Mac Pro 2023 = Another Mac Pro 2013 crisis and failure.

[bcortens]

Assuming that were done they'd ship a potential 80,000 MPX modules over a period of 4 years assuming the SoC has not already made the modules redundant.

Maybe - but it would allow for compute density that isn't currently possible on Mac Pro. IIRC there are 3d graphics studios that do movie VFX that use AMD Threadripper 64 core CPUs and having the ability to increase compute density in a single case with a good solution for distributed computing work would allow them to try and get this market back. Scientific big data computing is another market that could benefit from the increased compute density. It wouldn't be a huge engineering lift since the M2 Ultra already exists, cluster computing is pretty well figured out already, just a matter of changing the shape of the M2 Ultra PCB that they put into the Mac Studio and figuring out how to do power delivery and if they want to do something like ethernet or PCIE for the cluster access.

 

[Longplays]

Yeah you're right

It's also a much bigger chassis
A much bigger power supply, significally more than the Studio
A much bigger cooling system
And a lot more IO, including two 10gb ethernet ports

The size, power supply, and cooling system will probably yield much higher performance of that M2 Ultra than on the Mac Studio since you can dump more power in the SoC. But we will have to see in benchmarks. If in tests the performance gain is minor compared to the Studio, then yeah it'll be an L.

Not to mention it is at most 14.51kg heavier to ship in a far larger box.

How many Mac Studios can fit on a shipping pallet vs a Mac Pro?

 

[NT1440]

For years Apple sold a power tower: G4, G4, G5 and Mac Pro. These were solid machines that were priced appropriately. I had several of these with a typical price tag of $3K - $5K. My 2013 was around $7K and 2019… I’m embarrassed to say.

Apple’s focus on the pro market made the Mac the standard for video editing and graphic design. Now, they are killing off what makes the Mac great. Yes, most sales are for laptops and the iMac, but what is missing from the conversation is that if a company does high end work on a MacPro they probably buy several other non-MacPro Apple machines. Loose the pro business and the entire computer lineup suffers.

I understand the limitations of current Apple Silicone; all I’m suggesting is that Apple acknowledge these limitations and price the MacPro accordingly instead of arbitrarily raising the price, alienating, and turning off many customers who would spend $500-$1,000 more for the MacPro over the Studio.

If you were doing pro level video editing why in the world would you need a Pro over the Studio? If your workflow requires some special card (not a graphics card) then you have it today with the Mac Pro. What’s missing here?

 

[bcortens]

Yeah you're right

It's also a much bigger chassis
A much bigger power supply, significally more than the Studio
A much bigger cooling system
And a lot more IO, including two 10gb ethernet ports

The size, power supply, and cooling system will probably yield much higher performance of that M2 Ultra than on the Mac Studio since you can dump more power in the SoC. But we will have to see in benchmarks. If in tests the performance gain is minor compared to the Studio, then yeah it'll be an L.

It's not a bigger cooling system, it has 1 (maybe 2) more fans than the Mac Studio... it has maybe a bit more metal than the cooler in the studio. The studio is not some cheaply made case with minimal engineering effort. The studio chassis is itself expensive.
The extra 10GB ethernet port, and power supply and extra metal needed in the chassis do not a $3000 premium make, in no sensible world is that the price premium. It is Apple being apple and gouging on price for components again. Even at a $1000 premium I expect the margin on the upgrade would be close to 75%, the margin on $3000 is probably close to 350%...

 

[AlphaCentauri]

I already addressed your concerns in another post. In brief; losing the Pro business results in customers moving the entire process away from Mac.

The point is they’ve already moved to Win/Linux in 2013 and 2019 and those who stayed on Mac and want dGPUs and expandable memory are so few in numbers that Apple doesn’t really care about them.

 

[avkills]

For those moving their pre-PCIe 4.0 cards to the 2023 Mac Pro... would they even notice?

If it is a high end video capture card or high speed SSD storage, yes if the PCIe on the 2023 Mac Pro is actually being fed using thunderbolt 4 speeds. There will be a huge difference and some cards will not even work.

But we will not know until more reviews and testing is done.

 

[avkills]

Mac Pro 2023 has at least 64 PCIe gen 4 lanes. How many come directly from the CPU? How many come from the PCIe gen 4 switch (if there is one)? How much do PCIe gen 4 switches cost?

Mac Pro 2021 had 64 lanes from the CPU. 32 went directly to slots. Other 32 went to a PCIe gen 3 switch with 64 downstream lanes. So that's 96 lanes total.

Here's the technology overview for the Mac Pro 2021 https://www.apple.com/ua/mac-pro/pdf/Mac_Pro_White_Paper_Aug_2021.pdf
Will Apple make a technology overview for the 2023 Mac Pro? Or is the 2023 Mac Pro too simple for that?

Don't you mean the 2019 Mac Pro. There is no such thing as a 2021 Mac Pro.

And prove to us that the PCIe slots on the 2023 Mac Pro are actually full bandwidth slots; oh right you can't. Saying it is so because Apple marketing says it is is not actual proof.

 

[Longplays]

Maybe - but it would allow for compute density that isn't currently possible on Mac Pro. IIRC there are 3d graphics studios that do movie VFX that use AMD Threadripper 64 core CPUs and having the ability to increase compute density in a single case with a good solution for distributed computing work would allow them to try and get this market back.

Which Mac Pro are you pertaining to? The 2019, that is well documented, understood and left unchanged for 4 years or the 2023?

I'd hold off judgement of the 2023 until benchmarks and lengthy reviews about it is published.

Odds are very likely that either M2 Ultra SKUs will outperform any of the 2019 SKUs.

Though very very unlikely there will be one esoteric benchmark that will show a less than best result.

But then again if almost all benchmarks outperform the 2019 model and even other workstations within the same price point and use case then who should complain?

I read of one MR user complaining that they cannot use their afterburner cards on the 2023. During Apple's livestream it was claimed M2 Ultra has built-in afterburners of many multitudes. So I do not understand why anyone's complaining about that.

How about GPU performance? Everyone's demanding Nvidia & AMD GPU support even when no 3rd party GPU benchmarks are published.

Scientific big data computing is another market that could benefit from the increased compute density. It wouldn't be a huge engineering lift since the M2 Ultra already exists, cluster computing is pretty well figured out already, just a matter of changing the shape of the M2 Ultra PCB that they put into the Mac Studio and figuring out how to do power delivery and if they want to do something like ethernet or PCIE for the cluster access.

Ultra niche of the ultra niche on the Mac.

Based on the more than 192GB memory requirements that some big data MR users have Apple would be better off not designing for them.

With Intel selling to them does not cost Apple extra. This was a + for Apple when they went with Intel as they enjoyed the economies of scale of the worldwide desktop workstation market.

Apple's leveraging the economies scale of iPhone chips when applied to Mac chips. IIRC worldwide shipments iPhone + iPad chips = PC chips.

If it were not for the iPhone Apple would still be stuck on 14nm Intel chips in 2024.

 

[bcortens]

Which Mac Pro are you pertaining to? The 2019, that is well documented, understood and left unchanged for 4 years or the 2023?

I'm thinking about the new one, the 2019 got new GPUs as they became available increasing compute density over time. (I believe/hope new one will be updated at a reasonable 18-24 month schedule to keep it up to date with the rest of the M series).

I'd hold off judgement until benchmarks and lengthy reviews about it is published.

Odds are very likely that either M2 Ultra SKUs will outperform any of the 2019 SKUs.

Not quite, Apples own numbers only compare to a single WX6900 while there exist both the WX6800 Duo AND you can put two WX6900s in a single chassis. I believe for GPU compute the new machine might fall slightly behind a fully maxed out 2019 model. (I don't think this use case is super relevant but it is what it is)

I read of one MR user complaining that they cannot use their afterburner cards on the 2023. During Apple's livestream it was claimed M2 Ultra has built-in afterburners of many multitudes. So I do not understand why anyone's complaining about that.

Yeah that is a weird complaint. The M2 Ultra is going to be far better than the afterburner cards at video encode.

How about GPU performance? Everyone's demanding Nvidia & AMD GPU support even when no 3rd party GPU benchmarks are published.

Ultra niche of the ultra niche on the Mac.

Based on the more than 192GB memory requirements that some big data MR users have Apple would be better off not designing for them.

With Intel selling to them does not cost Apple extra. This was a + for Apple when they went with Intel as they enjoyed the economies of scale of the worldwide desktop workstation market.

Apple's leveraging the economies scale of iPhone chips when applied to Mac chips. IIRC worldwide shipments iPhone + iPad chips = PC chips.

If it were not for the iPhone Apple would still be stuck on 14nm Intel chips in 2024.

I know ultra niche on the Mac, but I think that giving up the professional workstation market is potentially a mistake, not because of the numbers but because of the influence. Professional workstation machines like this mean that high end 3d rendering software gets supported. If Apple can get a win in any VFX house then it would help them get into this space and help push more support for the super high end of 3d workflows.
I also thought maybe they could build these clusters and then allow anyone to rent access to one for $250 a month or something, let pro users who don't have the budget to buy one rent one for a time or remote into one from their iPad to do high end work on the go... They could also start dogfooding the SoCs to their own Siri teams to build the next gen Siri LLM models.

 

[avkills]

If it were not for the iPhone Apple would still be stuck on 14nm Intel chips in 2024.

Except that the current crop of Intel chips in 2023 are not 14nm; and support PCIe 5.0; so what's your point?

This type of complacent thinking is just how Intel got into trouble. And what happened, Intel has a new CEO and they are starting to turn things around.

Apple making the first iPhone CPU was out of *a must do* because a CPU capable of doing what they wanted to do simply did not exist at that time.

 

[Longplays]

If it is a high end video capture card or high speed SSD storage, yes if the PCIe on the 2023 Mac Pro is actually being fed using thunderbolt 4 speeds. There will be a huge difference and some cards will not even work.

But we will not know until more reviews and testing is done.

What I meant was many here who are ultra concerned and emphasing e-waste will likely move their PCIe 1.0, 2.0 and 3.0 cards to PCIe 4.0 slots.

So if PCIe 4.0 is "gimped" would they even notice it? The PCIe cards they've saved from eBay and garage sales will carry over since 2003.

 

[avkills]

What I meant was many here who are ultra concerned and emphasing e-waste will likely move their PCIe 1.0, 2.0 and 3.0 cards to PCIe 4.0 slots.

So if PCIe 4.0 is "gimped" would they even notice it?

Yes if those PCIe4.0 slots are being fed at Thunderbolt 4 bandwidth; which is at best PCIe 3.0 4x. Any card that requires more than a 3.0 4x connection speed will be gimped; which is pretty much every single video capture card and high speed SSD storage card.

 

[Rychiar]

yeah buying the studio ultra just makes way more sense. There’s nothing about this that appeals to me and that price is ******* insane

 

[Longplays]

Except that the current crop of Intel chips in 2023 are not 14nm; and support PCIe 5.0; so what's your point?

Intel stayed stuck on 14nm from 2014-2020. What made them materially move in 2020 was Apple moving to 5nm Apple Silicon that November.

This type of complacent thinking is just how Intel got into trouble. And what happened, Intel has a new CEO and they are starting to turn things around.

Apple leaving likely forced Intel to improve.

Apple making the first iPhone CPU was out of *a must do* because a CPU capable of doing what they wanted to do simply did not exist at that time.

2007 iPhone had an outsourced chip that was designed by Samsung for DVD players. Apple did eventually create an in-house custom iPhone chip to get the results they wanted. As to not make this off topic I will not address any further iPhone chip concerns.

 

[Longplays]

Yes if those PCIe4.0 slots are being fed at Thunderbolt 4 bandwidth; which is at best PCIe 3.0 4x. Any card that requires more than a 3.0 4x connection speed will be gimped; which is pretty much every single video capture card and high speed SSD storage card.

Are you the Apple employee with insider knowledge of the 2023 being gimped? 🤣

 

[avkills]

Apple leaving likely forced Intel to improve.

I seriously doubt that. Intel isn't stupid enough to think Apple is going to be back anytime soon. AMD is who Intel needs to contend with; and nVidia since Intel decided to join the GPU war.

 

[avkills]

Are you the Apple employee with insider knowledge of the 2023 being gimped? 🤣

🤣

No, I bash them too much to be considered for employment. Lol.

I never said it WAS gimped; but right now there is no proof to say it isn't because of what is currently known about M2 silicon. Perhaps there is a yet unknown portion on the Ultra silicon that has tons of PCIe.

 

[0339327]

If you were doing pro level video editing why in the world would you need a Pro over the Studio? If your workflow requires some special card (not a graphics card) then you have it today with the Mac Pro. What’s missing here?

Thunderbolt ports and fiber IO

Yes 8 TB is not enough. My workflow calls for at least 9.

Edit: clarification that TB here is in reference to Thunderbolt, not terabyte.

 

[0339327]

The point is they’ve already moved to Win/Linux in 2013 and 2019 and those who stayed on Mac and want dGPUs and expandable memory are so few in numbers that Apple doesn’t really care about them.

I didn’t because in those situations I was able to make the machines work for me. My workflow now relies on TB ports for RAID enclosures. Lots of them.

The studio is a great little machine - but it’s not a pro workstation.

 

[bobcomer]

Intel stayed stuck on 14nm from 2014-2020. What made them materially move in 2020 was Apple moving to 5nm Apple Silicon that November.

More likely it was TSMC and Samsung that forced the change in intel.

Apple leaving likely forced Intel to improve.

No way, too small of a market. Like I said, TSMC and Samsung. (and their own BoD)

 

[Longplays]

More likely it was TSMC and Samsung that forced the change in intel.


No way, too small of a market. Like I said, TSMC and Samsung. (and their own BoD)

Partially but the bad press of Apple leaving and how much better 5nm Apple Silicon is/was forced the Intel capex.

Intel claimed technical limitations why they stuck to 14nm from 2014-2020 but if you read between the lines it is a way to reduce cost considering all the PC OEMs buy chips from them since 2006.

So what is the incentive for monopoly to improve.

Qualcomm buying NUVIA to make ARM laptops for Windows 11 heightens the need to improve.

2006-2020 Intel is pure gravy profitability even when iPhone and Android ate into the PC market since 2007 & 2008 respectively.

 

[AlphaCentauri]

I didn’t because in those situations I was able to make the machines work for me. My workflow now relies on TB ports for RAID enclosures. Lots of them.

The studio is a great little machine - but it’s not a pro workstation.

No, it’s not. I’ve never said it was. The thing is, the Studio covers the needs of about 90% of Mac professionals, so it doesn’t need to be. The new Mac Pro caters for the needs of about 9% out of remaing 10% of Mac professionals due to PCI-E ports.

This leaves us with about 1% of Mac users who really NEED expandable tower with dGPU - Apple doesn’t care about them, unfortunately.

 

[sunny5]

We'll have to wait and see. Best people to ask are those who work in 3D Studios and heavy music production... Can't wait to see opinions.

I'm not really the target audience for this as the M2 and M2 Pro are more than enough of for what I do.

It's already dead before it comes out when Mac Pro does not support powerful GPU. Beside Apple Silicon itself still slower than Nvidia GPU. You probably never used RTX 3090 or 4090.

 

[Darren.h]

THEY SHOULD HAVE HAD UPGRADABLE REMOVABLE SOC OR CPU IN THE NEW MAC PRO.

THEN MAYBE WORTH 7 GRAND

USE TO START AT 6 GRAND FOR THE INTEL MAC PRO. WENT UP 1000 DOLARS.

 

[AlphaCentauri]

THEY SHOULD HAVE HAD UPGRADABLE REMOVABLE SOC OR CPU IN THE NEW MAC PRO.

THEN MAYBE WORTH 7 GRAND

USE TO START AT 6 GRAND FOR THE INTEL MAC PRO. WENT UP 1000 DOLARS.

Yes we know and mostly agree. No need to shout 🤣