MP All Models 2023 Mac Pro! [MERGED]

[AndreeOnline]

and that was useing the high speed ssd as sawp to mask that.

Nothing is masked. The "consensus" on what RAM is and how much you need has been built up through decades of using spinning, slow HDD, meh interconnect and 'not quite enough RAM' (back then).

Systems have changed and moved on. All the sub-parts are faster. App memory requirements have increased, but the average amount of installed RAM has increased more.
To this you can add how Apple Silicon uses and addresses RAM, and you have a new, modern system that needs to be evaluated for what it is.

If fast SSDs and a moderate amount of RAM performs better with better memory management than slower disks and heaps of RAM with "traditional" memory management.... well, it just performs better—regardless of what you call it.

More RAM has never made, and will never make a computer faster. But too little can make it slower. This is not a contradiction by the way.

Not saying some use cases won't benefit from large amounts of RAM (already said this previously), but it might be a good idea to evaluate each architecture on its own merits.

 

[Longplays]

I wonder how we’d be reacting to the 2023 Mac Pro if it had launched with an M2 Extreme (2 x Ultra) chip?

It would still not be upgradable in terms of CPU, GPU and RAM, but it would presumably offer performance beyond any other Mac.

I wonder how much that would have changed the conversation?

Non-binned M2 Extreme with more GPU cores would be like this

Chips	M2 Extreme
CPU	48-Core
High-performance	32x
High-efficiency	16x
GPU	152-Core
Neural Engine	64-Core
Transistors	268 billion
Max unified memory	384GB
Memory bandwidth	1.6TB/s

Less people would complain about the doubling of 192GB unified memory.

Instead of complaining about the price of how expensive 192GB was they'd be more alarmed by how much 384GB is.

Fraction of the people would complain about the raw performance of the M2 Extreme in a vacuum relative to the cost of a near identically spec'd & TCO'd Intel/AMD/Nvidia workstation.

For those wondering it would likely start at $11k for the binned SKU. Add another $2k for the non-binned SKU with more GPU cores?

We'd all be amazed by how honkingly big it is. For those wondering the SoC package would likely be around the size of a Mac Studio logicbaord.

It would be too big to fit into & require more PSU than a Mac Studio can deliver but would perfectly be placed into and powered by a Mac Pro.

 

[jmho]

I wonder how we’d be reacting to the 2023 Mac Pro if it had launched with an M2 Extreme (2 x Ultra) chip?

It would still not be upgradable in terms of CPU, GPU and RAM, but it would presumably offer performance beyond any other Mac.

I wonder how much that would have changed the conversation?

I honestly think given how disappointing the overall package is with such limited expansion options, the only thing an M2 Extreme would do is make the Studio look disappointing.

Currently the Mac Pro acts as an incredible "pricing anchor" to make the Mac Studio look like a bargain where you can get "Mac Pro" performance in a tiny box for $3k less.

It wouldn't surprise me if one of the reasons they canned the M2 Extreme was that they realised that not only would most people not buy it, but its existence would turn the Mac Studio from a top-of-the-range product into an expensive mid-range option.

 

[mattspace]

If fast SSDs and a moderate amount of RAM performs better with better memory management than slower disks and heaps of RAM with "traditional" memory management.... well, it just performs better—regardless of what you call it.

Exceeeeeept that if you keep an eye on what developers and users are reporting - out of memory errors are becoming a common issue these days. Apple in fact does not appear to have better memory management than they used to.

It would be reasonable to suspect that with iOS as the dominant platform with Apple, and macOS becoming a UI-mod derivative of it, that virtual memory is eventually going to go away on the Mac.

https://mjtsai.com/blog/2023/01/26/what-happened-to-virtual-memory/

 

[Longplays]

I honestly think given how disappointing the overall package is with such limited expansion options, the only thing an M2 Extreme would do is make the Studio look disappointing.

M2 Extreme wouldn't fit inside a Mac Studio. If ever it did it would require a more robust active HSF & PSU.

Currently the Mac Pro acts as an incredible "pricing anchor" to make the Mac Studio look like a bargain where you can get "Mac Pro" performance in a tiny box for $3k less.

The benchmark results of the 2019 Mac Pro Intel vs 2021 Mac Studio M1 Max made that value proposition very apparent even without knowing the price, I/O and PSU difference.

80% or more Mac desktop workstation users do not want to pay for I/Os & PSUs they would never ever use. Much less subsidize the economies of scale for others.

It wouldn't surprise me if one of the reasons they canned the M2 Extreme was that they realised that not only would most people not buy it, but its existence would turn the Mac Studio from a top-of-the-range product into an expensive mid-range option.

It is likely the M2 Extreme yielded badly. Similarly to last year's iPhone chip that allegedly had Ray Tracing GPU cores that drew too much power.

They may try again in Q1 2025 with the M3 Extreme.

 

[Longplays]

Here is a great analysis from an actual TV/Film/Music professional.

This the typical use case of the Mac Pro.

3.5 years ago he was one of the few who deployed Apple's 7,1 Mac Pro Rack and used up all the PCIe slots for his professional audio workflow.

 

[theluggage]

I think too many people focus on the name as if it is a solid delineation of professionals and everyone else. When it comes to Apple's naming pro merely means more capable than our other products.

Sure, but from 2006- Apple's range included a "Mac Pro" that - in terms of power, expandability and price - took up more or less where the iMac left off (even the Trashcan sort-of met those criteria - it just bet the farm on external Thunderbolt expansion). The fully tricked-out versions were well into "if you need to ask the price you can't afford it" territory, but the entry level Mac Pros were (a) still within the means of a prosumer/enthusiast and (b) actually worth having and better than the top-end iMac/Mini.

Then, in 2019, wham the entry-level for Mac Pro jumps from around $3000 to $6000 for a base spec that was barely better than a $3k iMac (which included about $1k worth of display). That's when all this "No true professional...", "If you don't need a 1TB working set you don't need a Mac Pro" argument really started. Fine, for a 50k/1.5TB/Quad-high-end-GPU system with Afterburners up the wazoo - but it was a complete dumping of the "something a bit better than an iMac" market.

 

[impulse462]

Well, for 1k more than the 2019 Mac Pro you get:
- 24 CPU cores vs 8 (massive difference here, should be at least 3x faster. Even the 2019 28 core won't match the M2Ultra CPU)
- M2Ultra GPU is much more powerful than the base GPU in the 2019. We have yet to see if M2Ultra can match the high end AMD GPU, but if the M2Ultra is approx 2x that of M2Max it should be close.
- 64GB RAM vs 32GB RAM
So for 1k more, you are getting a whole lot more and to configure the 2019 to similar specs would more than double the price of the base 2023.

What an incredibly disingenuous post this is. You're comparing a brand new SoC to a CPU thats 4 years old. wow its so much faster! who would have thought?

Congratulations, an integrated GPU is more powerful than a base GPU from 2019. But I can't add my OWN GPU, something a machine of this caliber should be able to do. The fact that there are PCIe slots is even worse since its basically mocking anyone who wants to replace their GPU.

"So for 1k more, you are getting a whole lot more"
Are you actually serious here?

For 1k more I'm also getting the freedom to not be able to boot linux or windows and run GPUs of my choosing severely limiting what this "pro" machine can do and where it can be used.

I also have the freedom to be capped at 1/8 the amount of RAM (that is also less stable due to no ECC support) usage also limiting what this "pro" machine can be used for and what problems it can address...you know the purpose of this workstation computer.


For over a decade, apple has been unable (read: unwilling) to make a desktop workstation all for the glorification of their massive ego. If you like unified memory in small form factors, GREAT, GO GET a mac studio. Have fun paying apple 28k to put in extra RAM (depending on if your desired RAM capacity is even supported LMAO). This DESKTOP TOWER is supposed to have interchangeable parts. The fact that there are apple apologists regarding this aspect is just baffling and shocking.

I can only come to the conclusion that the people who are in charge of the mac pro at apple have their heads so far up their own ass that all they seem to care about are worthless youtubers who make shiny videos acting like giddy schoolchildren showing off a computer theyll either return or never use to the full capacity. People who will constantly kiss their own ass so that they keep getting free products to review. That is who this new mac pro is for.

/rant

 

[blackquartz]

What an incredibly disingenuous post this is. You're comparing a brand new SoC to a CPU thats 4 years old. wow its so much faster! who would have thought?

Congratulations, an integrated GPU is more powerful than a base GPU from 2019. But I can't add my OWN GPU, something a machine of this caliber should be able to do. The fact that there are PCIe slots is even worse since its basically mocking anyone who wants to replace their GPU.

"So for 1k more, you are getting a whole lot more"
Are you actually serious here?

For 1k more I'm also getting the freedom to not be able to boot linux or windows and run GPUs of my choosing severely limiting what this "pro" machine can do and where it can be used.

I also have the freedom to be capped at 1/8 the amount of RAM (that is also less stable due to no ECC support) usage also limiting what this "pro" machine can be used for and what problems it can address...you know the purpose of this workstation computer.


For over a decade, apple has been unable (read: unwilling) to make a desktop workstation all for the glorification of their massive ego. If you like unified memory in small form factors, GREAT, GO GET a mac studio. Have fun paying apple 28k to put in extra RAM (depending on if your desired RAM capacity is even supported LMAO). This DESKTOP TOWER is supposed to have interchangeable parts. The fact that there are apple apologists regarding this aspect is just baffling and shocking.

I can only come to the conclusion that the people who are in charge of the mac pro at apple have their heads so far up their own ass that all they seem to care about are worthless youtubers who make shiny videos acting like giddy schoolchildren showing off a computer theyll either return or never use to the full capacity. People who will constantly kiss their own ass so that they keep getting free products to review. That is who this new mac pro is for.

/rant

WOW, I mean, yeah you could be right. I understand the rant. But Maybe the Mac Pro time is up as a product, too insignificant compared to emojis and iPhones. Maybe this is its Swan song. I don't see Apple caring enough to compete with NVIDIA or AMD to create a GPU solution that puts up the fight. M2 and M1 just aren't graphic intensive processors and they never will be.

This 8,1 Mac Pro is the proof Marketing at Apple takes all the important choices no matter the segment and we as Pro users and enthusiasts were wrong to support Apple and this platform to begin with. It's on us. We should have moved on years ( maybe decades ) ago.

While I love my 7,1 I don't see myself investing to upgrade to the 8,1 under any circumstance.

 

[iPadified]

For over a decade, apple has been unable (read: unwilling) to make a desktop workstation all for the glorification of their massive ego. If you like unified memory in small form factors, GREAT, GO GET a mac studio. Have fun paying apple 28k to put in extra RAM (depending on if your desired RAM capacity is even supported LMAO). This DESKTOP TOWER is supposed to have interchangeable parts. The fact that there are apple apologists regarding this aspect is just baffling and shocking.

You just have to show Apple that there is a viable business case for a generic machine such as 2019 MP with multi CPU and GPU. I suspect that is difficult and it is therefore understandable that Apple has selected to compete in one professional market segment (video) and left the traditional generic high end workstation towers to Windows/Linux where there also exist much better softwares for such specialized work.

 

[fluidmodeler]

Just watched John Gruber’s WWDC 2023 interview with John Ternus and Greg Joswiak, which was excellent.

Gruber does a great job of explaining the central problem with the Apple Silicon Mac Pro — professional workloads heavily reliant on GPU compute (e.g., ML with large data sets) are simply not possible on Apple’s highest end Mac hardware given M2 Ultra’s relatively paltry GPU compute relative to pro workstations with multiple discrete GPUs.

Both Ternus and Joswiak seemed caught off guard by this line of questioning. After a bit of deflecting, Joswiak tacitly admitted that the Apple Silicon Mac Pro is simply not capable of serving this (growing) market of pro users (NVIDIA as their strengths, we have ours).

The interview quite entertaining and is worth a watch in its entirety:

 

[Varmann]

Here is a great analysis from an actual TV/Film/Music professional.

This the typical use case of the Mac Pro.

3.5 years ago he was one of the few who deployed Apple's 7,1 Mac Pro Rack and used up all the PCIe slots for his professional audio workflow.

Good points, in the end he says the current form factor is to big for the actual benefits from it, which I agree with.

However....
As a home user, I am certainly outside this particular segment. But the form factor was one of four deciding factors when I bought my old 3.1.

The specs was nice, especially in the beginning. For the first 3 years I used, on 24h average, 80% of the CPU capacity (I ripped my CD and DVD collection, 4k+ of discs). After 12 years it still did not feel obsolete.

The build quality was amazing, I got 12 years of use, partly 24/7, and that was very important since I bought it as a long term computer. Being able to easily upgrade GPU, memory, discs and network option was crucial and kept it alive.

After these two, the form factor came as number three. Being able to put everything inside a box was very nice. It actually took less space than my current solution (Macmini 2019, eGPU, external 3.5" raid 1, external enclosure for disk-backup, external raid-0 SSD M.2 enclosure, external bluray + a small factor gaming PC).

Number four was of course price, $3k was a bargain for a quality 2CPU machine. But over the years I invested 10x that amount, mainly in discs, but also other upgrades. Being able to buy things little by little is important to me.

So, while I did not use my macpro 3.1 professionally, I probably used it more heavy than many "Pro users" would have, and for a much longer time. Hence the build quality and form factor was important.

That may be a fringe use case, or a segment most of the computer industry have missed? Solid built computers for home/professional users that wants systems that "just works", year after year.

The 8.1 miss out on my list. Expensive one-time investment, probably a limited life span with few upgrade options.

 

[goMac]

It would be reasonable to suspect that with iOS as the dominant platform with Apple, and macOS becoming a UI-mod derivative of it, that virtual memory is eventually going to go away on the Mac.

iOS has virtual memory. iOS has _always_ had virtual memory.

I suspect what you mean is swap. macOS has had swap. iOS usually hasn't had swap.

However - iOS now has swap files on higher end hardware. So now macOS and iOS both have swapfiles.

So no, swapfiles are not going away on Mac.

 

[Longplays]

That may be a fringe use case, or a segment most of the computer industry have missed?

I think Apple and Intel would go bankrupt if people kept their computers for 6+ years.

That use case is what all computer companies hope to avoid.

Personally I like using my devices until after its final Security Update then upgrade to a product refreshed after it.

So say a device released in October 23, 2012 received its final Update on July 20, 2022. I'd buy whatever refresh came out soon after.

Solid built computers for home/professional users that wants systems that "just works", year after year.

Your use case is in decline as demonstrated how big box stores likely do not carry the parts for your Mac Pro on hand.

When going there you'd likely see a ~$799 Windows laptop or a gamer PC with LED lighting.

 

[mattspace]

iOS has virtual memory. iOS has _always_ had virtual memory.

I suspect what you mean is swap. macOS has had swap. iOS usually hasn't had swap.

Yes, you're right, Swap is what I was referring to, but what iOS has for "running low on memory" is "crash the app using memory".

What macOS appears to have now is:

^ This should never happen on a UNIX-derived operating system. Unless what it's saying is that the machine has run out of SSD for swapfile, this is Classic MacOS clownshoes territory.

 

[Melbourne Park]

I think Apple and Intel would go bankrupt if people kept their computers for 6+ years.

That use case is what all computer companies hope to avoid.

Your use case is in decline as demonstrated how big box stores likely do not carry the parts for your Mac Pro on hand.

When going there you'd likely see a ~$799 Windows laptop or a gamer PC with LED lighting.

I don't think that is valid.

Firstly, some usage includes long hours per week. Some perhaps 60 hours a week of intensive work. So 50 weeks @ 60 hours = 3,000 hours per annum. 4 years => 12,000 hours. But someone else might use the computer for 10 hours a week. After 4 years, that's 2,000 hours. In that case, it should last a lot longer than 6 years.

Secondly buying an Apple puts one in the Apple eco system. Which invites buying more stuff from Apple. Watches, phone, iCloud, notebooks, Apple TVs, software, subscriptions etc. If a machine failed early, that would hurt the whole eco system relationship.

Thirdly Value - Apple costs more than the alternatives in most cases. Hence it should last.

Fourthly - responsible manufacturers who want to conserve materials, should make long lasting hardware,

Fifthly Apples gear historically, has lasted and lasted. They've earned a reputation for longevity. Which has paid off with profits and loyalty. Short life products will destroy that.

 

[Longplays]

I don't think that is valid.

Firstly, some usage includes long hours per week. Some perhaps 60 hours a week of intensive work. So 50 weeks @ 60 hours = 3,000 hours per annum. 4 years => 12,000 hours. But someone else might use the computer for 10 hours a week. After 4 years, that's 2,000 hours. In that case, it should last a lot longer than 6 years.

Secondly buying an Apple puts one in the Apple eco system. Which invites buying more stuff from Apple. Watches, phone, iCloud, notebooks, Apple TVs, software, subscriptions etc. If a machine failed early, that would hurt the whole eco system relationship.

Thirdly Value - Apple costs more than the alternatives in most cases. Hence it should last.

Fourthly - responsible manufacturers who want to conserve materials, should make long lasting hardware,

Fifthly Apples gear historically, has lasted and lasted. They've earned a reputation for longevity. Which has paid off with profits and loyalty. Short life products will destroy that.

What I pointed to is market studies made by industry or individual companies.

In the 90s the replacement cycle was pegged at 3 years regardless of hours of use.

But in the last 1-2 decades that has lengthened 4-6 years.

Reason being that typical consumer laptops have becomes "good enough" for what it is being used for. Not to mention that people are more inclined to buy a post-PC device like a smartphone and refresh it every 2-3 years.

This is typical consumer use case.

This is why desktop workstations with PCIe slots have declined to become a niche. They're too "perfect" or "too much" for over 97% of use cases. Hence their worsening economies of scale that pushes up the price from top-end $3k price point in 2002 to a $7k base model in 2023.

That product category's 2.62%.

In 2022 292.3 million PCs were shipped worldwide. Of which less than 7.7 million were PC workstations like the Mac Pro.

To put it into perspective Apple's Q4 2022 Mac Shipments was greater than 7.5 million.

For example Apple pegged replacement cycle for their devices as:

- four years for macOS and tvOS devices
- three years for iOS, iPadOS, and watchOS devices

Most Apple products last longer and are often passed along, resold, or returned to Apple by the first owner for others to use.

Anyone's welcome to use their devices as long as they want. I personally prefer to use my Macs for a decade which coincides with after the final macOS Security Update.

 

[Varmann]

Your use case is in decline as demonstrated how big box stores likely do not carry the parts for your Mac Pro on hand.

When going there you'd likely see a ~$799 Windows laptop or a gamer PC with LED lighting.

I have a tiny hope that, evidently, there will be a (small) market for "sustainable" systems. Bought at a premium price but made for an extended life span.

At work we now cycle main work stations (used for shifts 24/7) every 4-5 years. After that they is kept as office computers with lighter workload until they give up.

 

[mattspace]

I have a tiny hope that, evidently, there will be a (small) market for "sustainable" systems. Bought at a premium price but made for an extended life span.

Go look at what Framework is doing in the laptop space. That's the future. Extended use, and reuse, not recycling.

 

[Longplays]

I have a tiny hope that, evidently, there will be a (small) market for "sustainable" systems. Bought at a premium price but made for an extended life span.

Typical consumer laptop is ~$799 with the expected life of 5-6 years. Even at $3k back in 2002 for a top-end PowerMac G4 that's a big ask.

From the consumer's point of view that means a bit too heavy on the cash flow, quality of life would suffer and technical overhead of having to open your case to replace parts.

So a desktop with PCIe slots, much less a workstation, really is a niche that is approaching that of a mainframe.

Hence the slower refresh cycle. Based on Apple's assumptions many 4+ year old Mac Pros will be upgrading to the 2023 model. Minority will balk at the change because of upgradeability but many will move for business reasons.

I'd be grateful that the Mac Pro has the M2 Ultra. It means that refreshes will occur every 15-19.5 months in sync with the Mac Studio instead of the previous 7 years or 3.5 years of Macs with PCIe slots.

At work we now cycle main work stations (used for shifts 24/7) every 4-5 years. After that they is kept as office computers with lighter workload until they give up.

Same at work but we tend to buy for the purpose of use of 2-3x longer years then sell. Use case hardly changes so new computers are bought based on additional seats or when the PC is too expensive to repair.

But I am putting a hard cut off at 10 years for preventive maintenance and so that everyone is on 2015 Win10 until October 14, 2025. I want 2021 Win11 to get 4 years of patches prior to deployment. Let others beta test patches and get potential downtime.

 

[Longplays]

Go look at what Framework is doing in the laptop space. That's the future. Extended use, and reuse, not recycling.

I have strong doubts that Framework will survive exceeding a decade.

Does their business model sustain operations when their customer base has 6+ years of user replacement cycle?

How expensive are their parts compared to the competition with larger economies of scale?

When right of repair laws are implemented bigger companies will crush them.

 

[mattspace]

I have strong doubts that Framework will survive exceeding a decade.

One company is not a paradigm.

Does their business model sustain operations when their customer base has 6+ years of user replacement cycle?

If a business depends on people re-buying things that already work, like screens and chassis, that business is going to be dragged to the guillotine as things get worse.

How expensive are their parts compared to the competition with larger economies of scale?

Th competition won't give up the economic waste heat of their profit margins, so input prices are largely irrelevant.

When right of repair laws are implemented bigger companies will crush them.

If their model is the one that is left after everyone else adopts it, then who has been crushed?

 

[Longplays]

One company is not a paradigm.

If a business depends on people re-buying things that already work, like screens and chassis, that business is going to be dragged to the guillotine as things get worse.

Th competition won't give up the economic waste heat of their profit margins, so input prices are largely irrelevant.

If their model is the one that is left after everyone else adopts it, then who has been crushed?

The point I am making is that Framework's customers are buying a future unsupported product.

Which is counter to the wanted outcome of repairability of more than 1-2 decades.

Right now big brands other than Apple allows for user servicable repairs like Dell and Lenovo.

Framework's just positioning themselves on what others are not advertising actively.

 

[fuchsdh]

There's absolutely no market that Apple sees worth catering to for enthusiasts who want to tinker with their rig. It didn't exist in the PowerMac era, it still doesn't. The people who bought used Mac Pros are hot-rodded them are a great bunch (I was among them) but it didn't make Apple any money, so it's not surprising they aren't trying to address it. Would I like such a product? Sure. But with the integration of Apple Silicon its time has passed for the platform. Anyone who's still building a Hackintosh these days seems nuts, you're on extremely borrowed time.

The bigger problem with the Mac Pro is it's doing a worse and worse job addressing the professional market (aka the people actually spending real money to buy these). I kind of agree with the ATP people that this Mac Pro feels like a bare minimum "we fulfilled our obligations" product by a company where the people who don't want a Mac Pro are winning against the people who do. I guess we'll see if they at least keep it an updated product in the lineup, because that's arguably still better than the 7,1 or any Mac Pro in the past decade, albeit just refreshing a product category is a low bar to clear. The $1K price jump I think is the biggest indicator of this—there's absolutely no reason for the new model to cost more than the old one, they just want to bilk the few people who will buy this, and that of course is going to further alienate potential customers.

 

[mattspace]

The point I am making is that Framework's customers are buying a future unsupported product.

No less unsupported than Apple's products, except that everything from Framework can be made by third parties.

Which is counter to the wanted outcome of repairability of more than 1-2 decades.

Right now big brands other than Apple allows for user servicable repairs like Dell and Lenovo.

Framework's just positioning themselves on what others are not advertising actively.

Framework is providing generational upgrades to motherboards, and repurposing superceded components into new enclosures so they keep working.

Apple provides ewaste chippers, so their secondhand devices don't compete with their new ones.