MP All Models Mac Pro X,Y…WHAT IF…

[mattspace]

It's because of the way development works. The actual developers love tools and high quality APIs like Metal, but their bosses are idiots who only care about money and say dumb things like "Hey, lets use <dumb technology choice> to save ourselves lots of money"

Games is always going to be a problem for the Mac, because the hardware is console performant, but high end Gaming PC with top of the range GPU in price.

Who's the market for that - people who can afford the expensive mac, but not the console or dedicated gaming pc? People who want to game occasionally... are they worth developing for?

Back when I worked in games, even the PC port was generally a labour of love where the higher ups would be all "Who cares about the PC, they have powerful hardware so don't waste time optimising the PC port when you should be spending your time on the Playstation version"

Hence Microsoft's value proposition for doing Windows / Xbox as the same codebase.

Mac ports are similar where there is literally a 0% chance that your boss is going to come and say "Hey let's make a Mac version" unless someone at Apple has given them a suitcase full of money (and even then it would probably involve doing the bare minimum). It would essentially be up to an individual developer to prototype a Mac port and literally hand it to the higher ups and say "Hey, can we have a Mac version, I've basically ported the game already and all you have to do is accept the free money". That's why these tools are such a massive game-changer.

It still leaves open who the target market is - people who don't want an extra device (aside from their iPad, iPhone, Airpods, Vision Pro etc).

It's also why I don't work in games anymore because it's the most dysfunctional industry on the planet.

I'll credit Apple this - their money-up-front payment model to get games made for their service is a MUCH better model than traditional game financing.

 

[orionquest]

"Apple just lost a customer" says man who was never bought anything from them.

Never said that, what planet are you on? Must be Venus as you are pretty dense!
3 iphone's, MBA, MBP, and 2 ipads, and a host of classic mac's.
Keep trying!

 

[Fravin]

Well, now you can do that.

Will the PSU in MacPro handle AMD GPUs?

 

[Longplays]

In a world where Apple released the 8,1, the Mac 'Doh!. With no 3rd party GPU upgrades. With no upgradable RAM. With no upgradable CPUs...


View attachment 2213152


...Apple actually cared about Pros and Enthusiasts.... Let's play make believe...

All they need to do to turn this around are these 3 things:
1) Add support for GPUs
2) Add the ability to upgrade additional ram (and either extend the SOC ram or treat it like cache like we’ve seen elsewhere). And lets make memory be ECC.
3) Add the ability to replace/upgrade the SOC itself

And they can end the embarrassing 8,1 Mac 'Doh! and give us a real Mac Pro... WHAT IF they actually cared what you (pros/enthusiasts) think... What do you think it should be?

How many "Pros and Enthusiasts" are these?

I pegged this as 20% or less Mac Pro users.

From 2019-2023 Onelifenofear points to about 80,000 MPX units Mac Pro upgrades.

Let us be generous and make that an even 100,000 Mac Pro users worldwide.

Is that sufficient to support R&D resources to comply with all your points?

Would AMD or even Nvidia be interested in developing Mac drivers?

In Mac gaming very few developers were interested in the Mac market and Apple ships millions of non-Mac Pros quarterly.

 

[ZombiePhysicist]

How many "Pros and Enthusiasts" are these?

I pegged this as 20% or less Mac Pro users.

From 2019-2023 Onelifenofear points to about 80,000 MPX units Mac Pro upgrades.

Let us be generous and make that an even 100,000 Mac Pro users worldwide.

Is that sufficient to support R&D resources to comply with all your points?

Would AMD or even Nvidia be interested in developing Mac drivers?

In Mac gaming very few developers were interested in the Mac market and Apple ships millions of non-Mac Pros quarterly.

This is a well beaten topic. It's not about return on investment. As such, it's not worth apple's time. It *is* worth apple's time as a halo product and a product they themselves could use for development. Developing drivers is TRIVIAL if they just spend a little time and money (drivers exist on Linux for goodness sakes), and yes, AMD would do it, I believe someone posted a statement in another thread stating AMD would be willing.

Agreed about gaming, but that partly goes with how awful apple has treated gaming, historically, on Macs. Look at this way. No one wants to touch apple at the enterprise level because of their history of being awful/hostile to the enterprise market, but other companies have sprung up that do provide enterprise services for apple equipment and are doing great (one of them even went public!). So a lot of that is on apple for just awful mismanagement of the space.

 

[Longplays]

It *is* worth apple's time as a halo product and a product they themselves could use for development.

PowerPC called and they want their "halo tower" back.

As late as 2006 almost everyone buying any personal computer will buy a laptop over a desktop without PCIe slots much less those with PCIe slots.

Because it is good enough for their use case.

That's one reason why replacement cycle of 3 years back in the 80s & 90s lengthened to 5-6 years for the past 10 years.

Very few born after the 1990s would ever see a tower Mac as a "halo" product.

Developing drivers is TRIVIAL if they just spend a little time and money (drivers exist on Linux for goodness sakes), and yes, AMD would do it, I believe someone posted a statement in another thread stating AMD would be willing.

AMD or Nvidia aren't interested to service that few customers. They're better off making Emojis.

Agreed about gaming, but that partly goes with how awful apple has treated gaming, historically, on Macs.

Apple couldn't leverage their business model and tech on PC gaming. It would require them to do modularization and they'll lose money.

That is OK because their casual gaming business through their App Store makes more money than Microsoft, Nintendo and Sony's game businesses combined.

 

[goMac]

Developing drivers is TRIVIAL if they just spend a little time and money (drivers exist on Linux for goodness sakes)

There are no Metal drivers for Linux. Linux driver work does not re-apply to Mac.

I'd guess the macOS 6000 series drivers are a good starting point. But Linux has nothing to do with anything for macOS drivers.

 

[ZombiePhysicist]

PowerPC called and they want their "halo tower" back.

As late as 2006 almost everyone buying any personal computer will buy a laptop over a desktop without PCIe slots much less those with PCIe slots.

Because it is good enough for their use case.

That's one reason why replacement cycle of 3 years back in the 80s & 90s lengthened to 5-6 years for the past 10 years.

Very few born after the 1990s would ever see a tower Mac as a "halo" product.


AMD or Nvidia aren't interested to service that few customers. They're better off making Emojis.

Apple couldn't leverage their business model and tech on PC gaming. It would require them to do modularization and they'll lose money.

That is OK because their casual gaming business through their App Store makes more money than Microsoft, Nintendo and Sony's game businesses combined.

It called and you didnt answer. You still dont get how halos work. Rest of your analysis is of equal quality. Dismissed.

 

[ZombiePhysicist]

There are no Metal drivers for Linux. Linux driver work does not re-apply to Mac.

I'd guess the macOS 6000 series drivers are a good starting point. But Linux has nothing to do with anything for macOS drivers.

It does if they get the new Mac Pro to work with 3rd party graphics cards. At the very least in proving there is no technical problem stopping it, and just a will to do it.

 

[Boil]

All right, let's get this thing going...!

Rampant speculation towards the second generation ASi Mac Pro...

Back to the whole asymmetrical SoC idea...!

M3 Ultra / M3 Extreme SoCs, built on a 3nm process, with A17-based cores...

"Standard" SoC (M3 Max)

• 16-core CPU (12P/4E)
• 48-core GPU (w/hardware ray-tracing)
• 16-core Neural Engine

"GPU-specific" SoC (ASi GPU)

• 64-core GPU
• 16-core Neural Engine

Combine one of each of the above for a M3 Ultra that has a higher GPU-to-CPU core ratio...

Combine two of those "GPU-heavy" M3 Ultra SoCs for a M3 Extreme SoC...

Offer the "GPU-heavy" variant in the Mn Ultra Mac Studio as well...

Also use the "GPU-specific" SoCs for ASi (GP)GPUs, up to four per add-in card; if Apple cannot derive a way to make said add-in cards part of the overall "core count/RAM pool", then they are still there as targets for compute/render jobs...

Ideally, Apple derives a way to have a SuperDuperUltraHighSpeed backplane, allowing (up to four) Mn Ultra ("GPU-heavy variants) daughtercards & (up to four) ASi (GP)GPU add-in cards to work together as one giant pool of CPU/GPU/Neural Engine cores & UMA RAM...

64-core CPU (48P/16E)
1,072-core GPU (w/hardware ray-tracing)
384-core Neural Engine
6TB LPDDR5X RAM

Stupid expensive, yes...

Hella fast, also yes...

;^p

 

[Longplays]

M3 Ultra / M3 Extreme SoCs, built on a 3nm process, with A17-based cores...

Q1 2025 release assuming 19.5 month refresh of M1 > M2 > M3 is maintained.

This can be shortened when Apple syncornizes it to iPhone chip's 12 month refresh cycle.

$1299 Mac mini M3 Pro 3nm will likely equal performance of a $1999 Mac Studio M2 Max 5nm.

 

[Boil]

Q1 2025 release assuming 19.5 month refresh of M1 > M2 > M3 is maintained.

This can be shortened when Apple synchronizes it to iPhone chips.

So maybe gen2 ASi Mac Pro is M4/A19 based...?

$1299 Mac mini M3 Pro 3nm will likely equal performance of a $1999 Mac Studio M2 Max 5nm.

That would be sweet, as it would bode well for performance gains going up the stack...?!?

 

[Longplays]

So maybe gen2 ASi Mac Pro is M4/A19 based...?

I would be more concerned about the waiting time. You willing to wait to 2026 or H1 2027?

That would be sweet, as it would bode well for performance gains going up the stack...?!?

Likely.

Apple has 1st right of refusal for any future die shrink.

We're 3 months away from the next iPhone with a 3nm chip.

 

[jmho]

At this point I'm just expecting to be disappointed / annoyed:

That means either:

a) M3 is a disappointment. No h/w raytracing, no real new features. Apple takes the efficiency gains and makes everything thinner and lighter. The Mac Pro 9,1 is just the 8,1 but with an M3 Ultra.

or

b) Apple pulls an iPad 4 and releases the M3 Ultra 9,1 in 6 months time and it is so good it makes the M1 and M2 machines obsolete. I sit there looking at the M2 Ultra Studio I just reluctantly ordered and cry into my Magic Keyboard.

 

[ZombiePhysicist]

Finally honest analysis starting.

New Apple Silicon Mac Pro badly misses the mark

The Apple Silicon Mac Pro is here two and a half years after the shift from Intel began, but it looks like the company only did it to say that they had.

appleinsider.com

We need to ridicule this 8,1 machine so it’s another trashcan failure. Apple needs to hear our disappointment. It’s not a Mac Pro. It’s a Mac Studio pro. It’s a Mac ‘Doh! If we complain enough the next Mac Pro might actually be a Mac Pro.

We need to demand it support 3rd party GPUs. It have upgradable ECC ram. And upgradable processors.

 

[ZombiePhysicist]

And to that end, let your voice be heard...

Sign the Petition

Force Apple to release AMD drivers for the 2023 Mac Pro

www.change.org

 

[Boil]

I am guilty myself of referencing MacPro8,1, but is it not actually Mac14,8...?

So the Gen2 ASi Mac Pro would be Mac15,x...? Imma go with Mac15,7 just for luck...!

So we have a Gen2 ASi Mac Pro, but Apple decides to lean in extra hard on the whole non-upgradable thing, offering a custom PCIe expansion chassis (partnered with Sonnet) that hooks up to multiple TB5 ports for customers needing PCIe slots...

Still no add-in GPUs, ASi or third-party...

Just the new Gen2 Mac Pro...

ASi Mac Pro Cube

• WWDC 2024 preview
• Fall 2024 release
• 3nm process
• A17 cores
• M3 Extreme SoC
• 64-core CPU (48P/16E)
• 192-core GPU (w/hardware ray-tracing)
• 64-core Neural Engine
• 512GB ECC LPDDR5X RAM
• 32TB SSD
• Eight Thunderbolt 5 ports
• Two USB-A ports
• Two 10Gb Ethernet ports
• Two HDMI ports
• 3.5mm/optical audio output jack
• 7.9" x 7.9" x 7.9" chassis
• 700W PSU
• Silver or Space Gray
• US$19999

You get what you get and you don't throw a fit...!

;^p

 

[ZombiePhysicist]

I'm really hoping they increase the SSD to 16TB. 8TB is kind of weak sauce for a desktop.

 

[avro707]

The tech press has a similar issue to the video gaming press, in that sense. Being critical can easily cost you access, and potential career paths.

This I have seen elsewhere (not related to this) they do anything to get inside access to new stuff for their news sites, blogs or YouTube channels.

It’s infuriating for those who also have early access but for genuine feedback and testing purposes. These folk waltz in, then disappear not long after. I won’t say anything further than that.

 

[deconstruct60]

So we have a Gen2 ASi Mac Pro, but Apple decides to lean in extra hard on the whole non-upgradable thing, offering a custom PCIe expansion chassis (partnered with Sonnet) that hooks up to multiple TB5 ports for customers needing PCIe slots...




;^p

Really? 2017-mid 2019 ... rampant notions of 'lego' Mac Pro with Thunderbolt backplane ... blah blah blah.
.. 2019 Mac Pro .. no such thing.

2020-mid 2023 the 'tube' is coming back. Thunderbolt solves all PCI problems . Apple is going to use TB like ketchup.
... 2023 Mac Pro ... no such thing.


So what another year or two of pushing this rock up the hill again?

Thunderbolt is a useful tool for lots of systems. Thunderbolt 5 will be an even more useful tool. But to flip that into a "if all you have is a hammer , everything looks like a nail" solution is cuckoo for cocoa puffs.

Thunderbolt 5 goes off into dealing with DisplayPort v2.1 issues just about as much as it deals with PCI-e transport ones. TBv5 is not trying to eradicate internal PCI-e expansion.


If you have multiple TB ports being connected to the same peripheral , it is highly likely you can creating some Rube Goldberg solution to what is really a band-aid for bad system design.

The foundation of TB is being run by USB-IF now. It is a large committee of member who have significantly differing agendas. It is just going to move slower than when Intel effectively solely ran the show and Apple was the main , very eager, adopter. Thunderbolt is going to constantly trail PCI-e progress by several years.

That sizable trailing gap is why having one system that isn't forced to follow so far back is useful to have in the line up.

The whole "one and only one" internal drive notion for a workstation was dubious in 2013 and has just gotten even more dubious since. Apple doesn't have to sell the optional drives at all , but to pour TB on like ketchup as a substitute solution just do not work. Apple even admitted that.

2023 bleeding edge NAND chips have huge number of layers , but NAND is running into the same "Moore's Law is slowing" headwinds as the rest of logic. And at the high end of the spectrum data just keeps getting bigger. Storing absolutely everything on a solely single drive is nuts as capacity needs grow.

 

[Longplays]

I am guilty myself of referencing MacPro8,1, but is it not actually Mac14,8...?

So the Gen2 ASi Mac Pro would be Mac15,x...? Imma go with Mac15,7 just for luck...!

So we have a Gen2 ASi Mac Pro, but Apple decides to lean in extra hard on the whole non-upgradable thing, offering a custom PCIe expansion chassis (partnered with Sonnet) that hooks up to multiple TB5 ports for customers needing PCIe slots...

Still no add-in GPUs, ASi or third-party...

Just the new Gen2 Mac Pro...

ASi Mac Pro Cube

• WWDC 2024 preview
• Fall 2024 release
• 3nm process
• A17 cores
• M3 Extreme SoC
• 64-core CPU (48P/16E)
• 192-core GPU (w/hardware ray-tracing)
• 64-core Neural Engine
• 512GB ECC LPDDR5X RAM
• 32TB SSD
• Eight Thunderbolt 5 ports
• Two USB-A ports
• Two 10Gb Ethernet ports
• Two HDMI ports
• 3.5mm/optical audio output jack
• 7.9" x 7.9" x 7.9" chassis
• 700W PSU
• Silver or Space Gray
• US$19999

You get what you get and you don't throw a fit...!

;^p

I rather buy $AAPL

I dont want a gaming machine.

 

[goMac]

2020-mid 2023 the 'tube' is coming back. Thunderbolt solves all PCI problems . Apple is going to use TB like ketchup.
... 2023 Mac Pro ... no such thing.

A certain someone has been real quiet recently.

 

[deconstruct60]

A certain someone has been real quiet recently.

If Apple put the Mac Pro on a mostly regular upgrade cycle ( 2 year +/- 6 months ) . Not fixed in time date every year, but enough regularity that would know there is an solid expectation of an incremental upgrade coming. ( Doing the Mac Pro on an iPhone monomaniacal yearly update schedule likely isn't economical at all. In 2006-2010 there were huge non-Mac revenue flows that covered up how relatively low Mac Pro volume's problems. No 'sugar daddy' to cover up expensiveness anymore. )

It is the length of the information vacuum that drives these threads way , way , way out deep into the swamp.

It is still hard to see what the iteration time gap is going to be for M-series long term. Lots of folks seem to want it to get set for 12 months. Not sure Apple could keep that pace for more expensive SoCs. But something less than 3 years would help a lot .

i think some of these folks presenting the "sucks more" options is so that when Apple rolls out the Mac Pro, and it isn't a completely Windows PC Intel Clone, then you can put a "sucks less than swamp option Apple could have done" label on it. Perhaps that is intended to be helpful , but i think it does more damage than good (or at least tied).

Apple execs sitting on stage saying "yeah we got key customers who have non GPU PCI-e cards and that's worth doing" kind of shines too much light to make the dark swamp Mac Pro look viable. It is when Apple goes dark for a long time ... it is swampy mushroom time.

This M2 Ultra probably isn't all of what they wanted to do so, it is doubtful we will see another 4 year Rip Van Winkle exercise. The next iteration time gap probably won't be indicative of the natural long term iteration state either.

 

[Longplays]

If Apple put the Mac Pro on a mostly regular upgrade cycle ( 2 year +/- 6 months ) . Not fixed in time date every year, but enough regularity that would know there is an solid expectation of an incremental upgrade coming. ( Doing the Mac Pro on an iPhone monomaniacal yearly update schedule likely isn't economical at all. In 2006-2010 there were huge non-Mac revenue flows that covered up how relatively low Mac Pro volume's problems. No 'sugar daddy' to cover up expensiveness anymore. )

It is the length of the information vacuum that drives these threads way , way , way out deep into the swamp.

It is still hard to see what the iteration time gap is going to be for M-series long term. Lots of folks seem to want it to get set for 12 months. Not sure Apple could keep that pace for more expensive SoCs. But something less than 3 years would help a lot .

i think some of these folks presenting the "sucks more" options is so that when Apple rolls out the Mac Pro, and it isn't a completely Windows PC Intel Clone, then you can put a "sucks less than swamp option Apple could have done" label on it. Perhaps that is intended to be helpful , but i think it does more damage than good (or at least tied).

Apple execs sitting on stage saying "yeah we got key customers who have non GPU PCI-e cards and that's worth doing" kind of shines too much light to make the dark swamp Mac Pro look viable. It is when Apple goes dark for a long time ... it is swampy mushroom time.

This M2 Ultra probably isn't all of what they wanted to do so, it is doubtful we will see another 4 year Rip Van Winkle exercise. The next iteration time gap probably won't be indicative of the natural long term iteration state either.

M1 Ultra > M2 Ultra took 15 months.

So it isn't unreasonable that by Q1 2025 a M3 Ultra/Extreme be out by then.

The 19.5 month cycle between M1 > M2 likely has to do with COVID impacting supply chain.

It is likely Apple decided it best that they prioritize their money maker iPhone be least impacted by China's lockdowns.

If the rumors are true about M3 coming out 1-2 months after September iPhone event rather than Q1 2024 then that could be the start of Mac chips aligning with iPhone's 12 month refresh cycle.

This would materially help Mac chips align their CPU, GPU, Neural Engine, etc with the latest iPhones.

With how Apple sets up Mac chips it wouldnt cost them all that much to start collecting binned and unbinned M3 Ultra/Extreme from M3 Maxs.

Mac Studio & Mac Pro cases are already standardized. I/O internal/external will hardly change in terms of mounting points and machined ports position the foreseeable future. So putting in the new SoC is like plugging in a higher effienct LED lamp with better lumens per Watt.

 

[deconstruct60]

I'm really hoping they increase the SSD to 16TB. 8TB is kind of weak sauce for a desktop.

Apple gimped the M2 generation laptops to fewer bigger NANDs chips ( and tossed bandwidth) to hit the same price points. I wouldn't hold my breath.

There is a NAND chip glut right now, but until $/GB prices look like they permanently are going lower , Apple is not likely to move. ( Apple did make the MBA 13" and 15" prices a bit better. But SSD capacity didn't go up).