told, I will be surprised that IA mobile makes a half-decent showing.

Geekbench can't show off a systems balance of traits at all. And Cinebench is a single workload, whereas GeekB is at least a composite of 10 or so workloads. So we've got a yardstick, but it provides no insights into the most important evolutionary advances of these designs, which are about the SW stack's interplay between dedicated units, the cache-core complex, the GPU-as-an attached processor suitable for a divided realm of workstation-class compute (3D vs tensor/AI) and the primacy of efficiency in network+storage+security.

To restate, Apple's approach is about obviously about bringing the right amount of power to where it's needed in an overall balanced design plan for a personal device and its developers-developers-developers (to quote old Balmer) while MSFT as competition is a herd of cats dragging around some very antiquated assumptions about "PC", not the least of which are phat gaming and backwards compatibility.

Intel, bless their hearts, completely gets this balance perspective, which is shown by the fact that their server / workstation products are no longer applicable to Apple at all. Where brute force is valued, Intel continues to offer 25K–50K workstation-class designs. Apple had to track the lower-end these Intel chipsets over a decade in the form of the 2023–2019 Mac Pros, where in retrospect it's obvious why they let the Mac Pro languish, because they believed they would succeed on the M-series (mobile to Max)— And they did.

While I did not need to upgrade I choose to anyways From 16" MBP M2 Max 32gb ----> 16" MBP M3 Max 64gb. I did it for three reasons First is 32gb was a mistake for me, Second is the ray tracing, Third my wife is starting school and needs a new laptop.

This should allow me faster development of Architectural Renderings in Twin Motion / Unreal. It should also allow me to run Parallels/windows for quick Revit model corrections while working on TM/UE. Hopefully Epic will enable RT for Mac with the M3.
 
Over at Macrumors someone posted results of a test exporting several thousand images via Lightroom comparing M2 and 12th gen mobile:


The bulk of the thread is typical benchmark bickering over validity of claims, but here's the key post:

Adult80HD Wrote: I'll give you actual data. I ran an export of 1465 Sony A1 raw files (51megapixels) to jpg files. This task pushes the CPU and GPU very hard--they will both spike up to near continuous full usage. It took my 16" M2 Max (96GB/38GPU) 18 minutes and 26 seconds to do the task plugged in and 18 minutes and 29 seconds to do so on battery--letting MacOS manage power/battery use. That used of 10% of the battery capacity, so extrapolating it would have eaten up the whole battery in three hours or so. In the meantime, the Lenovo (12th Generation Intel® Core™ i7-12800H vPro® Processor with 14 cores; 128GB of RAM; Nvidia RTX 3070 Ti 8GB) I mentioned above also worked the CPU and GPU very hard on the same task; all verified in the task manager. It took it 33 minutes and 41s to finish the task plugged in--handily beat by the M2 Max--and when I put it on battery the same task took 49 minutes and 22s and chewed up 57% of the battery. The M2 Max has basically a 100Whr battery vs. the 90Whr in the Lenovo, so it has about a 10% edge in total life, but basic math, the exact same task took 5x more power to accomplish on the PC. When I tried to turn off the power savings feature on the Lenovo so that it could run on battery at full speed like the M2 Max can, it killed the battery before the task finished. In fact, it took less than 25 minutes to 100% drain the battery.

The rhetoric on behalf of the PC ends up focusing on gaming framerates, which is a typical turn in these dialogs, where no comparison data is provided because there's no culture of AAA gaming on Apple... Such arguments get tedious.

But there are other observations which make the thread worthwhile to peruse.

It's a Apple-oriented blog, so of course there's the implicit bias of the community, but the best points paint a picture that's not surprising to anyone who has been tracking the Apple HW trajectory: there's no comparison IA, AppleSi is in league of its own and it's crystal clear why Apple went their own way.
 
Over at Macrumors someone posted results of a test exporting several thousand images via Lightroom comparing M2 and 12th gen mobile:


The bulk of the thread is typical benchmark bickering over validity of claims, but here's the key post:



The rhetoric on behalf of the PC ends up focusing on gaming framerates, which is a typical turn in these dialogs, where no comparison data is provided because there's no culture of AAA gaming on Apple... Such arguments get tedious.

But there are other observations which make the thread worthwhile to peruse.

It's a Apple-oriented blog, so of course there's the implicit bias of the community, but the best points paint a picture that's not surprising to anyone who has been tracking the Apple HW trajectory: there's no comparison IA, AppleSi is in league of its own and it's crystal clear why Apple went their own way.
This is actually quite the perfect argument for my use case, I do play one game but I am not sure WoW is considered a "AAA Game". However, I do use Twinmotion and Unreal and those are both heavily based on FPS Especially while working in realtime rendering modes. Unreal now has an AS version and hoping that Twinmotion will follow suit now that Ray Tracing has been added.
 
This is actually quite the perfect argument for my use case, I do play one game but I am not sure WoW is considered a "AAA Game". However, I do use Twinmotion and Unreal and those are both heavily based on FPS Especially while working in realtime rendering modes. Unreal now has an AS version and hoping that Twinmotion will follow suit now that Ray Tracing has been added.

Can you say a bit more about the perfection of the argument?
 
Buying a new Macbook Air/Pro laptop ? It seems obvious but you really must choose the ram upgrade if you can't afford the SSD upgrade as well. 8GB in a laptop in 2023 just won't cut it. You're stuck with what you choose to buy, for the life of the Mac laptop.

Alex Ziskind is a developer who tests out Macbook laptops in this video.


If you don't want to watch the lengthy video testing, skip ahead to about 10:25 for the conclusion.
 
Last edited:
That video is manifest attention deficit disorder!
 
I don't have good news for Logic Pro X users.



 
Last edited by a moderator:
iMac M3 Logic Benchmark test:


1699389222844.png



1699389173427.png
 
iMac M3 Logic Benchmark test
How does the $2000 M2 Max Mac Studio do in this benchmark test ? Looks like 181 tracks.

Screen Shot 8.jpg


I think Apple expects Pro audio customers to buy that or the Ultra instead of an M3 iMac with 8GB of ram. The iMac M3 is meant for consumers and their basic media consumption needs. They don't even know what Logic Pro is.

The M2 Max Mac Studio is probably the best value going in the Mac desktop line right now.
 
Last edited:
Back
Top