Apple Silicon Mac Pro Reportedly Not Coming at WWDC, Mac Studio Refresh Likely Delayed Until M3

[CWallace]

My point is: adding more lanes isn't quite the same level of work as adding PCIe from scratch. Any Mx chip already has PCIe, but yes, to your point, it may not be enough lanes to support meaningful internal expansion.

But does an Apple Silicon Mac Pro need a significant number of PCIe lanes? I do not see Apple allowing third-party GPUs so I do not think they will need the eight total slots of the 2019 model, nor having four of those eight x16 slots.

I think they could get away with two x16 slots for networking and storage cards and then a few x8 slots for everything else.

 

[DavidSchaub]

But does an Apple Silicon Mac Pro need a significant number of PCIe lanes? I do not see Apple allowing third-party GPUs so I do not think they will need the eight total slots of the 2019 model, nor having four of those eight x16 slots.

I think they could get away with two x16 slots for networking and storage cards and then a few x8 slots for everything else.

Well... that is why I made this art:

DavidSchaub (@dschaub@mstdn.social)

Attached: 1 image @dmoren@zeppelin.flights Back in 2021, I made this Mac Pro based on the rumours of the day

mstdn.social

... but the rumours are of full sized case, so I would assume 8 slots?

If they're going to bother doing that, even without GPUs (though they may have a hardware accelerator story still!), I would assume they would just build the PCIe bridge chip with sufficient bandwidth for them all. That beats them having to put a bridge chip on the motherboard like with the Mac Pro (2019).

As I previously posted, there are lots of cool bandwidth intensive cards out there.

 

[Boil]

... but the rumours are of full sized case, so I would assume 8 slots?

If they're going to bother doing that, even without GPUs (though they may have a hardware accelerator story still!), I would assume they would just build the PCIe bridge chip with sufficient bandwidth for them all. That beats them having to put a bridge chip on the motherboard like with the Mac Pro (2019).

As I previously posted, there are lots of cool bandwidth intensive cards out there.

Could do six slots by removing the "in-between MPX" slots (# 2 & 4), leaving two x16 slots for ASi "MPX-ish" GPGPUs (mainly the power delivery & the massive "passive" heat sink); remaining four slots could be a x16 for a M.2 RAID card, x8 for 8K video I/O card, & two x4 for audio I/O DSP cards...?

 

[257Loner]

The current version of iMac is essentially a giant iPad with a build in stand but no touch screen.

The essential features of the iMac are the same as any modern computer since 1984's Macintosh:

1. Metaphorical Object-oriented Graphical User Interface featuring a virtual desktop environment.
2. Windowing system that allows users to work with several programs at the same time.
3. Built-in software for productivity, creativity, media creation and enjoyment.
4. Internet-based communication apps such as web browsing and email.
5. Screen
6. Keyboard
7. Mouse

 

[deconstruct60]

The Mac Pro could have about 90 lanes worth of PCIe slots... which is probably WAY more than even multiple M* Max silicon can offer.

Mac Pro can get by with WAY less than 90 lanes. The current MP 2019 has a two input x16 lane PCI-e switch that provisions out most of the slots on the system. Slots 2 , 4 , 5-8 and the MPX x4 connectors. If Apple did two x16 PCI-e v4 lanes they have the same total aggregate bandwidth as the MP 2019 has ( with 64 PCI-e v3 slots). Send that to an updated PCI-e v4 switch and could easily provision 6 lanes as they did before just fine. ( even better bandwidth distribution if drop the MPX lanes.. so down to just those 6 for general use all the time. ) .


I highly doubt Apple is going to be in the whole "top end bandwidth pissing match" mode. Even doing two x16 PCIe- v4 controllers would be large stretch from what Apple has been doing so far.

The other problem too is that I am somewhat doubtful the internal die data distrubtion network really has that much 'spare' , mostly unused bandwidth to allocate to PCI-e. There is going to be a ton of NPU , CPU , and GPU core traffic. Traffic to the display controllers. To/from the uncompressed video handlers . etc. The "backhaul" to ploy another 90 lanes of v4 ( or v5) probably isn't there. And it makes about zero sense to put some complicated PCI-e switch on the die. That clearly could be 'outsourced' to someone else .

The die edge space that 60 PCI-e lanes would take up would likely be better overall system allocated to more memory ( where Apple is throwing most of their edge space too now. ) . More GPU/NPU non-generic cores are likely going to need more memory bandwidth ; not less. SRAM/cache are harder to get ( because have basically stopped scaling. ) Relatively long distance ( multiple inches ), external ( off-die) SERDES I/O links really have basically stopped scaling so well also.

Long term if Apple picks up PCI-e v5 (and CXL 2.x ) then get a future bandwidth bump in future version while keeping lane allocation about the same.


Benchmarking against the lane counts of the AMD Epyc and Xeon SP is probably not where Apple is going. Different market.

 

[deconstruct60]

But does an Apple Silicon Mac Pro need a significant number of PCIe lanes? I do not see Apple allowing third-party GPUs so I do not think they will need the eight total slots of the 2019 model, nor having four of those eight x16 slots.

Going to likely want matched electrical lane allocation out to x16 for PCI-e v2 and v3 x16 cards even if the new Mac Pro just goes to version v4.

Two x16 v3 cards could be 'switched down' onto one PCI-e v4 x16 backhaul just fine. (actually better than the MP 2019 works. )

A substantive number of folks have the "need a container" criterion because want new home for older cards they want to keep using. Someone with a x16 PCI-e v3 four M.2 SSD card provably won't want to backslide on a x8 electrical slot that is PCI-e v4 . Can theoretically claim that isn't a backslide on bandwidth , but pragmatically it is going to be one ( electrically dumped half the lanes and still running at the previous gen speeds. that will be slower).

Future storage bandwidth requirements are extremely likely not going to get lower.


Circa CES 2023

All the PCIe 5.0 SSDs Coming Out in the Next Year or So

We finally have some model names for upcoming PCIe 5.0 drives.

www.tomshardware.com

by 2024-2026 getting a PCI-e v5 x16 card to host four of these won't be difficult. having just PCI-e v4 x16 would be a handicap. But not as much as limited v4 x8 or some v3 x4 .

I think they could get away with two x16 slots for networking and storage cards and then a few x8 slots for everything else.

Apple likely needs to justify the $6K ... So a one or two input PCI-e v4 switch that can adjusted to allocate bandwidth to several slots ( i.e., what they have in the MP 2019 ) is probably going to be a welcomed 'value add'.

Just two x16 slots is too focused on riling up the GPUs are the only cards that are important crowd.

Apple's networking support (drivers and certified cards) for over 100GbE is relatively poor. so deep need for x16 v4 for networking is huge stretch. Probably not.

And pretty good chance that Apple will need a 'compute card' accelerator support. Even if continue to avoid display focused GPUs.

 

[deconstruct60]

Could do six slots by removing the "in-between MPX" slots (# 2 & 4), leaving two x16 slots for ASi "MPX-ish" GPGPUs (mainly the power delivery & the massive "passive" heat sink); remaining four slots could be a x16 for a M.2 RAID card, x8 for 8K video I/O card, & two x4 for audio I/O DSP cards...?

No slots 2 and 4 provisioned off the PCI-e swtich is likely what want to keep. It is slot 1 and 3 with direct links to the CPU package that are far more sensible to dump ( if Apple has a limited PCI-e lane budget coming of the primary package. )

The MPX connector portion deadiated to PCI-e v3 x4 is for additional Thuderbolt ports . Very likley Mac Pro already has enough Thunderbolt ports already provisioned by the main package. it is a "solution" in search of a problem that has largely been eliminated.

 

[Boil]

No slots 2 and 4 provisioned off the PCIe switch is likely what want to keep. It is slot 1 and 3 with direct links to the CPU package that are far more sensible to dump (if Apple has a limited PCI-e lane budget coming of the primary package.)

I was thinking more of the actual physical location in regards to an ASi GPGPU/compute card having a MPX-style heat sink...

The MPX connector portion dedicated to PCI-e v3 x4 is for additional Thunderbolt ports . Very likely Mac Pro already has enough Thunderbolt ports already provisioned by the main package. it is a "solution" in search of a problem that has largely been eliminated.

Is that not also the portion which carries the extra power...?

As I have said many times before, when I am talking about "MPX-style", I am mainly interested in the massive passive heat sink & the extra power delivery; I fully realize it is not needed for Thunderbolt provisioning anymore...

 

[DavidSchaub]

I highly doubt Apple is going to be in the whole "top end bandwidth pissing match" mode. Even doing two x16 PCIe- v4 controllers would be large stretch from what Apple has been doing so far.
...
The die edge space that 60 PCI-e lanes would take up would likely be better overall system allocated to more memory ( where Apple is throwing most of their edge space too now. ) . More GPU/NPU non-generic cores are likely going to need more memory bandwidth ; not less. SRAM/cache are harder to get ( because have basically stopped scaling. ) Relatively long distance ( multiple inches ), external ( off-die) SERDES I/O links really have basically stopped scaling so well also.
...
Benchmarking against the lane counts of the AMD Epyc and Xeon SP is probably not where Apple is going. Different market.

Oh, I agree... It is WAY more likely that Apple will continue to under-provision the PCIe lanes bandwidth. Probably in the ballpark of the current 64 lanes the Mac Pro (2019) has.

There are a pile of unknowns:

- I'm still guessing that Apple won't utilize any of the M* dies for PCIe slots, they'll add an independent bridge chip to the UltraFusion fabric? But that is just pulled out of some dark place, and is probably wrong.

- Whether Apple will still utilizes bridge chips on the motherboard like on the Mac Pro (2019)?

- Whether the M* Ultra and M* Quadra Mac Pros would have the same PCIe configuration or same PCIe bandwidth? (it could if they use an on package bridge chip)?

- How is the 4x M* Max scaling working with UltraFusion (given that it sounds like Apple hasn't gotten it shippable, I expect this is not going well), and will any nasty bottlenecks show up?

 

[DavidSchaub]

Is that not also the portion which carries the extra power...?

As I have said many times before, when I am talking about "MPX-style", I am mainly interested in the massive passive heat sink & the extra power delivery; I fully realize it is not needed for Thunderbolt provisioning anymore...

I would hope MPX simply goes away and Apple doesn't over-engineer the power delivery... PCIe power cables are just NOT that bad, and if the TB/video backchannels aren't needed, drop it entirely.

Hot take, I suppose: MPX was an over-engineered mess that just fed Apple's "single cable to monitor" addiction... It just wasn't worth the investment.

 

[Twigg90]

I’m confused, am I suppose to be saving for the Apple Silicon Mac Pro or just purchase the 2019 model?

I hope we get some answers at WWDC!


Also, there’s been some conversation as to wether we can “hold” Apple responsible for not truly completing this in the 2 year deadline. I think my biggest issue is when they announced the Mac Studio they specifically said “There is only one Mac left and that is for another day”.

Well, it’s been over a year now!
We still know nothing!

I think if you’re going to make that statement you need to provide something within a year. I took that statement as being within 365 days a device will be released, or at least announced or we will have solid information at to what the future holds for the Mac Pro. They made it sound like they truly had something special in store and were looking forward to announcing it “soon”. Now, they still might but it’s disappointing and confusing because there is no timeframe.

He should have said “There is only one Mac left and that is for next year”, but I know Apple would never say that. Still, they HAD to have known it was delayed and that statement to me was INTENTIONALLY misleading.

 

[JippaLippa]

I really like Mac OS. It's easily Apple's best product... but for some screwed up reason, half the population will gladly get ripped off on the price of iPhone but they will NOT get a Mac computer to go with their phone? WTF. Why aren't Macs more popular?

What makes a computer decent isn't subjective; we can pretty easily give a value to a number of factors and pretty much all of these are in the hardware. And with all the hype Apple tried to sell us years ago about how great M1 chips were, they still haven't delivered a solid desktop for power users.

I wonder too...

I'm the only person I know that's all into the ecosystem, and it's not about money as I know for sure I have less than them.

I know people with macs, and expensive android phones...
People with iPhone Pros and cheap windows computers...

Sure, you can do what you want, but having most stuff apple made gets you more benefits that each item combined.

I could never think of letting go of my mac-iphone-airpods pro triad, and I wish I could comfortably afford an iPad pro.

 

[Allen_Wentz]

If it's real and it doesn't have good games, it will 100% be a failure.

That is absurd. Apple has never considered games any kind of priority, yet they became the 4th largest PC seller and by far the most profitable PC vendor. Not having good games is pretty much irrelevant. If Apple does better gaming it is just a little bonus.

 

[Gadger Snr]

I think the Mac Studio refresh will come in late 2024 or earlier 2025 using the M3. I think that’s when we will get the new Studio display at the same time. The a 27-inch mini-LED panel, featuring ProMotion support for refresh rates up to 120Hz.

 

[Allen_Wentz]

I think the Mac Studio refresh will come in late 2024 or earlier 2025 using the M3. I think that’s when we will get the new Studio display at the same time. The a 27-inch mini-LED panel, featuring ProMotion support for refresh rates up to 120Hz.

I doubt Apple would intentionally wait that long to refresh the Studio. This is tech and delay means lost sales like we saw with the delay releasing M2 MBPs.

 

[DavidSchaub]

I doubt Apple would intentionally wait that long to refresh the Studio. This is tech and delay means lost sales like we saw with the delay releasing M2 MBPs.

I would like to agree with this, but if anything the years has shown to me is: Apple will happily get products rot on the vine for longer than expected.

I would hope that has changed, but the lack of an M2 iMac 24" suggests that Apple hasn't changed.

If it doesn't cost them *enough* sales, maybe Apple doesn't care.

 

[opeter]

Well, AMD does plenty of interesting things. The entire current x86-64 architecture is AMD's design, which Intel licensed from them.

AMD and Intel can share technologies. That's why you see AMD's x86-64 (or AMD64) instructions in Intel CPUs from 2004 onward.

From wikipedia page:
Historically, AMD has developed and produced processors with instruction sets patterned after Intel's original designs, but with x86-64, roles were reversed: Intel found itself in the position of adopting the ISA that AMD created as an extension to Intel's own x86 processor line.

Intel's IA-64 (Itanium) was not compatible with the older x86 instructions.

 

[chucker23n1]

I think they could get away with two x16 slots for networking and storage cards and then a few x8 slots for everything else.

x16 storage? Like… I suppose you can do a RAID 0 of 8 NVMe SSDs and come close to saturating that…

For networking, that seems even more of a stretch. 100GbE would only need x4.

Sounds very niche to me, much more so even than the 2019 Mac Pro was. For most I/O stuff, Thunderbolt 5 should be plenty. Internal PCIe slots, as before, seem mostly interesting for GPUs, and if Apple isn't going to do that, I'm not sure there's a point.

No that's wrong, they built the CPU without support for it.

Support for what? PCIe 5? That'll happen sooner or later. eGPU? What in particular is the SoC lacking? It seems to be mainly a firmware (iBoot probably doesn't have a notion of "your pre-OS display output goes to a different GPU") + OS thing. The OS already supports it on Intel; it might already be compiled for ARM anyway.

 

[chucker23n1]

I'm a bit confused by the above...

Apple could support eGPU for video-out on ARM Macs if they wanted to. They have seemed to just build the OS without support for it. That's a choice Apple has made.

Yeah. I suspect it's mostly a matter of implementing it in iBoot. macOS already has eGPU support on Intel, and I don't see how that would be architecture-specific. (I suspect their Thunderbolt stuff largely isn't architecture-specific either.) But the boot loader is completely different on ARM than on Intel, and it hasn't so far needed anything like this.

I'm not sure there was ever a decision either way on will they/won't they; rather, they have simply decided that it wasn't enough of a priority yet. If my hunch is right, the feature could still come as a firmware update even dating back all the way to the 2020 M1. (They coooooould even do that for M1/M2 iPads. "All" it would require is for them to port the dGPU stuff from macOS to iPadOS.)

 

[opeter]

The worst problem with the MacPro is the fact that you have to unplug every single device to get the cover off. The old MacPro has a nice drop-down panel. The new MacPro one has to lift the entire housing up and off and if anything is plugged in, you have to unplug it. So dumb.

That problem was introduced with the trash can Mac Pro (2013)... and the 2019 model just expanded on that.

 

[Realityck]

Almost nobody upgrades their phone every year, so this isn't a relevant concern in terms of the environment.

Even the 2 year get the new iPhone free contracts from T-Mobile, Verizon, and AT&T reflect that fact.

 

[steve123]

I really like Mac OS. It's easily Apple's best product... but for some screwed up reason, half the population will gladly get ripped off on the price of iPhone but they will NOT get a Mac computer to go with their phone? WTF. Why aren't Macs more popular?

I think this is because the purchasing influence is much greater in the other direction. The iPhone sells well because people that buy macs have a very large influence on the people who buy iPhones. Probably because an iPhone on average is cheaper than a mac, the average iPhone user is not very tech savvy, you can easily finance an iPhone through a telecom carrier and iPhones are perceived as a sort of "status" symbol, that is, you are part of that crowd that can afford a mac. This effect explains why Mac Pro is the Halo product for Apple. iPhone sales far exceed macs but the influence of mac users on iPhone sales is enormous. This also explains why adoption of new Apple products generally takes time (iPad, Watch, AppleTV). The influencers need to be satisfied they can recommend it and the influencers are primarily made up of mac users. And Mac users generally aspire to own one if they could justify and afford it. The Mac Pro quite frankly is the ultimate status symbol for a Mac user (followed closely by a maxed out MBP).

 

[DavidSchaub]

I think this is because the purchasing influence is much greater in the other direction. The iPhone sells well because people that buy macs have a very large influence on the people who buy iPhones. Probably because an iPhone on average is cheaper than a mac, the average iPhone user is not very tech savvy, you can easily finance an iPhone through a telecom carrier and iPhones are perceived as a sort of "status" symbol, that is, you are part of that crowd that can afford a mac. This effect explains why Mac Pro is the Halo product for Apple.

The trick is that iPhones are status symbols, and Macs are definitely not status symbols. Most people could care less who uses a Mac, and care even less how much money was spent on one.

The realty of the Mac Pro truly becoming a "Pro" piece of hardware is, I kinda think consumers buying Mac Pros are not "high status", but are just "kinda wasteful goofs" for burning so much money. Mac Pros should be bought by people who's work will pay for it within a few years.

 

[steve123]

That is absurd. Apple has never considered games any kind of priority, yet they became the 4th largest PC seller and by far the most profitable PC vendor. Not having good games is pretty much irrelevant. If Apple does better gaming it is just a little bonus.

I think this is consistent with the way Apple used to think and still does to a large extent. Gaming does not dominate their thinking but when they can, they will take steps to improve the gaming experience because now they have the ability to do that more cohesively.

I recall an interview, https://www.macrumors.com/2023/02/06/apple-executives-techcrunch-m2-chip-interview/, with Apple Executives who said Apple was taking a long view on gaming and there would be an evolution with Apple Silicon to position it more favourably with game developers. The results would take some time to evolve but that gaming was now on their radar. It was a rather vague comment I know but none the less, it was interesting that it was discussed at all.

 

[DavidSchaub]

I recall an interview, https://www.macrumors.com/2023/02/06/apple-executives-techcrunch-m2-chip-interview/, with Apple Executives who said Apple was taking a long view on gaming and there would be an evolution with Apple Silicon to position it more favourably with game developers. The results would take some time to evolve but that gaming was now on their radar. It was a rather vague comment I know but none the less, it was interesting that it was discussed at all.

I think I read that as: "We aren't going to invest much money or resources on gaming, but we may consider gaming in our decisions as long as they don't cost us anything" ;P