MP 7,1 Mac Pro 8.1 - WHAT IF...

[maikerukun]

it's pretty sure they will make an "afterburner like" card for rendering/graphics, an apple mpx module in the m2 ultra mac pro. They are not crazy, they know the power of nvidia gpus

I'm literally praying that what I started this whole thread out with turns out to be correct...something in the form of a brand new chip they've been working on and haven't said a single word about...their own GPU very specifically designed to compete against Nvidia and AMD on THEIR highest levels...

 

[deconstruct60]

I really really hope Apple and AMD release 7000 series for the 2019 Mac Pro. Although my guess is if the new Mac Pro can't even beat the top tier MPX modules available for the 2019, then it isn't going to happen.

There is lots of stuff that goes into GPU drivers. Some of it can be redundant.

"...

Driver Version (GPUs)             Download Size (KiB)    Size (MiB)
AMD 22.11.2 (RDNA 2 and earlier)     558,886           545.8
AMD 23.1.1 (RDNA 3 only)                603,71             589.6
Nvidia 528.02 (RTX 40 and earlier)    832,540           813.0
Intel 31.0.101.4034 (Arc and Xe)      1,243,656        1,214.5

...
"

AMD, Intel and Nvidia GPU Driver Sizes Compared: Feature Creep

Bloated drivers can't be good for performance

www.tomshardware.com

If there was relatively super small gaps in the RDNA2 and RDNA3 driver code then it is somewhat odd that AMD has to about as large downloads for each one. The change from monolithic to chiplets at the low level is probably leading to substantive differences in the low level code. If AMD's 'Pro' GPU drivers for RDNA3 arrive with the same 'duplicative' impact on footprint even more indicative that there is substantially 'new ground' being covered with the new drivers.

If RDNA2 and 3 had been very close at the driver level then maybe Apple would be willing to throw money at an update. If it ls lots of work , then that gets doubtful. If Apple was/is intent on making AMD 'eat the upfront costs' of doing the drive update.... even more doubtful.

If Apple intends to sell the MP 2019 alongside a M-series version for a while , they are unlikely going to 'move the goal posts' on the comparisons between the two.

Apple already said at the Mac Studio introduction that the Ultra version was better than the 16 core MP and W5700 (the two most common chosen configurations ). It is extremely unlikely that Apple is looking at the MP 2019 as a 'growth' market. Some folks will still buy the Mac Pro 2019 , but some will drop out. (that drop out is the shrinking part).

Lots of the "gotta have a 4090 killer' talk is not aimed at the W5700 zone but at the W6800 Duo and W6900 and Apple would just address the maximum GPU for the max "make it rain" big spenders. Again, not a growth market. Not even close. If toss out the most popular GPU class out of the coverage then pretty much guarantee that it is a shrinking market that they would be addressing. Throw on top the folks who will just bootcamp install Window 10/11 and run GPU drivers that already exist, it is an even smaller market. ( users squatting on MP 2019 with Windows buys Apple additional time about just as much as releasing a card drive on macOS. Not buying a PC leaves the door ajar.

Technically, I'm not sure there is an Infinity Fabric link on the 7900 base compute tile. If mainly just moving from Duo to Solo MPX cards the growth is going to be what? The W6900 solo is being sold for more than a W6800 Duo. I don't think Apple is shooting for volume, 'lowest cost leader' sales there. )


I think there is a bit of a 'tail wags dog' notion in which Mac products did most of the motivated the baseline discrete GPU driver development. RNDA being used in iMac , MBP systems probably did more baseline driver inertia contributions than the W5700 did. Likewise the mid 6000 series did more than the 6900 sales did.
There is a notion that it was all 'trickle down' . An extremely few top end , new GPU architecture was the main driver to spur the motivation onto the next round. Without the millions of embedded dGPU placements there are major issues getting a new GPU driver development going. Hence, the last two years of largely nothing.


Mac drivers tended to come after (or around the same time as) the AMD 'Pro' Windows ones. It is a little early to see the clear pattern. But if follow the money, it doesn't look good.

 

[maikerukun]

With the 4090s out this may have changed, but I believe a 2019 Mac Pro with a pair of W6800X Duos pretty much had the benchmark crown for GPU rendering in Octane. Probably no longer the case, but even with a single W6800X Duo Octane is very fast on the 2019 Mac Pro.

I really really hope Apple and AMD release 7000 series for the 2019 Mac Pro. Although my guess is if the new Mac Pro can't even beat the top tier MPX modules available for the 2019, then it isn't going to happen.

You're not wrong. And if comparing it to 1 4090, then yes, it still holds the crown. To put it into perspective. 1 w6800x duo card was approximately %33 faster than 1 RTX 3090 card. so 2 w6800x duo's "which is my config so I can confirm the scores" was running in Octane %66 faster than 2 RTX 3090's. So yes, until the 4090 dropped, it was the king of Octane.

With the 4090 being approximately %83 faster than the 3090, that places it at about %50 faster than 1 w6800x duo, meaning my current setup, and anyone with 2 w6800x duo's is still running a machine that is basically the exact same speed in Octane as 1 RTX 4090.

Meaning...the ONLY way that a 2023 AS Mac Pro becomes enticing to someone like me with the setup that I have, is if it can be equivalent to or BEAT a PC with an RTX 4090 in it in Octane and Redshift Benchmarks (and real world usage btw.).

This is EXACTLY why I just want them to make ONE MORE INTEL Mac Pro! Because the only 2 issues with my system is that the CPU is trash, and there aren't enough CPU cores in the first place. If I could get a 56 core CPU minimum running on the most up to date Intel architecture, next to the M1 neural engine rocking the exact same GPU setup I currently have (2 w6800x duos) don't even need next gen GPUS lololol, alongside updated PCIe 5 slots...I will genuinely, literally, and without hesitation cough up $60k for that machine and write a personalized thank you email to Tim Cook daily for the entirety of my AppleCare contract on it.

The 2019 is a freaking Dragon missing one scale exposing the soft outer layer of skin enclosing it's heart...which is why that one fatal arrow stops it dead in it's tracks when it comes to anything outside of 3D and VFX work (I NEVER edit 4k+ footage on it running beyond 30fps) because my M1 Max MacBook Pro 16inch chews through 8k Red footage like butter.

So each of these machines, 2019 Mac Pro and M1 Max MacBook Pro...need to basically just be 1 machine and I'll never ask Apple for anything else ever again...

...well, until next year lol.

 

[deconstruct60]

it's pretty sure they will make an "afterburner like" card for rendering/graphics, an apple mpx module in the m2 ultra mac pro. They are not crazy, they know the power of nvidia gpus

They know and yet won't sign any driver that Nvidia makes. There is a huge gap between knowning and desperately feeling that it is essential. If take a rational look, it is pretty clear that Apple does not consider Nvidia absoletely essential. It is a "nice to have it is aligns with the rest of the strategy" level performance, but it isn't critical to the overall Mac ecosystem.

Afterburner is transparent to the software calling it. If it is not there Apple's libary dispatches the work to come CPU and/or GPU cores to be done. If present then dropped on the FPGA. There is no user application level code difference.

A 'discrete Apple GPU' would be. Apple has spent the last 2.5 years trying to get developers to write more code that is highly presumptive of there being no NUMA and all unified memory. Changing the code from what it was to the 'new world order'. It is not completely transparent at all.

Afterburner was a wide release Prototype of ProRes acceleration before Apple weaved it into the standard SoC logic. That is primarily it. There very likely will not be another follow on Afterburner. Just a place to work out the kinks before it got far, far, far more expensive to fix. ( also where they could more easily adapt to changes in the encoding driven the be RED lawsuit over patented compression. ). Afterburner is a "one and done".

So why would Apple do a "one and done" dGPU? If it is not a "one and done" then it isn't Afterbuner-like. So where is the representative example they are suppose to be following?

 

[maikerukun]

How do you even use octane on macpro2019 these days? Do you have to stay on pr14? With what host dcc?

Yep. Unfortunately I run Cinema 4D R2023 on my M1 Max MacBook Pro and I use C4D R23-S26 on the 2019 Mac Pro very specifically because Octane ONLY USES ALL 4 26800x GPUs on PRE-2023 builds of C4D.

If you want to use Octane in Cinema 4D 2023, you have to both pay for it, AND settle for only using 2 of your 4 GPU's...because again, Apple killed Octane for Intel and OTOY is only moving forward with heavy development updates for the M series chips smh.

 

[innerproduct]

Yep. Unfortunately I run Cinema 4D R2023 on my M1 Max MacBook Pro and I use C4D R23-S26 on the 2019 Mac Pro very specifically because Octane ONLY USES ALL 4 26800x GPUs on PRE-2023 builds of C4D.

If you want to use Octane in Cinema 4D 2023, you have to both pay for it, AND settle for only using 2 of your 4 GPU's...because again, Apple killed Octane for Intel and OTOY is only moving forward with heavy development updates for the M series chips smh.

This is the reason I never got a mac pro 2019. I wanted the software to be ready before I pulled the trigger and that sadly happened long after the ASi announcement. I was eyeing lower tier solutions though with standard gpus instead of mpx modules for some “bang for the buck”. Anyway, glad you can use your system but sad about apple and octane not having the decency to support its most loyal and spendy customers. Hope they fix this in the end.

 

[ZombiePhysicist]

This is the reason I never got a mac pro 2019. I wanted the software to be ready before I pulled the trigger and that sadly happened long after the ASi announcement. I was eyeing lower tier solutions though with standard gpus instead of mpx modules for some “bang for the buck”. Anyway, glad you can use your system but sad about apple and octane not having the decency to support its most loyal and spendy customers. Hope they fix this in the end.

Sadly this is the case for lord of software. Quick recent example, using handbrake uses like 20% of the cores available. You cannot renice it into doing more. Solution. Make 4 or 5 copies of the app and manually distribute the task. Hell, handbrake won’t even let you click to mass confer a folder full of ripped dvds. You have to manually add each one at a time like an animal.

I love OCRKit. It is borderline magic in how fast it ocr’s large documents. However, it too cannot take up a significant share of the available CPUs.

‘This seems to be mostly lame OS cpu scheduling And lack of optimization. I hope as “standard system” core counts increase this will improve.

 

[deconstruct60]

You're not wrong. And if comparing it to 1 4090, then yes, it still holds the crown. To put it into perspective. 1 w6800x duo card was approximately %33 faster than 1 RTX 3090 card. so 2 w6800x duo's "which is my config so I can confirm the scores" was running in Octane %66 faster than 2 RTX 3090's. So yes, until the 4090 dropped, it was the king of Octane.

With the 4090 being approximately %83 faster than the 3090, that places it at about %50 faster than 1 w6800x duo, meaning my current setup, and anyone with 2 w6800x duo's is still running a machine that is basically the exact same speed in Octane as 1 RTX 4090.

That whole segment is a bit detached from real market factors. The "double duo" is competitive with the 4090 is completely detached from price. The double Duo is $10,000. The 4090 is $2,000 ( card plus maybe some power and or thermal upgrades for the rest of the box. ) So these 'competes with' costs 5x as much. Is that really going to complete? The Performance/$ it is not even close. Sure there are some folks who can 'dump' wildly uncompetitive Pref/$ costs onto their clients, but that isn't going to work for the bulk of the market.


If (and that is a big 'if' ) Apple looked over at the Nvidia side 'looking' for competition it is far more likely that they would be looking at the RTX 6000 Ada . That card lists for $6,800

NVIDIA RTX A6000 "Ada" workstation GPU officially costs $6,800 - VideoCardz.com

NVIDIA lists RTX A6000 ADA GPU at $6,800 The fastest RTX workstation card launches a month after it was supposed to. NVIDIA RTX 6000 (Ada Generation) We have already reported on RTX A6000 ADA pricing a few weeks ago, as it was revealed by some US retailers. However, the card did not actually go...

videocardz.com

Apple's W6900X , which goes unmentioned in the above quote about 'duos' , goes for $6,000 . That is probably a 'floor' of what Apple would charge for some hyper low volume , hard to source package. Apple would be eyeballing problems that were more in the > 32GB VRAM range that the RTX 6000 Ada goes after to leverage same reason paying lots more in both cases.


The reality is that a large fraction of folks yelping for dGPUs are really focus on priced anchored mainstream consumer level priced dGPUs. Not high end Pro GPUs pricing. The clamor for a 7900 MPX module is largely so they can skip the MPX module and buy some retail off-the-shelf card at much lower prices (either now or at even lower prices much later in used market). It is the same unhealthy ecosystem market dynamics that plague the early Mac Pro 2006-2013 EFI (not UEFI) era where folks went through lots of gyrations to not buy the 'official Mac GPU" and bootleg ROM imagines onto mainstream cards at lower prices. People wanted the cards but didn't want to pay for the drivers.


Apple merging into UEFI over time was largely a mirage that Apple was excited about the retail consumer dGPU card market. They were not. That was largely mostly a side effect of requirements of the Intel CPU (and iGPU) a spill over into Mac systems sitting on top of the same major components ecosystem. Apple make complaints that they didn't like UEFI. ( T2 scrubbed/validating firmware for security holes before handing only a copy to the Intel CPU). And basically dumped the whole boot stack on the M-series transition. If Apple was hyper , super duper in love with retail consumer GPUs cards they wouldn't have made that move.

Every full size MPX module that Apple made was not priced at retail consumer GPU norms . The retail GPUs were not the primary design objective. MP 2019 allowed them to be used , but that was more a secondary consideration after design decisions that Intel and/or AMD made .

.I will genuinely, literally, and without hesitation cough up $60k for that machine and write a personalized thank you email to Tim Cook daily for the entirety of my AppleCare contract on it.

I highly doubt the vast bulk of the Mac Pro market is out to spend > $50k per system. That Mac Pro being the 'conspicuous consumption" system of the Mac line up isn't a role that I think Apple wants it to be in. The Mac Pro pressing problem now is actually delivery value/$ , not getting to the highest price tag ever. The gold Apple Watch was a goofy tangent to what really needed to be done in the Watch space. It is not what Apple historically does really well. Apple isn't about delivering the most affordable system possible. However, they are about as equally bad at delivering the most expensive system possible too. $50k and a default 1 year warranty. Really? No onsite tech support. No grand multiyear roadmaps . etc etc. Apple as a whole isn't that kind of vendor.

$50K can get you a Threadripper 5000 series 64 cores and two RTX 6000 Adas.


The next Mac Pro would probably do much better to put more value in the $6K-36K price range than trying to come up with some $45K 'new shiny' that they sold to less than 4,000 a year units. If Apple added 4,000 units to the 6K-36K zone and lost 4000 in the over $36k zone then that would be a net loss of zero of units sold. A $6K computer with just a AMD 5500 or 6600 GPU is not really competitive in most of the single user workstation market. The Mac Pro as a product line would be helped greatly if $6K was at least 'North' of a W5700. (i.e., they need to move more GPU performance down the price ladder. Not crank the ladder up to reach the top of the skyscraper. ). The bottom "half" of the current Mac Pro price range has more problems than the upper 5 percentile does.

The 2019 is a freaking Dragon missing one scale exposing the soft outer layer of skin enclosing it's heart...which is why that one fatal arrow stops it dead in it's tracks when it comes to anything outside of 3D and VFX work (I NEVER edit 4k+ footage on it running beyond 30fps) because my M1 Max MacBook Pro 16inch chews through 8k Red footage like butter.

So Apple has a strength outside the 3D and super high end VFX work zone , but the Mac Pro shouldn't lean into that area where Apple has a significant lead? Really? The Mac Pro could be an even 'hotter knife' for even 'thicker butter'... but skip that.

 

[maikerukun]

This is the reason I never got a mac pro 2019. I wanted the software to be ready before I pulled the trigger and that sadly happened long after the ASi announcement. I was eyeing lower tier solutions though with standard gpus instead of mpx modules for some “bang for the buck”. Anyway, glad you can use your system but sad about apple and octane not having the decency to support its most loyal and spendy customers. Hope they fix this in the end.

Trust me fam, I get it. And yeah, I hope they fix this grave error as well. As I explained above, this system will be easily a top of the line system (top 3 including PCs sub $20k) for at least the next 3 years or so considering the turnaround time from the 30 series to the 40 series for Nvidia...and don't let Apple actually decide to create a 7000 series with AMD for the system because at that point, it immediately gains another 5 to 7 years of top 3 for high end graphic usage life.

But let's see what they're hiding. They HAVE to be hiding SOMETHING...There is a secret Apple DGPU that they haven't announced and it's coming...there has to be...otherwise the 2019 will remain King in the world of Apple for the high end and I don't think they want that even in the forums...

 

[maikerukun]

Sadly this is the case for lord of software. Quick recent example, using handbrake uses like 20% of the cores available. You cannot renice it into doing more. Solution. Make 4 or 5 copies of the app and manually distribute the task. Hell, handbrake won’t even let you click to mass confer a folder full of ripped dvds. You have to manually add each one at a time like an animal.

I love OCRKit. It is borderline magic in how fast it ocr’s large documents. However, it too cannot take up a significant share of the available CPUs.

‘This seems to be mostly lame OS cpu scheduling And lack of optimization. I hope as “standard system” core counts increase this will improve.

"You have to manually add each one, one at a time like an animal." Very possibly making this a t-shirt to sell in my online store

 

[maikerukun]

That whole segment is a bit detached from real market factors. The "double duo" is competitive with the 4090 is completely detached from price. The double Duo is $10,000. The 4090 is $2,000 ( card plus maybe some power and or thermal upgrades for the rest of the box. ) So these 'competes with' costs 5x as much. Is that really going to complete? The Performance/$ it is not even close. Sure there are some folks who can 'dump' wildly uncompetitive Pref/$ costs onto their clients, but that isn't going to work for the bulk of the market.


If (and that is a big 'if' ) Apple looked over at the Nvidia side 'looking' for competition it is far more likely that they would be looking at the RTX 6000 Ada . That card lists for $6,800

NVIDIA RTX A6000 "Ada" workstation GPU officially costs $6,800 - VideoCardz.com

NVIDIA lists RTX A6000 ADA GPU at $6,800 The fastest RTX workstation card launches a month after it was supposed to. NVIDIA RTX 6000 (Ada Generation) We have already reported on RTX A6000 ADA pricing a few weeks ago, as it was revealed by some US retailers. However, the card did not actually go...

videocardz.com

Apple's W6900X , which goes unmentioned in the above quote about 'duos' , goes for $6,000 . That is probably a 'floor' of what Apple would charge for some hyper low volume , hard to source package. Apple would be eyeballing problems that were more in the > 32GB VRAM range that the RTX 6000 Ada goes after to leverage same reason paying lots more in both cases.


The reality is that a large fraction of folks yelping for dGPUs are really focus on priced anchored mainstream consumer level priced dGPUs. Not high end Pro GPUs pricing. The clamor for a 7900 MPX module is largely so they can skip the MPX module and buy some retail off-the-shelf card at much lower prices (either now or at even lower prices much later in used market). It is the same unhealthy ecosystem market dynamics that plague the early Mac Pro 2006-2013 EFI (not UEFI) era where folks went through lots of gyrations to not buy the 'official Mac GPU" and bootleg ROM imagines onto mainstream cards at lower prices. People wanted the cards but didn't want to pay for the drivers.


Apple merging into UEFI over time was largely a mirage that Apple was excited about the retail consumer dGPU card market. They were not. That was largely mostly a side effect of requirements of the Intel CPU (and iGPU) a spill over into Mac systems sitting on top of the same major components ecosystem. Apple make complaints that they didn't like UEFI. ( T2 scrubbed/validating firmware for security holes before handing only a copy to the Intel CPU). And basically dumped the whole boot stack on the M-series transition. If Apple was hyper , super duper in love with retail consumer GPUs cards they wouldn't have made that move.

Every full size MPX module that Apple made was not priced at retail consumer GPU norms . The retail GPUs were not the primary design objective. MP 2019 allowed them to be used , but that was more a secondary consideration after design decisions that Intel and/or AMD made .






I highly doubt the vast bulk of the Mac Pro market is out to spend > $50k per system. That Mac Pro being the 'conspicuous consumption" system of the Mac line up isn't a role that I think Apple wants it to be in. The Mac Pro pressing problem now is actually delivery value/$ , not getting to the highest price tag ever. The gold Apple Watch was a goofy tangent to what really needed to be done in the Watch space. It is not what Apple historically does really well. Apple isn't about delivering the most affordable system possible. However, they are about as equally bad at delivering the most expensive system possible too. $50k and a default 1 year warranty. Really? No onsite tech support. No grand multiyear roadmaps . etc etc. Apple as a whole isn't that kind of vendor.

$50K can get you a Threadripper 5000 series 64 cores and two RTX 6000 Adas.


The next Mac Pro would probably do much better to put more value in the $6K-36K price range than trying to come up with some $45K 'new shiny' that they sold to less than 4,000 a year units. If Apple added 4,000 units to the 6K-36K zone and lost 4000 in the over $36k zone then that would be a net loss of zero of units sold. A $6K computer with just a AMD 5500 or 6600 GPU is not really competitive in most of the single user workstation market. The Mac Pro as a product line would be helped greatly if $6K was at least 'North' of a W5700. (i.e., they need to move more GPU performance down the price ladder. Not crank the ladder up to reach the top of the skyscraper. ). The bottom "half" of the current Mac Pro price range has more problems than the upper 5 percentile does.






So Apple has a strength outside the 3D and super high end VFX work zone , but the Mac Pro shouldn't lean into that area where Apple has a significant lead? Really? The Mac Pro could be an even 'hotter knife' for even 'thicker butter'... but skip that.

Wow...a couple things lol:

1. You're not listening. As has been pointed out a million times in this thread, folks like me DO NOT CARE how much we spend on that system or the GPUs for it. We are completely aware that we are not the status quo but the anomaly. As such, MONEY DRIVING MY COMPARISON IS MOOT. Money isn't a factor if it's $2k or $50k. Performance is the factor. It's the only one that matters to me. Because reality is, 2 projects and a month later the system is paid off and it's getting written off on my taxes anyway.

2. I said SUB $20k...I'm aware what I can get on the PC side and have no problem doing so...but when 98% of both my work and home ecosystem is Apple...I don't really want to integrate a high end PC into the mix if possible. I currently have a $14k Puget System configed and sitting in the cart waiting for Apple to announce the next Mac Pro and based on what they say when they do, immediately after the announcement I will either complete my purchase or not.

3. Nobody ever said Apple SHOULD NOT lean into their significant lead, and if you look back through the thread, you would know that I LEAN INTO WHAT THEY DO BEST AS WELL...That's what the whole M series is all about...leaning into what they do best. The Mac Pro however, is not about that, and doesn't need to be...because that's what they're already doing with literally every other product they have. So not sure where you're getting your whole "skip that" argument lol.

I'm also gonna show you something...not to be a bu##hole lolol...but I just want you to understand...I literally don't care how much it would cost and neither does anybody else in my tax bracket. In the 3 days that February has existed, I've spent like $8k mostly on clothes as I'm going to Berlin, Germany in a week for the Berlin Film Festival to promote my new movie. I've spend like $25k so far in 2023, and we are literally just a month in lolol. So I'm not exaggerating when I say what I say. $2k, $60k...whatever. Just give me my GPUs and let me work lolol.

I'm not out of touch with reality nor the status quo though...which is why I think the M-Series is incredible, and I love the M series MacBook Pros and the M series Mac Studios (well, I like those...they're a little too close to the Mac minis for my taste, but they have crazy potential) and they are exactly what Apple should lean into and be for literally 99% of the world...but for the 1% of us that want to heat the neighborhood with our GPU's during the long winter months in sunny LA lol, they need the Mac Pro to be a monster. Not because that's what most will use it for...but because why even build it if it can't compete with Pro systems in the PC World? What is even the point if it's not capable of being a monster that can eat the 2019 beast like spicy nuggets at Wendy's? lol.

 

[AlphaCentauri]

I don’t have your budget, but I’m quite well off, no complaints here 🤣

I am willing to spend around £8000 on new Mac Pro (if I like it, if not - I’ll buy £6000 Mac Studio) and I really couldn’t care less how much an equivalent (or faster) PC would cost. If I wanted a PC, I would have bought one long time ago…

It’s a tax write off for me, too. So really not a problem.

 

[Calaveras]

I would love to see Apple make a nod to the inherent nature of workstations;
That they often run very expensive specialize software packages that cater to specific use case scenarios. Hence are not going to have a massive user base compared to stuff like Microsoft Office or Photoshop.
What I am getting at here is that a lot applications real professionals buy very expensive workstation hardware for may not get updates very often. Heck there are many cases of some abandonware livign on for years after the demise of the original company because it's the only thing that does what it does.
I'd love to see a hybrid system that is both.
Xeon and Apple silicon.
Yeah that is crazy. Not possible.
Except Apple has shipped computers in the past with a Motorola and an Intel CPU.
There have been dozens of 3rd party accessories for this. It was pretty much how it was done until virtualization happened.
Now I know a sufficiently powerful machine could virtualize a different architecture and OS.
But the kind of specialized apps I'm talking about usually have explicit hardware calls maybe even a $3000 dollar USB key that has to be plugged in all the time.

I'm dreaming I know. Apple always throws specialized pro users under the bus in favor of the far more numerous prosumers and hobbyists. Vast amounts of money have been made with that approach in other areas like music gear and photography.

 

[mode11]

It is the same unhealthy ecosystem market dynamics that plague the early Mac Pro 2006-2013 EFI (not UEFI) era where folks went through lots of gyrations to not buy the 'official Mac GPU" and bootleg ROM imagines onto mainstream cards at lower prices. People wanted the cards but didn't want to pay for the drivers.

Or to put it another way, it's just not economically viable for Apple to sell macOS towers. Laptops and laptop-like computers, yes, but apparently the only way they can justify selling a tower is to charge twice the price as for the same hardware running Windows. This is a weakness, not a strength, of the macOS platform.

Apple isn't about delivering the most affordable system possible. However, they are about as equally bad at delivering the most expensive system possible too. $50k and a default 1 year warranty. Really? No onsite tech support. No grand multiyear roadmaps . etc etc. Apple as a whole isn't that kind of vendor.

This is a good point. Essentially, they can only sell bad-value desktops. Even at the high end of the workstation market, if they're expected to update hardware regularly, provide comprehensive support and so on, it again becomes economically non-viable (or at least unattractive).

So Apple has a strength outside the 3D and super high end VFX work zone , but the Mac Pro shouldn't lean into that area where Apple has a significant lead? Really? The Mac Pro could be an even 'hotter knife' for even 'thicker butter'... but skip that.

I'd argue the opposite. It sounds like they have that area sewn up already. There's only so many streams of 8K footage people need to edit simultaneously. In any case, fixed-function blocks for video compression / decoding is not that ingenious, and not especially future proof either (e.g. they'd do nothing for "h.266"). Seems they need to concentrate now on general GPGPU muscle, which can turn its hand to anything software developers dream up.

 

[maikerukun]

I don’t have your budget, but I’m quite well off, no complaints here 🤣

I am willing to spend around £8000 on new Mac Pro (if I like it, if not - I’ll buy £6000 Mac Studio) and I really couldn’t care less how much an equivalent (or faster) PC would cost. If I wanted a PC, I would have bought one long time ago…

It’s a tax write off for me, too. So really not a problem.

Which was exactly my point. In the world of high end Apple electronics, Nobody this is purchasing the top end cares about the budget in relation to PC. You already know you're getting the Apple Tax and you've likely made peace with that LOLOL. I really didn't understand why he said that lol...

 

[maikerukun]

I would love to see Apple make a nod to the inherent nature of workstations;
That they often run very expensive specialize software packages that cater to specific use case scenarios. Hence are not going to have a massive user base compared to stuff like Microsoft Office or Photoshop.
What I am getting at here is that a lot applications real professionals buy very expensive workstation hardware for may not get updates very often. Heck there are many cases of some abandonware livign on for years after the demise of the original company because it's the only thing that does what it does.
I'd love to see a hybrid system that is both.
Xeon and Apple silicon.
Yeah that is crazy. Not possible.
Except Apple has shipped computers in the past with a Motorola and an Intel CPU.
There have been dozens of 3rd party accessories for this. It was pretty much how it was done until virtualization happened.
Now I know a sufficiently powerful machine could virtualize a different architecture and OS.
But the kind of specialized apps I'm talking about usually have explicit hardware calls maybe even a $3000 dollar USB key that has to be plugged in all the time.

I'm dreaming I know. Apple always throws specialized pro users under the bus in favor of the far more numerous prosumers and hobbyists. Vast amounts of money have been made with that approach in other areas like music gear and photography.

What an idea, and what a system that would be. I would love to see that system. It's exactly what I have described elsewhere in a sense. That's the trick of it for Apple though...They KNOW all of this, and they COULD DO literally everything everyone on this thread up to this point has suggested...which of these things THEY CHOOSE...that's the fun "and often frustrating" part lol.

 

[maikerukun]

Or to put it another way, it's just not economically viable for Apple to sell macOS towers. Laptops and laptop-like computers, yes, but apparently the only way they can justify selling a tower is to charge twice the price as for the same hardware running Windows. This is a weakness, not a strength, of the macOS platform.



This is a good point. Essentially, they can only sell bad-value desktops. Even at the high end of the workstation market, if they're expected to update hardware regularly, provide comprehensive support and so on, it again becomes economically non-viable (or at least unattractive).



I'd argue the opposite. It sounds like they have that area sewn up already. There's only so many streams of 8K footage people need to edit simultaneously. In any case, fixed-function blocks for video compression / decoding is not that ingenious, and not especially future proof either (e.g. they'd do nothing for "h.266"). Seems they need to concentrate now on general GPGPU muscle, which can turn its hand to anything software developers dream up.

Yep...I love your perspective on that last point.

 

[smckenzie]

Any chance we got this all wrong and the new MacPro will be Intel based? After all, Tim did say they had an Intel Mac coming and the new Xeons are set for the end of Feb.

Exclusive: Intel ‘Sapphire Rapids’ Xeon W3400, W2400 And W790 Chipset Launch Date And More — Wccftech

Okay so I have a short and sweet update on Intel's upcoming Sapphire Rapids-SP, Sapphire Rapids-112L and Sapphire Rapids-64L platforms. Officially dubbed the 4th Generation Intel Xeon Scalable Xeon W-3400 and Xeon W-2400 CPUs and powered by the company's W790 platform, Intel is gearing up for...

apple.news

 

[prefuse07]

Any chance we got this all wrong and the new MacPro will be Intel based? After all, Tim did say they had an Intel Mac coming and the new Xeons are set for the end of Feb.

Exclusive: Intel ‘Sapphire Rapids’ Xeon W3400, W2400 And W790 Chipset Launch Date And More — Wccftech

Okay so I have a short and sweet update on Intel's upcoming Sapphire Rapids-SP, Sapphire Rapids-112L and Sapphire Rapids-64L platforms. Officially dubbed the 4th Generation Intel Xeon Scalable Xeon W-3400 and Xeon W-2400 CPUs and powered by the company's W790 platform, Intel is gearing up for...

apple.news

I would die of happiness!

Though yes -- I do remember: the original rumors were that there would be one more Intel Mac Pro, potentially using the same case as the 7,1 with newer CPUs, and THEN they would perform the transition to ARM by releasing an AS Mac Pro.

This is the same unit that I believe @Amethyst's 'friend' was testing out, but word form him was that this project got scrapped.... Gurman also mentioned something being scrapped, and I'd bet he was potentially referring to the same project?

*shrug*

 

[smckenzie]

Is it fair to say that although AS is performant, its GPU power is still way off that of dGPU? It is from a rendering point of view. A maxed Ultra can’t come anywhere near my 2x Duo’s.

If so, could Apple just be waiting for new Xeons and then release an updated Intel MacPro with updated MPX modules? This would allow them to utilize the existing case and mpx design.

Then the choice becomes pretty clear. Mac Minis and Studios for most people, MacPro for the niche. It would also buy them time until AS is at such a point that it can match dGPU’s.

Maybe a gross over simplification but it makes sense to me. Just find it hard to believe they would go to all that effort of designing the 7’1 and giving it such a short shelf life.

 

[ZombiePhysicist]

Is it fair to say that although AS is performant, its GPU power is still way off that of dGPU? It is from a rendering point of view. A maxed Ultra can’t come anywhere near my 2x Duo’s.

If so, could Apple just be waiting for new Xeons and then release an updated Intel MacPro with updated MPX modules? This would allow them to utilize the existing case and mpx design.

Then the choice becomes pretty clear. Mac Minis and Studios for most people, MacPro for the niche. It would also buy them time until AS is at such a point that it can match dGPU’s.

Maybe a gross over simplification but it makes sense to me. Just find it hard to believe they would go to all that effort of designing the 7’1 and giving it such a short shelf life.

The ultra studio cannot compete with a SINGLE 6800xt. The M2 version of it won't help it beat it as if I recall, a single 6800xt was 2x the metal score of the ultra studio.

As for your hypothesis, it's well reasoned and has logic. Ergo it shall not apply to apple.

 

[prefuse07]

As Zombie pointed out -- my SINGLE RX-6800XT, in my lowly 5,1 smokes a maxed out Mac Studio Ultra (benchmark comparisons can be found in a few threads here)... Your Duos will just decimate it completely, and then some...

@smckenzie I agree with your theory and that would be my thinking as well -- apple definitely can still release one more Intel Mac Pro, thus allowing for continued support for the 7,1 as well, while they work behind the scenes to try to bring AS up to par with Nvidia. Not only this, but it also gives Rosetta more time to mature.

This is really the only way they can do this without embarrassing themselves.... That or they pull another trashcan failure, which I REALLY hope they don't do... but then again, look who's driving the mothership

 

[smckenzie]

They would have to have something pretty special up there sleeve to beat the MacPro on GPU and there’s nothing to suggest they do at this point. Just seems risky to me to release an AS MacPro if it can’t beat the old one. Not a good formula for good publicity and they would look stupid.

On the other hand, if they release an updated Intel MacPro the narrative would be the opposite. They can spout how they still support and value the pro market whilst still moving forward with AS for everyone else and I think most would people would understand. Would be a win win.

Apart from a few instances, Apple seem to be really good at making the most of a design or chips. I mean they got their moneys worth out of the iMac and Mini design for sure. And when they have a new chip, they use the old one in other products for years afterwards. So why would that not be the case with the 7’1?

 

[mattspace]

With the iPad now a bigger business than the Mac, I just can't fathom how the economics work for building a dedicated halo workstation CPU, given how much more powerful / how architecturally different it would need to be to provide a significant differentiator to the existing chips, for a tiny subset of sales. Those of us old enough remember what the prices were like when all the UNIX RISC workstations ran on their own chips.

If I was a shareholder I'd be asking serious questions about whether the money put into developing an AS Mac Pro would be better spent elsewhere, and the Mac Pro left as an IA64 machine (which keeps the cross-platform compiler tools production-ready), with AS based accelerator cards.

 

[ZombiePhysicist]

With the iPad now a bigger business than the Mac, I just can't fathom how the economics work for building a dedicated halo workstation CPU, given how much more powerful / how architecturally different it would need to be to provide a significant differentiator to the existing chips, for a tiny subset of sales. Those of us old enough remember what the prices were like when all the UNIX RISC workstations ran on their own chips.

If I was a shareholder I'd be asking serious questions about whether the money put into developing an AS Mac Pro would be better spent elsewhere, and the Mac Pro left as an IA64 machine (which keeps the cross-platform compiler tools production-ready), with AS based accelerator cards.

That chart would look worse for Porsche. Would you as a shareholder ask them why they would put , forget even the 911, insane r&d money in to the 918 supercar and their next halo supercar when all they really sell are Cayenne SUVs? If they stop with the halo cars, the sales of their overpriced pimp SUVs dry up.

Of course car world isn’t the same, but by pushing developments in heavy duty cpu they can help other parts of their business. Some parts through halo “association” like Porsche. But other parts are like maybe developing server farms. Or heavy graphics ability that eventually can be brought down to mobile devices.

Underestimate the power of halo at the company’s peril, imo.