Hard Truths About The Apple Silicon Replacement to the Mac Pro (2019)

[Yebubbleman]

Well thought out logical post but stop misusing the word fact. What apple stated 2.5 years back is their strategy/approach at that time, and Apple can change their approach any time. It’s more logical to believe what apple says over other clickbait analysts.

That video wasn't a strategy/approach overview. That video is literally the guide to the platform architecture given to developers. Those pieces are fundamental. THEY STATE AS MUCH IN THE VIDEO!

 

[jmho]

With how integrated the Apple Silicon SoCs are, at that point the rest of the machine is basically just a Thunderbolt docking station / expansion chassis?

If so, it would be hilarious if the "Replaceable Processor Modules" were just Mac Minis.

We could finally get back to talking about Beowulf clusters again. It's been way too long.

I'm pretty much convinced this is where the Mac Pro is going given this year's "Accelerating machine learning with Metal" WWDC talk where they connected 4 M1 Ultra Studios together with thunderbolt.

Nobody wants to pay $20k for 4x Mac Studios cobbled together with thunderbolt, but $10k for 4x "Studios" connected together via PCI-e could be very interesting....

 

[257Loner]

But, what if someone wants to use a 100Gbps network card? Or, what if someone wants to use a quad-link 12G SDI card? That's 48Gbps... Thunderbolt is only 40Gbps... and only 32Gbps of that is usable for data.

They will buy a much cheaper, much faster Puget System. They will not spend an extra $6,000 over the Mac Studio to upgrade a 40Gbps Thunderbolt 4 connection to 100Gbps network card.

 

[TechnoMonk]

That video wasn't a strategy/approach overview. That video is literally the guide to the platform architecture given to developers. Those pieces are fundamental. THEY STATE AS MUCH IN THE VIDEO!

Architecture is also an approach, that can change, Apple changes more than most major companies. There is nothing preventing Apple from releasing a new/modified developer guidelines or architecture. What apple is doing currently isn’t a fact, unless you have insider information about future road maps. Apple like any other company will change its course on revenue, performance, and other factors. If Architecture was a fact we would be stuck with old x86 or other outdated chips.

 

[MayaUser]

As an heavy maya user, i dont care about upgradability as long as Apple will upgrade the Mac Pro every 3-4 years with at least double the performance on both cpu and gpu parts
They cannot make for me an Mac Pro with 0 upgradaility and leave it like this for 7-10 years

 

[BWNYC]

They will buy a much cheaper, much faster Puget System. They will not spend an extra $6,000 over the Mac Studio to upgrade a 40Gbps Thunderbolt 4 connection to 100Gbps network card.

as a person who has an 18-core Intel system from Bizon (a Puget competitor), I'd rather have an Apple Silicon Mac Pro with 100Gbps networking. For any price. 100Gbps networking is essential for SMPTE 2110 (uncompressed) workflows.

 

[257Loner]

as a person who has an 18-core Intel system from Bizon (a Puget competitor), I'd rather have an Apple Silicon Mac Pro with 100Gbps networking. For any price. 100Gbps networking is essential for SMPTE 2110 (uncompressed) workflows.

Well, I hope they can meet your needs.

 

[dawnrazor]

Here are some facts based on knowns and unknowns….

I really hope this is a typo… otherwise you undermine your entire post with this one sentence…

 

[Jumpthesnark]

There are no facts here just speculation. Until Apple releases or announces the new Mac Pro there are no hard truths.

Exactly. I know this is a rumors site, but until a Mac Pro is actually announced, all of this is speculation.

Here are some facts based on knowns and unknowns...

The word "facts" are not based on "unknowns," by definition. Facts are things that exist, they are provable. What follows this statement is based on a lot of evidence and speculation and even likelihood, but the bottom line is only Apple knows what is going to be in their next Mac Pro. No one here knows for a fact.

It's Schrödinger's Mac Pro. Its features simultaneously exists and don't exist, because nothing has been proven yet. Yes, we all come here in part for the speculative exercise, which is great. That's part of the community. But don't confuse speculation with fact.

 

[apparatchik]

Wouldn’t an M3 Extreme (up to 512GB RAM) with say 8 Thunderbolt 5 lanes (PCIE 4 / 80 Gbps bidirectional / 120 uni) be a feasible outcome?

There’s nothing stopping Apple from debuting the M3 series of chips starting from the top, the 3nm process from TMSC supposedly went into full capacity a few weeks back.

 

[jz0309]

I don't know who needs to hear this. Based on the hubbub on multiple Mac news sites, I'm guessing many. Here are some facts based on knowns and unknowns regarding whatever machine Apple is going to replace the Mac Pro (2019) aka MacPro7,1 with:


1. The RAM will not be separately user-upgradeable; it will be tied to the SoC.

2. There will be no PCIe GPU nor eGPU upgradability or expansion; the only GPU will be the one on the SoC.

Before anyone challenges me with that, make sure you have watched this video from WWDC 2020 (where the Intel to Apple Silicon transition was first announced): https://developer.apple.com/videos/play/wwdc2020/10686/ (at about the 1-minute mark)

3. This doesn't mean that there's no point to a Mac tower with PCIe expansion; there are plenty of professionals that need broadcast cards or special video tuners or audio interface boards in their Mac Pro; these things are not the kinds of things you can solve with Thunderbolt 4 or a Thunderbolt 3/4 breakout box. It's just not practical.

4. There's nothing in the referenced video above that negates the notion that Apple could socket the SoC and/or make it user-upgradeable/replaceable. Those of you that have used or operated a 2009-2012 Mac Pro (aka MacPro4,1 or MacPro5,1) have seen a similar concept in the form of the processor tray and backplane. There's nothing stopping Apple from doing something similar here. That's not to say that SoC upgrades likely won't cost an arm and a leg. They probably will be very expensive (assuming Apple goes this route). But it will still be possible to upgrade RAM and graphics this way.

5. The internal SSD on a 2019 Mac Pro is already proprietary, requires a DFU restore of the T2 chip in order to replace the storage; modules become useless when removed from the Mac Pro they came from; this won't be different on an Apple Silicon Mac Pro replacement either. Furthermore, if the SoC is to be user-replaceable due to being socketed or on a processor tray, the internal storage will need to be wiped when performing an SoC replacement/upgrade. This is how Apple Silicon and T2 Mac Storage works. This has no bearing on SATA or PCIe SSDs; just storage controlled by the SoC.

6. The base model SoC offered for this new Mac Pro will most likely run rings around the least expensive Mac Pro (2019) MPX AMD video card option. This is a safe bet. Less safe of a bet, but still perfectly plausible, is that it also runs rings around the MOST expensive Mac Pro (2019) MPX AMD video card option. This won't fully soften the blow of having the GPU be tied to the SoC and not upgradeable separately from it, but it will soften it for a decent amount of Mac Pro customers.

7. Apple likely won't introduce a dual-socket Apple Silicon Mac, let alone Mac Pro. This isn't a guarantee, but given everything that they said about using two discrete SoCs when first unveiling the M1 Ultra shows that they'd rather take two SoCs and bridge them internally into one mega SoC than go the dual-socket route. They could introduce a totally different technology that makes this feasible for the Mac Pro, but this seems unlikely.

8. "M2 Extreme" may have been cancelled, but it is extremely unlikely that an M2 Ultra, born out of two M2 Max SoCs with "Ultra-Fusion" will be the only SoC going into the next Mac Pro. You can customize a Mac Pro (2019) with 1.5TB of RAM. I'm sure that very few Mac Pro customers do this, but I'm also sure that there are some that do. Apple may not replace the current Mac Pro with a Mac Pro that goes all the way to 1.5TB of RAM, but it's safe to assume that they'd at least try to get halfway there. At best, an M2 Ultra, born out of the highest end M2 Max SoC times 2 would only yield 192GB of RAM. I'm not saying that isn't a ton of RAM even still. But a far cry from even half of the current Mac Pro's maximum. Let's assume that M3 Max is able to offer 128GB of RAM (by virtue of M3 being able to go to 32GB of RAM from M2's maximum of 24GB - up from M1's 16GB). That still only gives M3 Ultra a maximum of 256GB. Apple is going to continue the Intel Mac Pro's tradition of offering an entirely different class of SoC unique to Mac Pro. That's not to say that a "Max" or "Ultra" SoC won't still be on offer. That's totally possible too. There are probably many folks that would be fine with a "Max" chip's performance, but needing PCIe slots for specialized cards. But, you'd probably also have folks that would need to go to Ultra before eventually building a Mac Pro with that next level tier.

9. No, Apple hasn't forgotten about the Pros. In 2019, they released two products that all but outright admitted that they messed up. One was the current Mac Pro. The other was the first and last Intel 16-inch MacBook Pro (the first Mac since the butterfly keyboard to not have a butterfly keyboard and to be thicker than its predecessor for the sake of better performance). They did these moves for Pros. We're not getting another trash can. The "Ultra" configuration of Mac Studio is not going to be the best high-end desktop Mac that Apple is going to offer. You won't see regular upgrades to the Mac Pro. And, per that video linked above (which is to say "per how Apple Silicon is fundamentally designed as a Macintosh hardware platform"), you will not have the level of easy aftermarket upgradeability you had with the 2019 (let alone 2009-2012) Mac Pro. But it ought to still be a decent upgrade and not a trash can upgrade.

another post on a rumors site that claims "truth" and "facts" about an unreleased product - enough said

 

[StuAff]

It's Schrödinger's Mac Pro.

A phrase I've used myself. Very apt.

 

[MallardDuck]

I don't know who needs to hear this. Based on the hubbub on multiple Mac news sites, I'm guessing many. Here are some facts based on knowns and unknowns regarding whatever machine Apple is going to replace the Mac Pro (2019) aka MacPro7,1 with:


1. The RAM will not be separately user-upgradeable; it will be tied to the SoC.

2. There will be no PCIe GPU nor eGPU upgradability or expansion; the only GPU will be the one on the SoC.

Before anyone challenges me with that, make sure you have watched this video from WWDC 2020 (where the Intel to Apple Silicon transition was first announced): https://developer.apple.com/videos/play/wwdc2020/10686/ (at about the 1-minute mark)

2 is not obvious. They've done switched graphics before. eGPU is unlikely, but a mac pro without a PCIe GPU option would be vastly less successful than one with it. There are a lot of workloads that aren't metal, aren't going to be metal, and are critical for this market.

 

[Bug-Creator]

If they reach a ceiling in this time frame and have trouble bumping the specs, losing their market share to Intel / AMD, the logical step would be going back to x86 / x64.

This implies that performance was/is/will be the reason for AS Macs and that most users who buy a mac are overly concerned about performance.

Neither is true.

!!IF!! AS falls sofar behind to make it unsuitable for Macs then there will be no more Macs.

 

[BWNYC]

Wouldn’t an M3 Extreme (up to 512GB RAM) with say 8 Thunderbolt 5 lanes (PCIE 4 / 80 Gbps bidirectional / 120 uni) be a feasible outcome?

Yes! This would be completely acceptable to me!

 

[Joe Dohn]

This implies that performance was/is/will be the reason for AS Macs and that most users who buy a mac are overly concerned about performance.

Neither is true.

!!IF!! AS falls sofar behind to make it unsuitable for Macs then there will be no more Macs.

SOMETHING has to justify the premium you pay for Apple. It used to be the ecosystem, but Apple has been neglecting it. The next thing is performance. If the Ecosystem is neglected and there is no performance, what does have have going for Apple Silicon? Pretty icons? Saturated colors?

 

[bobcomer]

Top of the market wants performance and reliability. Nobody in the enterprise gives even the smallest damn about upgradeability. The only people upgrading are hobbyists and smaller companies who’s IT departments are run by PC enthusiasts. Most professionals don’t care about that stuff.

Well --- yes and no. They don't like upgrading a server or workstation after its in service, but they definitely do like configuring it to do the job from the vendor. If it needs TB's of RAM for the job its going to do, it gets configured that way to begin with and if they can't order it that way, they wont buy it. That's the kind of upgrading we do...

Xeons right now are dead in the water. The only things that keeps them going is the established name. Anyway, you got a point there. Apple needs to deliver performance and reliability. If they can’t a Mac Pro doesn’t make much sense.

There's one thing a Xeon does over other intel x86 alternatives -- ECC RAM. And some servers or workstations it's actually a good idea to have. (depending on criticality or environmental reasons)

 

[bobcomer]

their current top notch machines max out at 96 GB RAM,

128G -- Studio Ultra...

 

[bobcomer]

No, actually pretty much everything I posted here is fact or a stated likelihood (which, itself is factual). The first two points come from Apple in a video I reference from WWDC 2020.

Not set in stone until they actually release the machine.

Apple can change it's mind after all, especially after 2 years.

 

[StuAff]

'Apple hasn't done it' does not mean 'Apple will not do it', nor 'Apple cannot do it', 'Apple has no plans to do it', etc.
If SGI managed to build a unified memory system with expandable RAM, pretty sure Apple could.

 

[Dan From Canada]

If Apple gives customers zero of what they want, Apple will have zero customers. No one will buy a $6000-$24000 Mac Studio Pro for internal expansion when you can buy a $2000-$4000 Mac Studio and use Thunderbolt for external expansion.

That also applies to "normal" M1, M2 Macs.
As a long time Mac user I am looking at returning to windows just because I can change/upgrade things in it.

 

[Joe Dohn]

As someone whose owned every design of the MacPro I offer the following:

1. I am very disappointed about the lack of user upgradable ram due to the high cost of Apple ram. Further, sometimes you just mess up and don’t order enough. For my current unit I have over 200 GB. I ordered 128 GB and had to upgrade about three months in.

2. I’m less concerned about the GPU as with video editing, most Apple Silicone GPUs today are plenty, even for multicam 4K editing.

3. The OP is correct that the tower has greater functions that just the ram and GPU. The trash can was horrible as it didn’t have enough Thunderbolt ports and no ability to add more hard drives.

4. Ultimately, though, I will likely buy the new Mac Pro despite my hesitancy of the aforementioned. I need to work, and if this machine will make my workflow more efficient, highly likely, it will still make financial sense.

Do you really think though that a Mac Pro will add more value than x64 + Nvidia in this scenario? I doubt so. And if you say it will, I doubt it's something a Mac Studio can't offer instead at a much more effective price point.

 

[Joe Dohn]

For rendering FCP, Apple Silicone is king. That’s the thing about Apple hardware and software. You get the gestalt, where the benefits are greater because the software is specifically meant to maximize the hardware.

Excuse me, what is FCP? Final Cut Pro?

 

[sam_dean]

So this will be a Mac Studio with expansion slots for non-SSD & non-GPU expansion.

Could anyone point out what sort of internal expansion are available currently for the market now that can use these expansion slots?

Last non AIO desktop I had was a 2002 Power Mac G4 QuickSilver. We never used its expansion slots or even drive bays.

If the 2021 Mac Studio M1 Pro equivalent was available back then at 1/3rd the cost of a Power Mac then I'd have opted for that.

 

[senttoschool]

There are a lot of valid points here.

I'd like to add that the high-end workstation market is simply much smaller than it was before. Heavy workloads that could move to the cloud have moved to the cloud. This isn't 2010 anymore.

Take for example, Threadripper (Zen3) was 1.5 years late. Intel has all but given up on the workstation market. They're barely even trying.

There are only so many ultra professional video editors and much can just be done on as little as a Macbook Pro 16" or an affordable Ryzen 5950x + Nvidia desktop.

I maintain that for Apple to make the "M2/3 Extreme" work economically, they will have to create an Apple Silicon Cloud where businesses can rent a Mac Pro for hours at a time. This is in addition to selling a traditional Mac Pro workstation. https://forums.macrumors.com/thread...t-a-40-core-soc-for-mac-pro-now-what.2306486/