Apple May Not Launch Updated Mac Studio With M2 Ultra Chip Due to Similarity With Upcoming Mac Pro

[theluggage]

For example, the Intel Core i9 13900 is almost 50% faster than the Intel Xeon W 3.2 GHz, which is the base CPU of the 2019 Mac Pro.

But it's not all about raw CPU benchmarks. The M-suffix Xeon Ws used in the MP support up to 2TB of RAM (Apple only supports 1.5TB so don't open too many tabs in Safari ) and 64 lanes of PCIe (needed to multiple high-end GPU cards that each need 16 PCIe lanes for full performance).

According to ark.intel.com that i9 "only" supports a maximum of 128GB RAM and 20 PCIe lanes. Hey, it would do me fine, but it's not in the same league as the Mac Pro. One reason why the Mx Ultra is being ruled out as a Mac Pro successor is not lack of CPU power, but its 128GB RAM cap (maybe 192GB on the M2 Ultra) and limited PCIe (the M1 Ultra uses the internal equivalent of 24 lanes of PCIe for Thunderbolt - a Mac Pro would still need a few Thunderbolt ports - plus it appears to have some spares PCIe for Ethernet etc.).

As for AMD - well, yeah, I'd probably choose AMD over Intel at the moment and Apple could have gone AMD if they hadn't gone for their own chips instead. However, Xeon has a reputation (deserved or otherwise) for stability and is probably a "tick list" feature in some institutions because nobody ever got fired for buying Intel. Otherwise, AMD would have wiped Intel off the face of the Earth by now. Of course, Apple's problem is that those sorts of conservative buyer won't touch Apple Silicon with a bargepole, either...

For the moment, if Apple wanted a drop-in (or nearly) upgrade for the Mac Pro they'd be dependent on Intel producing a direct successor to the Xeon-W (...maybe they have - I don't know).

 

[Joe Dohn]

But it's not all about raw CPU benchmarks. The M-suffix Xeon Ws used in the MP support up to 2TB of RAM (Apple only supports 1.5TB so don't open too many tabs in Safari ) and 64 lanes of PCIe (needed to multiple high-end GPU cards that each need 16 PCIe lanes for full performance).

According to ark.intel.com that i9 "only" supports a maximum of 128GB RAM and 20 PCIe lanes. Hey, it would do me fine, but it's not in the same league as the Mac Pro.

Which is why I mentioned in the comment:

But suppose you NEED server features, such as ECC memory and very large amounts of RAM available.

However, Xeon has a reputation (deserved or otherwise) for stability and is probably a "tick list" feature in some institutions because nobody ever got fired for buying Intel. Otherwise, AMD would have wiped Intel off the face of the Earth by now.

I don't see why the customer which buys Xeon processors wouldn't buy a Mac Pro on the processor alone, as it's the same processor.

The main issue here is not the brand, but the price: Apple is NOT currently offering a good bang for the buck. For example, AMD's latest EPYC processor, which costs a whopping $10,000-$11,000, would STILL offer better value than a Macbook pro. You could pair it with 1TB RAM and it would only cost you an extra $1,000.

If the benchmark I posted is correct, we are talking about at around 13% speed increase single thread and 75% increase multithread if we consider the processor alone. If we add GPUs to the mix, it probably gets much worse.

 

[chfilm]

Not necessarily, especially if you’re using an application that has been coded, and tweaked for, Intel processors for years and years. And, potentially, has a GPU pipeline that assumes IMR and not TBDR (trying to fit an IMR into TBDR is bound to lead to performance issues).

So, while the potential is there, the software’s going to have to catch up. It really depends on what you’re doing though. If you’re running Intel based benchmarks, no, Apple Silicon doesn’t have to beat that. If you’re running Final Cut Pro, where the vendor was focused on being as performant as possible on Apple Silicon, then yeah, there would be a big difference.

I think davinci resolve is pretty much one of the most opitimized apps out there for apple silicon, no?

 

[chucker23n1]

The Trashcan was a noble effort at producing a media creation "appliance" that could hit above its weight when using software that was optimised for using dual GPUs and maybe open up new markets. I think the 3 fundamental mistakes were:

1. letting the former Mac Pro PCIe tower get too out of date and then discontinuing it, so the trashcan had to serve as a replacement for something that it wasn't.

I honestly never understood why they gave it two GPUs, but… it would've been interesting if they had simply called it the Mac Studio and released it alongside a regular Mac Pro update.

Would people have been interested in that? Maybe if the Mac Pro were $1,000 more?

 

[theluggage]

I honestly never understood why they gave it two GPUs,

Because GPUs were increasingly being used for all sorts of processing tasks, rather than just driving displays, and with GPU-based computing being heavily multi-threaded anyway, two GPUs should be better than one.

I guess what happened in practice was AMD et. al. just crammed more cores/shaders/whatever into their single-chip GPUs - but those were too hot for the Trashcan design which assumed 3 roughly equal heat sources...

 

[chucker23n1]

Because GPUs were increasingly being used for all sorts of processing tasks, rather than just driving displays, and with GPU-based computing being heavily multi-threaded anyway, two GPUs should be better than one.

Oh, I know, but that made the 2013 Mac Pro's design only useful for some pro apps. Quite a few, but not all. For software developers, for example, it's a very silly approach. I wish they'd either made the branding reflect this specialization, or offered a configuration that, instead of 1 CPU + 2 GPU did 2 CPU + 1 GPU.

Imagine if all configurations of the 2023 MacBook Pro were as GPU-focused as the 2013 Mac Pro was. Instead, you can buy the M2 Pro, and that gives you fewer GPU cores, because for some pro apps, they'd be wasted anyway.

I guess what happened in practice was AMD et. al. just crammed more cores/shaders/whatever into their single-chip GPUs - but those were too hot for the Trashcan design which assumed 3 roughly equal heat sources...

Well, that, and Apple didn't really put much effort into driving GPGPU development on the Mac. They didn't want CUDA, but also offered a very confused hardware and software story on what you should be doing instead.

 

[rasputin666]

5120 x 1440 in 49"? What a joke!

I'll stick with my 27" 5K iMac used by Mac Studio as an extended display and easily switch to universal control to actually work on it with a single set keyboard/trackpad.

why is it a joke? every use case is different, what works for one doesn't for another.

 

[duck apple]

The 2019 Mac Pro's distinguishing features were really:

• Support for multiple high-end workstation-class AMD GPUs or other 8/16x PCIe cards
• Lots of PCIe expansion space for specialist interface cards with 1x-4x PCIe
• Capacity for >> 512GB RAM
• ECC RAM support (even if it's only a tick-list feature with LPDDR).
• Space for internal storage expansion via proprietary flash, PCIe-to-M.2 and even a couple of SATA rust spinners

External GPU is the first thing Apple to avoid as plague.
It needs effort to adapt and maintain driver (without full source code as AMD engineers enjoy) , consumes too much power, wastes too much PCIe slot space, and emits too much noise and heat.

Instead, with the 3nm process, M3 could easily increases its GPU count to 128 or even 256 if max CPU cores and/or RAM capacity didn't scale up proportionally. Or they can design daughter board based on the GPU in Apple Silicon.

With 3nm and improving EDA pipeline, Apple should be able to design a more modular architecture that provide variant combination of CPU/CPU count and RAM capacity to fit more usage scenarios.

M3 based MacPro could be configured with 80 CPU cores with 512G RAM but only 12 or 24 GPU cores,or 128 or even more GPU cores with the base 128 RAM and 24 or even 12 CPU cores.

And I guess MacBook Air and Mac Mini will be positioned to qualified only for previous generation SOC or the minimum base of latest generation to keep its price competitive.

Mac Studio will be in the middle, rides on latest generation of SOC but could only reach the base capability of MacPro and no NVMe expansion and PCIe connectivity.

MacBook Pro will be less capable than top end of Mac Studio for obvious reason.

 

[duck apple]

Hmmm… now I have a decision to make. I was waiting for the M2 update for the Max Studio before I upgraded my 2018 Max Mini. Mac Pro will most likely be priced above what I’d like to spend.

Go Studio Ultra, and upgrade to whatever better fit with M5 or M6.
Myself got the top configure Max (with 2TB SSD); it brings back fun of computing and I'm regretting not getting the top Ultra for headroom of RAM an GPU power and enough CPU cores for VM then I can stick with it for the next 5 or 8 years.

I recently got an 2nd hand iMac 2020 with 10-core i9 and top configured Pro WX5700 XT (about the same price as my Studio Max) because my project needs Linux running on native X86 (in a VM). Summing up, it's a not-worth-of-it experience for personal daily usage, task like A/V conversion is 2/3 of Studio Max and simply web browsing (columns of silly GIF ad.) would spin up the fan noise.

 

[sam_dean]

TLDNR: my prediction - M2 Max and M2 Pro "Mac Studios/Fat Mac Minis" get launched as "new Mac Pro" at $2000/$4000 + inflation. Maybe with a "proper" rackmount option.

The 2019 Mac Pro's distinguishing features were really:

• Support for multiple high-end workstation-class AMD GPUs or other 8/16x PCIe cards
• Lots of PCIe expansion space for specialist interface cards with 1x-4x PCIe
• Capacity for >> 512GB RAM
• ECC RAM support (even if it's only a tick-list feature with LPDDR).
• Space for internal storage expansion via proprietary flash, PCIe-to-M.2 and even a couple of SATA rust spinners

I doubt 2023 Mac Pro will lower from $6k as the economies of scale was impacted by the cheaper 2022 Mac Studio.

Over 50% of Mac Pro users are unwilling to pay an extra $2k for those bullet pointed features.

It may be more economical to keep the N5 process node Mac Studio for 4-6 years then upgrade to a A14 or A10 process node chip.

The $6k Mac Pro would have the same tech specs of a $4k Mac Studio M2 Ultra without PCIe expansion slots.

If Apple were to produce a two die M2 Ultra to make a M2 Extreme then expect the price point to increase to $10k at double the RAM & SSD of the $6k model.

 

[duck apple]

why is it a joke? every use case is different, what works for one doesn't for another.

But you didn't say it's an awesome display for you.

1. One needs to sit distant from the display to have a good angle of view, not good for my aged eye.
2. The vertical 1440 resolution of that screen size is going to be pixelating and the font will be ugly if one sit at the proper distance.

One may need a ultra wide monitor for application with a long timeline as possible, but with this resolution, the photo or video pane won't be of any good for prove reading.

 

[duck apple]

why is it a joke? every use case is different, what works for one doesn't for another.

But you say it's a decent
The vertical 1440 resolution of that screen size is going to be pixelating.

 

[duck apple]

There is the xMac Studio Thunderbolt chassis from Sonnet, that’s basically what you’re imagining. Although it just uses regular Thunderbolt, sounds like you’re wishing for more like full bandwidth PCIe 4.0 x16.

There’s also the OWC ThunderBay Flex 8, that’s PCIe slots plus HDD bays. I haven’t seriously considered either, as the price/performance doesn’t seem worth it for me, but I’d be curious to hear from anyone who’s tried them!

The OWC ThunderBay Flex 8 is a brain damage product.

A single TB3 only allows 22Gb ~ 24Gb for PCIe tunneling. You can't get more bandwidth by just adding a PCIe switch upstreaming to the host TB3.
And you can't aggregate all the 6 TB3 ports on the Studio Ultra into a fat PCIe port without Apple messing around its PCIe lanes-port assignment (a daunting job if ever possible).

 

[atonaldenim]

Certainly the Studio has been out long enough to be updated. M2 Max chips are out in the MBPs, so that makes the M1 Studios now officially last year's old tech. I can see why the Ultra may remain M! due to conflict with the coming MP, but the Studio Max needs (trivial) upgrading to M2 right now.

I’d guess the yields aren’t there for M2 Max to be offered in additional products beyond their most important laptop models, yet.

If M2 Ultra launches first in Mac Pro and doesn’t come right away to Mac Studio, probably same reason.

Hopefully later in the year they’ll have enough supply of M2 Max and Ultra for the Studio to get refreshed.

But my M1 Ultra Studio is running quite well, I think most people in the market for a Studio have one by now, and aren’t in a big rush for an update already.

 

[duck apple]

for the majority of the people who want this type of product, the regular M2 Max chip would be more than sufficient

I need Ultra's 20 CPU cores + 128G RAM for heavy duty VM; the additional GPU cores could be of use in places you never expected.

 

[atonaldenim]

There was a time theatrical films and TV were edited on Final Cut on Mac, but try to find any major project since about 2011 that admits they used Final cut. I’ll wait.

It’s true FCP is no longer an indie filmmaker darling. But fun fact: 2020 Oscar winner “Parasite” is said to have been edited in FCP7, on a Macbook Pro.

 

[duck apple]

it makes much less sense to make a Mac PRO with pci express slots that would be used for adding … what??? Am inferior AMD GPU? A raid card that can easily be attached via thunderbolt 4?

How about A 16 lanes PCIe slot to fit with a RAID board connecting to a 96 lanes PCIe switch for twenty 6GB/s of various connector format of NVMe SSD?

Or, simply 4 or 8 Apple designed, spring-loaded M.2 bays for user replaceable NVMe SSD?

The above storage expansion solution needs zero labor in software.

And may Broadcom be very willing to port their SAS/PCIe combo RAID board for almost limitless pool of disks that gives 16GB/s or even 32GB/s RAID?

TB4 won't see the tail light of performance of PCIe based or even SAS based solution.

Display is not going to play a role here since it's exclusive redundant to the function of the powerful and efficient native GPU and needs massive software effort from AMD or Apple. The heat and noise will not be ignorable if it's to be a lot more powerful than native GPU, which will very possibly be consisting of 64 or 96 cores, without needing two turbulent fans .

 

[duck apple]

Except you can't just cram more transistors on a chip and make it faster. There are practical limits to how many transistors you can put on a single chip/how big you can make chips, and the M1 Ultra (if you count it as a single chip) and M2 Max are already well ahead in the "how many transistors can you put on a chip" race (https://en.wikipedia.org/wiki/Transistor_count).

TSMC claims its 3nm process gives 70% increment of density (vs. 5nm).

 

[duck apple]

Now they are on a proprietary processor with no true compatibility with Intel and you think making a pro machine makes any sense when software developers are most likely going to start backing out supporting mac's again (just like the 90's)? They ripped out every standard for graphics and are pushing their own iOS centric metal platform. They provide no Nvidia compatibility for expanding GPU. It's a clear done deal on the pro market. They will ride the Final Cut Pro and Logic wave till the company eventually can't compete anymore. People will realize what a huge waste of money it is to buy their machines with the limited software and hardware support.

The devices running the most restricted OS and almost zero expansion beat all the other brands combined in terms of profit.

According to a survey, Mac is the only one with market growth in computer field. You just can't fall in love with the MacBook, Mini and Studio due to its performance and energy efficiency.

Software vendors see the turn around created by the M and will jump on the wagon.

 

[duck apple]

I will NEVER buy the studio display to connect it to another Mac. Neither do I need a MacBook because working on a laptop is unergonomic. Just release a 30” iMac.

Agree, but why not 32" in 6K.

 

[Detnator]

So many it's hard to enumerate them here. In the case of NVIDIA cards, a striking example is CUDA support. Apple cannot currently offer any alternative if you need CUDA acceleration.

Nope they can’t. And they don’t want to nor need to. Apple makes Apple stuff. Nvidia makes Nvidia stuff. Different companies, different products, for different needs. Each delivers on their targets extremely well. This obsession people have with Apple needing to make products that perform Nvidia tasks totally misses the point.

If people want to do Nvidia stuff then buy Nvidia stuff and put the Nvidia stuff into machines that take Nvidia stuff. Problem solved. None of that takes away from how good Apple stuff is at doing what Apple stuff is for.

If you need a toaster don’t buy a fridge and complain it doesn’t toast. 😉

 

[Detnator]

All of my statements are accurate, I use apple products daily. Heck, I am a multi decade stock owner with many thousands of shares. Yet, I call apple for what it is. I see the computer industry and what is happening, and apple is on a path that looks to be very wrong with people that make decisions for purchasing professionals equipment for business. The move to intel was a huge leap that afforded them a second look by many businesses. They were adopting open standards making developers give them a second look. They have effectively undone all of this advantage. Everything is back to proprietary and exclusively standards for development. The home market is limited and when people start finding programs that run on windows exclusively because developers will not port their apps or redevelop software to apple's graphic frameworks, that's a HUGE problem. They can only ride out the fruit ninja style games and tap based photo touch-up apps for so long. Adobe apps are getting worse on apple platforms and it is turning to 1995 adobe is better on PC all over again. I hold no bias. I am worried they are really botching things up with all these drastic changes and unproven proprietary architectures. It is a company losing touch with the computing community and professionals that have been using their product for over 3 decades, like myself. They should have kept standard pc ideology in their pro line of products. perhaps they will surprise us with the Mac Pro. but so far.. it is all grossly overpriced, look cool at Starbucks, nonsense. It's a bicycle for your mind, not an accessory. They serve a generation of iPhone/tiktok users now, not professionals. The last Mac Pro was so promising, but the cost greed just made it embarrassing.

Umm…. No. All your statements are unsubstantiated opinions in multiple long winded rants. I can’t think of any message you’ve written that had any facts supporting it. The above is no exception.

You’re entitled to your opinions. But for the most part you’re just plain wrong, or you just don’t get it (Apple).

Apple have plenty of missteps but no more than any other company. For the most part they know what they’re doing and millions upon millions of people are buying their stuff and are happy. It’s not just people buying their stuff that speaks. Apple tops the customer satisfaction charts year after year.

 

[Detnator]

How did it work on Mac systems with DOS-compatible systems installed into expansion slots back in the 90s?

Not particularly well. 😉

I had one for a short time. I’m not certain but if I recall correctly the entire Windows OS ran in a window in a dedicated Mac app. There was little integration if any between the Mac and Windows OS’s. But it ran natively. Better than emulation, for sure and better than virtualization even since it was the only thing running on that CPU, dedicated RAM, etc. just sharing monitor space, keyboard and mouse.

Of course someone please correct me if I’m remembering that wrong.

 

[Detnator]

why are you so angry? your comment literally verifies my point that apple has stopped being a true professional computer computer. I need to go elsewhere now? hmmm.

I’m not angry. I’m wondering why you are! Post after post of rants about how Apple has got it all wrong for your needs. So maybe give the ranting a rest and just go buy something else that meets your needs…? The market is full of them. They just don’t run macOS. But what’s more important?

 

[Detnator]

Apple could invest a truckload of money in developing a new ARM ISA CPU with huge PCIe and RAM capacity in direct competition with Xeon, AMD (and possibly Amazon et. al. who already have server/high-density-computing-grade ARM chips that are closer to what is needed than Apple Silicon), that could support AMDs latest and greatest GPUs, but the resulting GPU power would be no better than any other PC system stuffed with the same GPUs. So they'd have to invest another truckload of money trying to develop a proprietary GPU that could beat AMD and NVIDIA.... and then recoup that investment from their smallest-selling Mac when their competitors have far larger established markets... and to keep the market for that smallest-selling Mac they'd have to make their whizzy new GPU perform well on legacy software that was optimised for AMD GPUs which limits the extent to which they could do anything radical.

Compare that with the M2 Pro/Max range (presumably to be joined by a M2 ultra at some stage) which scales nicely from a 14" Mac Pro to a Mac Studio Ultra using just one basic die design - and while it doesn't have the GPU to match AMD's latest, what it does do is pretty much perfect for the ultra-portable laptops that are Apple's bread and butter - especially when paired with Apple Silicon-optimised software - giving them a substantial boost over what Intel/AMD can do in a comparable form-factor.

The Trashcan was a noble effort at producing a media creation "appliance" that could hit above its weight when using software that was optimised for using dual GPUs and maybe open up new markets. I think the 3 fundamental mistakes were:

1. letting the former Mac Pro PCIe tower get too out of date and then discontinuing it, so the trashcan had to serve as a replacement for something that it wasn't.
2. depending on Intel and AMD producing future chips that would work with the thermal design.
3. a few years down the line, trying to replace it with the iMac Pro all in one (YMMV - but I suspect that is what actually happened, with the whole mea culpa news conference kicking off just when they'd have been showing iMac Pro prototypes to key customers/developers).

The Mac Studio/Max certainly looks like the return of the trashcan, but in rather better circumstances: (1) the 2019 Mac Pro is still available as viable solution for those who need it, (2) it's now up to Apple to produce whatever future chips they need to fit their products and (3) the iMac seems to have been put back in its proper place as a low/mid-range general purpose desktop where most users aren't going to expand it. Also, thunderbolt is faster and more mature than it was (when the Trashcan came out there was stuff all available to plug into it).

I think Apple's best bet would be to keep the Intel Mac Pro on life support for a few years - maybe a CPU bump if/when Intel offer something better - as well as putting the last Intel-compatible version of MacOS into long-term support (it's not like Pro users with hard-to-change workflows are begging for annual OS upgrades). That leaves them free to "think different" about how the existing Apple Silicon range could scale to some radical new approach to high-end workstations - probably via some clustering system based on M2 Ultra "compute modules" linked by Thunderbolt - without trying to pretend it could be a like-for-like replacement for a traditional tower.

Worst thing they could do would be to kludge together some parody of a tower workstation using M2 tech especially while annoying people by not bumping the Studio to M2. Better to openly sideline the Mac Pro, take the hit, and focus on what Apple Silicon was good at rather than drawing attention to what it wasn't suitable for.

I’ve now read every comment on this article and this is one of the best I think. Some very insightful points here.

I’m not sure I like or agree with the idea that they let the intel MP hang around and I’m just holding out hope they have something up their sleeve that can be a worthy ASi replacement for the existing Mac Pro sooner rather than later. But if they can’t yet then keeping the Intel one around until they do is a good approach.

I think the key differences between Mac Studio and MP are gobs of RAM and gobs of PCIe expansion, but now for things other than 3rd party GPU’s — I think a commitment by Apple to their own GPUs is worthwhile and they should put everything they can into making those worthy competitors to what you’d otherwise put in the MP’s PCIe.

I think that’s what they’re doing and hope they’re getting their GPUs to levels comparable to the other companies’ top tiers to deliver in a MP and then provide the slots for all the other PCIe options beside that.

I agree if they haven’t figured that out yet then they shouldn’t release something half assed, and the Intel MP should continue to fill that gap until they do. I just hope that’s not too far away. With all the rumors of an ASi MP coming real soon one can only hope they’ve got it right.

I love this assessment of how the Mac Studio differs from the trash can in timing and the environment it’s in now (alongside a decent MP, not replacing it).

Maybe they’re not quite there yet and they’ll provide both at once. Release a new ASi MP that serves some of the needs above the Mac Studio even if we can’t put TB’s of RAM in it, etc. while keeping the Intel MP alive alongside it (eg. like they did with Intel Mac minis for 18 months) until they can eventually bridge the gap.

Whatever they do, I’m 90% certain they’re not interested in any ASi Mac supporting non-Apple GPUs and they’ll eventually, later if not sooner, deliver on Apple GPUs that deliver at the highest levels compared with AMD’s best. As for Nvidia in Macs, that ship sailed years ago - long before ASi - and it’s not coming back.