Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP All Models 40 core ASI with PCI-E slot is around the corner!!
- Thread starter Amethyst
- Start date
- Sort by reaction score
0 PCI-e slots and yet ...
Post #1 in this thread ... 1 slot
Post #427 in this thread ... 6 slots
to bet on zero seems to be a bit of a stretch when reportedly there are already non-zero slot systems running. Betting on one of the numbers between 1 and 8 would be safer bet.
The "extreme" package may have been trimmed off , but unlikely that would also trim off the slot allocation of the system board the package sits on.
Even if Apple wanted to resort to cheesy hackery they could take an 'extra' TB controller or two and provision some slots just by adding discrete TB controller and cheap PCI-e switch to roll out some PCI-e v3 slots. That would be almost as easy as falling off a log just to get to some non zero number. If add a one or two x16 PCI-e v4 controller to the package, then dramatically even less cheesy. 6 slots is strongly suggestive that they have something better than an extra TB controller.
Kind of hard to have an 'extra' M.2 slot when don't even have one in the first place. MP 2019 doesn't have one. No other Mac system has one ... so where is the first one coming from.
Again it is going to be super simple to take one or two x1 PCI-e v4 lanes that already come out of other M1/M2 SoCs and add a USB controller ( to provision an internal USB socket like on MP 2019 board) and another to a discrete SATA controller to again replicate the minimal two SATA socket interfaces of the MP 2019 board. ( there are external discrete SATA controllers that work on external cards via thunderbolt so the drivers would be 'off-the-shelf' to get for Apple. ) . No guarantees going to get an internal USB and SATA sockets, but roughly the same user use cases that made those valuable in 2019 are still in play today. ( iLok licensing and bulk data internally).
It would be nice if Apple put a M.2 slot where the Memory used to be, but doubtful they are going mess with M.2 as doing a standard PCI-e slot is even cheaper to implement. ( and folks can just stick a M.2 carrier card in that slot).
Leaning almost entirely on ports begs the question of what Apple needs yet another Mac Studio. They have got one.
Leaning too hard on Thunderbolt was a problem in 2017. It is an even bigger problem now in 2023.
Apple likes to setup leakers based on wrong information, there is no way we are going to see a trashcan MacPro
Yeo. I feel like what this thread keeps circling around is the Trash Can is already back and it's the Mac Studio.
Assuming what Mago is "hearing" is true (which if it's not clear I am extremely skeptical of) - the machine he's describing sounds like a likely Mac Studio update. Not a Mac Pro.
- WWDC 2023
- ASi Mac Pro preview
- Shipping December 2023
- N3E process / A17-based cores
- Base M3 Ultra Tower - $6K
- Six PCIe slots (Gen4/Gen5)
- Rackmount conversion kit option
- M3 Ultra / M3 Extreme options
- Hardware ray-tracing
- Maximum 512GB LPDDR5X RAM
- Possible ASi (GP)GPU options
- Base M3 Extreme Cube - $8K
- 7.7" tall Mac Studio-style chassis
There is likely truth to what Mago is hearing. Consider the ARM developer transition kit comprised a A12Z in a Mac Mini form factor. Apple never released this as an end user product. I suspect Apple has other transition kits used internally and externally by developers for each Mx transition. Much of the information we hear is likely leaking from developers using these transition kits. The actual end user product prototypes are likely very closely guarded.
N3E doesn't start high volume production unitil 2H '23 (July or later). It is too late for the A17 roll out. So if there are A17 elements ( P core clusters , E cores clusters , GPU core clusters , etc) then pretty likely N3B. N3B and N3E are not design rules compatible. It is not a very straightforward flip.
Decent chance Apple doesn't use N3E at all. One or two cycles on N3B and then move on to N3P. Really shouldn't be in a hurry to 'burn up' all of the N3 family iterations because more than decent chance N2 is going to have a delay hiccup.
If there are six slots doubtful there is some kind of "conversion kit". Probably reusing the same two chassis have now. The rack variant might get a simplification mode if Apple removes all the "removables" from the bottom of the logicboard. Then don't need an awkward to access 'door' on the bottom.
What is in the other two slots the chassis provides? Decent chance non standard , proprietary Apple stuff. Lots of folks have wished for a variant of the old MP 2009-2010 riser card. It could be that, only far , far , far more expensive.
Or Apple could just leave the filled to banks with no target slots behind them. ( holes where folks who want to do odd-ball experiments where pump even more power into the box have a path). Just cheaper.
A smaller cube cost more than a taller tower? Not going to happen.
An extreme SoC package in the Studio's desktop footprint isn't going to work. Adding some limited height vertically above that 'too small' logic board footprint isn't going to solve the problem. And one , now even more dubious, fan on the bottom isn't going to work well either.
Having highly overlapping products up in the $8K and up range is nuts (for the Mac ecosystem). The user base up that far is barely large enough to support one product, let alone two. If Apple uses an Ultra in both a Studio and Mac there are going to be demonstrably segmented on price. One will start around $4 and the other around $6K. Yes, you'll be able to add non-SoC BTO options to push the Studio in the $6-8K range, but if add the same stuff to a Mac Pro it would be there; it will be $2K higher than that.
As soon as can optionally computational augment that $6K tower with something that "taller studio" is going likely going to loose in many contexts ( even if go thru gryations to stuff an 'extreme' into one. ) .
Finally, the Studio could have been made taller to ease the cooling issues with the Ultra , only Apple didn't. In part it the "unobtrusive desktop space/volume" constraint that the Mini and Studio are working under. Studio is higher but it is still low enough that it can slide 'underneath' most monitors set to a normal height adjustment. So it doesn't have to occupy highly llmited space outside the monitor's "shadow" footprint onto the desktop. Too tall and just way passed what the Mini is anchored around. So why stay attached to the mini's design constraint?
A cube isn't going to make the "I want a mini tower xMac" folks happy. And the NeXT cube lost against the Sun 'pizza boxes' on the desktops for good reasons. It is a "look at me , look at me" stunt that makes not efficient desktop utilization sense at all. Didn't then, doesn't now.
I pray for that as wwdc is closer less chances for it to be a cheese grater with expandable pcie etc and more on a trashcan or compact unexpandable cheese grater or cube whatever.
Supply chains are extremely complicated - and it doesn't even sound like the Mac Pro has entered production at this point. There are people like Ming-Chi Kuo who's full time job is watching Apple supply chain sources for information.
I'm extremely skeptical : random Macrumors poster : has somehow got better supply chain sources than Mark Gurman or Kuo. If that's true and Mago is actually that good, Mago shouldn't be here. Mago should be charging $20 a month for a newsletter like those other guys.
Amethyst claimed to have sources - probably at Apple - who know bits and pieces of information. That's far more realistic for how development works. It's very hard to get knowledge of the entire widget. And Mago changing their story every few weeks should make clear they likely have no idea what's going on and they're just making it up.
N3B, N3E, N3P, whatever... As long as it is on a 3nm process utilizing A17-based cores... No "old stuff" in what should be Apple's most advanced & performant Mac ever...?
SoC(s) on riser cards would allow SoC & RAM upgrades, replacing the whole riser card with a paid exchange from Apple; like upgrading a Silicon Graphics workstation worked "back in the day"...
Six slots because the "in-between" slots in the "MPX area" are removed, they would be blocked by the 4-slot heat sinks from the two ASi (GP)GPUs anyway...
Considering I had the base Tower with a M3 Ultra and the base Cube with a M3 Extreme, yes...
Internals would be different than Mac Studio, just the uniformity of the exterior chassis housing is what I want to see with the Cube; Mac mini, taller Mac Studio, even taller Mac Cube...
Internally, Cube would have vertical logic board up one side & Mac mini-style (just taller & higher power output) PSU up the other with heat sink in between; bottom intake (Mac Studio "foot") and top exhaust via 180mm high static pressure fan...
What overlap...? For the non-PCIe slot needing crowd the Mac mini goes up to Mn Pro, Mac Studio up to Mn Ultra, Mac Cube is Mn Extreme only...
If one needs PCIe slots, there is the Mn Ultra / Mn Extreme Mac Pro tower...
The Mn Extreme Mac Cube would very much so be a "LOOK AT ME" halo product, Phil Schiller would return just to drop the "can't innovate my ass" line again; and when paired with a 16" iPad Studio & Mixed Reality Headset, no worries about monitor "shadow" placement issues...?
Look bro, I just want an ASi Mac Cube, let me dream on this forum built upon rumors & conjecture...! ;^p
Probably keep the two MPX bays (but not as MPX). It is definitely slot 8 that is in danger. That's where the Thunderbolt controller was . If the TB controller is now inside the SoC then that SoC needs to be near where the old ports where. Probably better if they just put at least 4 TB ports there and just make the 'top'/'fron't ports USB4. If soak up slot 7 also can also do another couple TB ports , HDMI , and some USB ports to along with those. Slot 8 has special GPIO and USB connector on it also ( that regular x4 cards wouldn't 'see') that whole thing goes away.
So the riser card could fill the holes allocated to both slot 7 and 8. The formerly quite tall horizontal (in tower configuration) heat sink for the very large package is not going to get turn 90 degrees and go vertical. The space where the old CPU package was in 2019 model is going to move 'down' in the tower to give the heat sink more clearance. Going 'down' is likely going to minimally run into slot 8. And if want to give more clearance for a J2i bracket moving down to slot 7 can at least get some 'clean air' on at least part of the J2i bracket's top device (i.e., not pre-heated by SoC package). RAM, SSD module sockets , nominal ports driven by SoC controllers .... all on the riser card.
Nominally the classic place for the Afterburner card was slot 5. If that is in the SoC now also , then what would have been pushed from slot 5 to slot 7 can now just go in slot 5.
[ and if Apple keeping the USB/SATA internal connectors could move those up closer to where the J2i bracket so don't need quite as long cabling to get to the drives. ]
The layout for slots 1-6 can stay the same.
The crazy 400->500->700 W trend on GPU cards means probably do need two 4 double wide slots left after finish here. There is going to be no shorter of cards with monsterous heat sinks in the "Compute accelerater" market going forward. However, decent chance though the MPX connector get removed from slot 1 and 3 though.
The TARDIS aspect of getting that Extreme SoC on a Studio footprint is the bigger problem.
Even taller Studio ... with one, only one fan, is highly dubious. The lack of proper ventilation in the nominal mini design constraints is going to wreck this thing. ( The Ive the cooling system has to be magically absent philosophy is goofy. It isn't magical at all. ).
Apple would be way, way , way better served to throw that road-to-nowhere-nostalgia R&D money they might throw at that , over to making one (or two) Mac-on-Card systems for hypermodular folks who are going to pick up and leave when can't run Nvidia GPUs and can install their own DIMMs. They could put a Mac inside a slot of their Boxx/Pudget System/Lenovo/Dell/HP box to run the subset of macOS stuff they want to hold onto long term and the rest Windows/Linux. ( as a PCI-e card don't need a AC/DC power supply (get DC power off board). Two TB4 ports , Ethernet , one USB4 port would simpliy the other discrete components on the limited board space. two SSD modules. And pack a fan cooler on top). Mac users could even throw that card into one of the double-double wide slots if wanted another instance to hand off workload to.
That has a way bigger , expansive target market than trying to depress higher end Mac Pro sales. Try to GROW desktop aggregate sales . not balkanize them into as small as possible pools of the same size.
Again if twerk the orientation of the large, tall heat sink 90 degrees and don't have room ... going to run into placement problems. The Extreme package is bigger but the heat sink is ALSO going to get bigger. It bloats out in 3D; not just two.
Missing the point that in the $8K there is hardly anyone up there. The vast majority of the "no PCIe slot" crowd is down in the $2K-4K price space. Dramatically smaller group in the $4K - $7K space . By time get to $8K there is almost nobody left. Pragmatically need to herd those folks onto something they might not totally prefer but will buy because there is no other option; the Mac Pro.
If the riser card path is around in the Mac Pro it isn't just don't need any PCI-e card... also don't never need any RAM/CPU/GPU upgrades ever either. If bought the extreme because needed more RAM what make you then they won't need more RAM later? If folks are on the lunatic fringe of needing more CPU cores , GPU cores, and/or RAM also diving into the group of folks who won't want those all permanently soldered down.
At some point the target market is so small can't keep chopping it into smaller pieces. There are substantial number of old school hypermodular Mac Pro users who are going to walk away. Pretty likely the eligible pool of users is going to shrink smaller. Taking something that is already small and making it smaller isn't going to help much.
When the Mini got a M1 Pro option that was expansive. The Studio is probably somewhat expansive of the old iMac 27" market. The big CPU/GPU iMac folks herded into Studio. Some of the xMac , "I'll never buy an integrated screen" folks into Studio. And pulls some Mac Pro folks down into Studio. (or opened door back up to them after closing it by increasing the entry Mac Pro price 100%. folks in the 3K-5K zone of old Mac Pro).
The Cube thing isn't going grow nothing. Most of the folks probably want to point to were previously herded into buying a Mac Pro and were happy enough to pay/use it. That will still be true if don't try to balkanize what is left of the Mac Pro user space. [ the very unhappy folks probably left for Windows/Linux. And at $8K price point not going to bring many of those back. If have an insatiable core count or insatiable GPU core count problem ... $8K is going to buy more there if relatively price/cost constrained ]
Finally, folks who really need an Extreme level of processing power probably need more than just one internal disk drive. The notion that most of these folks are not going to put another PCI-e SSD inside their system is likely bankrupt. That much processing power likely leads to high data capacity storage needs . Apple's SSD pricing is quite high. More money for Mac Pro can be offset by far cheaper market rates for your high capacity storage needs.
Folks can leave lots of their data on SAN/NAS networks , but those don't have to be capped at 10GbE either.
The Extreme's SoC major problem though is that is far , far , far more needs BUYERS not just lookers. Need to grow the desktop Mac market bigger so that it can get foundation so it can econoically justify an appropriate die design that is effective.
The whole thing of Readily Pro as a primarily replacement for a mainstream , primary display is ergonomically and battery challenged hot mess. Besides, how is it going to work any significantly better with a Cube than it would a Mac Pro? Give up using 27-32" monitor so can use 16" ipad. Yeah OK.
If Apple wants a box to sell to the "I'm an Apple super duper fan", hipster crowd for conspicuous consumption then just come out with a model of the Studio Utlra that is painted Space Grey and has one extra 'exclusive you don't have one' GPU core activated. It would be far less cheaper R&D and less of a distraction.
Seven slots total; specialty SoC Riser Card slot between Gen5 x16 (GP)GPU slots, for shortest distance to SoC; remaining four PCIe slots above & below (two & two) those...
SoC Riser Card & ASi (GP)GPUs all use 4-slot heat sinks...
When one looks at the cooling system in the Mac Studio it occupies a substantial "layer" in the chassis...
Half of this layer is heat sink, the other half the "quad" blower fans...
Rotating the logic board and switching to a (higher power output) Mac mini-style PSU opens up the remaining interior volume for a massive heat sink, with a single 180mm HSP fan exhausting out the top (and no funky PSU restricting airflow intake from the bottom "foot")...
At this point not arguing about if a Cube makes sense, but how a Cube could be functionally laid out...?
Safe or not, 0 PCIe slots is still my bet. #1 and #427 are wrong. The 2019 Mac Pro has plenty of extra m.2 slots, you just need a PCIe adapter to use them. Without PCIe, an m.2 slot could serve as a reason to buy the Mac Pro over the Mac Studio. Apple is free to prove me wrong about this but I don’t think they will.
GPUs are on die, Afterburner is on die, Thunderbolt and USB controllers are packaged with the CPU. Apple is cutting ties with both Intel and AMD and going their own way.
This comments section is going to be funny when WWDC comes and goes without even a mention of the Mac Pro.
Yeah, I have a funny feeling they will talk only about goggles 🤪
...and it'll require a dedicated top of the range black anodised Mac Studio, which will run a customised OS that does nothing but drive the goggles as a basestation, and can only communicate with AS Macs which have to run the actual apps.
I'm 50/50. But I don't think it will have anything to do with the headset. Just feels like Mac Pro development is in a real troubled place. But I guess they could do a preview like they did with the 2019 even if they are running behind.
If it is in a troubled place, it suggests that Apple jumped into ASi with both feet before confirming that whatever plan they had for the Mac Pro was solid. Which in itself tells you something about its position in their priorities. They wouldn't have taken that kind of risk with e.g. the MacBook Pro.
Even if a Mac Pro does eventually come out, I wouldn't hold your breath for updates.
Weirdly I disagree.
I even hope they don’t even mention it during WWDC. The longer it’s delayed the more likely we get a real good m3 extreme based machine (and pci5) we all really want (Versus the compromised non extreme m2 variant).
I don’t mind at all that it will take longer but they will do it right and well as a consequence.
I think this makes sense, but it would be good of Apple to at least let us know what's going on (not that they will, of course). Their pledge of 'renewed commitment' to the pro market is starting to wear a little thin.
It is not in a troubled space. There are good indications it will be announced at WWDC, will use an M3 variant and it will outperform the competition. AS already outperforms almost everything else and keeps improving as the API's for the hardware accelerators mature. Apple is not in this game to hop scotch back and forth with Intel and AMD ... the M3 will put them in the leadership position by a wide margin and they will stay there.
This is sounding like a car forum with the Tesla devotees pushing out Tesla PR messages.
It will be interesting if Apple announces nothing.
It will be interesting if Apple announces nothing.
There’s no indications of it being announced at WWDC. We’re 2-3 weeks away and there’s no leaks so far. MR doesn’t even mention it in their list of expected reveals.
Mobile devices are likely 95% of Apple’s hardware business. ASi is totally optimised for mobile and does very well there, but on the desktop, AMD and Intel have the performance lead. You’re correct that Apple doesn’t plan to hopscotch with those companies, but unfortunately, not in the way you hope.
it was one thing when Apple could raid the Xeon parts bin, but those days are over. Apple would need to compete despite having a very small workstation business, and no server presence at all. The idea they are going to consistently dominate Intel, AMD and Nvidia in situations that are not power constrained is frankly laughable.
Last edited:
apple needs to do something to address the delay going that long with no updates is bad and makes it look like that apple silicon can't do pro workstations / can only do pro workstations at X3-4+ the cost of other pro workstations (not talking about ones you can build on your own no ones from pro level system builder)