Apple Silicon Mac Pro Revealed at WWDC 2023

[pastrychef]

Geekbench 5 or 6 is not a reliable benchmark. See on Cinebench R23 ; i9-13900K is more faster than any new Silicon MacPro. In real render apps of course. We dont buy i9 or m2 ultra for browsing net or use excell )

Which i9-13900K system renders Pro Res faster than an M2 Ultra in FCPX?

 

[trs96]

But what is it for now?

It's for a much smaller niche Pro market that will still buy it, use the PCIe slots and write off the total cost as a business expense. It's really not for about 99.9% of the people here at tmx86. The Mac studio is the best way to go now if PCIe add in cards are not needed. The constant geekery YT channel made a good video explaining all this clearly.

 

[craighazan]

I’m just upset and sad… (sigh). It feels like the end of something to me, a split from the past, the end of the tower.

 

[trs96]

It feels like the end of something to me

I know, the first Intel MP towers were some of the best machines ever made by Apple. Some still use them to this day. Or at least they gut them out and put in their hackintosh hardware. Intel support in macOS is soon to be history and we'll likely never see something that truly modular and easily upgradable from Apple.

 

[zipb]

I know, the first Intel MP towers were some of the best machines ever made by Apple. Some still use them to this day. Or at least they gut them out and put in their hackintosh hardware. Intel support in macOS is soon to be history and we'll likely never see something that truly modular and easily upgradable from Apple.

They(the towers of old) represented a computer generation that could not do a pro job without (lots) of extra hardware.

I had 30K of just computer hardware(10k for the mac, 20k for peripherals) in the 90-ties to do pro audio work.

Nowadays, even a standard 2k laptop is plenty powerful.

But in the 90-ties I could charge much more for my services, as pro audio was a specialized niche, expensive and valued. Nowadays, not so much.

I could afford a real Mac then. Now, hardly.

 

[bombardier10]

Which i9-13900K system renders Pro Res faster than an M2 Ultra in FCPX?

Pro Res files is a apple trick to make it seem that FCPX is faster . Try video native files (directly from camera) on FCPX and all slown down...The trick is that first the files are converted to Pro Res and then you work on them. Many apple apps are special optimized for Silicons. I works on Adobe Premiere and i9-13900K and here the situation is different.

 

[bombardier10]

Ok time for final conclusion about old and new MacPro.

MacPro 2019 on the day of release

CPU - up to HighEnd 28 core Xeon . One of the fastest processors four years ago .
GPU - Computer was always ready to use fastest Radeon graphic on the market.
So we always get fastest Radeon technology.
Modular architecture - easy replace memory, CPU, GPU .

MacPro 2023 on the day of release

CPU - M2 Ultra comparable or slower to the consumer i7-13700K.
GPU - two times or more slower than Radeon RX 7900 XT.
closed architecture. No possible change/replace memory , CPU , GPU.


In this brief overview, you can already see that now the new Mac Pro is
far behind the most powerful PCs and much less attractive than old MacPro.
…On the day of release, of course.

After that, this distance will increase . Soon the release of Intel's 14th generation
processors and in some time newer Radeon/Nvidia graphics will come in.
And the Mac Pro will remain the same for years .

MacPro is obviously a fast machine (or super fast) but only in the closed
world of Apple Technology.

If the goal of the new Silicon CPU/GPU technology was to deliver more efficient
PC computers, it is clear that it has failed.
If the goal was to reduce the price of Apple computers then that didn't work either.
If, on the other hand, the goal was to increase Appel's profits then the answer is yes.

 

[pastrychef]

Pro Res files is a apple trick to make it seem that FCPX is faster . Try video native files (directly from camera) on FCPX and all slown down...The trick is that first the files are converted to Pro Res and then you work on them. Many apple apps are special optimized for Silicons. I works on Adobe Premiere and i9-13900K and here the situation is different.

Wait... What?? ProRes is a trick??

Okay... Premiere... When I searched the PugetBench for Premiere Pro 0.98.0 over the last month, the best macOS scores came from M2 Ultras...



Premiere doesn't seem very well optimized considering these scores from M2 Max and M1 Max... (*Note: result from M1 Max is from an older version of Premiere Pro)



The closest I saw from an x86 systems running macOS were a pair of Ryzen builds...



I also found a build with an i5 and an RX 6800 XT on macOS...





Link to PugetBench:https://benchmarks.pugetsystems.com...re+Pro&application=&specs=macos#results-table

Of course, these results don't tell you how much noise is coming from their builds or how hot the systems are running.

 

[pastrychef]

MacPro is obviously a fast machine (or super fast) but only in the closed
world of Apple Technology.

Of course it's only about macOS. This entire site is about running macOS.

If you don't care about running macOS, why are you here?

 

[c-o-pr]

Is the new Mac Pro still using ECC RAM?

It's ECC that set the previous Mac Pros apart from the larger customer market, and partly why they cost more.

It turns out that for some market segments, esp. CAD, medicine, but also AAA media, etc. ECC (inc the GPU) is important from a compatibility and liability where high confidence is needed in data integrity. Random bit errors are not uncommon, and can break a project.

There's a corollary in software architecture where going from single threaded to multithreaded approaches can lead to differences in calculations.

I read an analysis of Adobe After Effects SW evolution that it was partly trapped on a single CPU, and no GPU acceleration because to port it to parallel HW caused computational charges, which leads to different data which upsets compositing downstream. Even given that AE is purely a visual effects medium and that there's no "visual" difference of the data in situ, a workflow might be disrupted by side-effects of the data having different numbers, blowing compatibility and therefore contracts.

I don't have direct experience with these specific topics, but I have no trouble seeing a point of complexity that may only be examined in more rarified domains of IT. When companies are involved in huge collaborative ventures with big investments are on the line, they take care to avoid contributing errors to shared data sets.

This concern of data integrity can be more general, and partly explains why Apple has taken to gluing their products together, just to better control reliability. With very deep pockets, a vertical stack, and direct end-user application exposure, Apple needs to concentrate on data integrity in their lineups.

So to put a previous post's observation about Cinebench scores into more gross terms, the 13700K and 4080 are built for gamers, who are, as a market, very tolerant of errors!

But there are plenty of ventures where getting the highest score doesn't matter if you get disqualified from the game due to wrong data.

It so happens that hackintoshing mindset substantially overlaps with gaming market in its bias towards happiness to get something working at all, with less concern for any refined notion of "working."

Apple surely is a far more interested in the topic of correctness of its designs, which is something to keep in mind when comparing oranges to app... (I didn't say it!)