MP All Models Mac Pro X,Y…WHAT IF…

[deconstruct60]

https://twitter.com/i/web/status/1773135334567788661

Perhaps an Extreme?

If those M1-M3 die pictures are all to the same scale, then probably not. The M3 area size > M2 area size > M1 area size. The M3 bigger ( to point had to subsume the UltraFusion area for 'stuff' ) means there is a serious incremental bloat issue here.

Multiple chips would mean even bigger areas consumed which leads to even bigger latencies to resolve.

If there is a change here , it is more likely it is to a smaller 'building block' chiplet. If building with smaller blocks then more likely not getting 2x and 4x more peformance. Getting more performance than a laptop Max with a 'double die' but not chasing mega Nvidia dies either.

Apple used InFo-LSI for the Ultra's previously. That is limited to reticle size ( last slide deck I saw) so if your 'building block" gets too big than can't fit inside of reticle with even two 'blocks'. Let alone even more of those ever increasing size blocks. Decent chance here that Apple has blown past the limit and might get a monolithic version that is a bit les than 2x more. ( or move to more chiplet decomposition and get more than just a two , but not massively much more as will need to cover going 'down' from Ultra-zone as well as up.).

Decoupled from the Max volumes there is a serious cost issue here.

 

[deconstruct60]

i'm wondering if the next iteration of the mac pro will be delayed because of this: https://arstechnica.com/security/20...secret-encryption-keys-from-apples-mac-chips/

heres the actual paper: https://gofetch.fail

Likely not anymore than the rest of the M3 generation line up was. Probably very far off base if think the Mac Pro is going to get a solution for there that is any different to what has been rolled out for the M3/Pro/Max systems right now.

 

[ZombiePhysicist]

So a lot of things could get me to get a new Mac Pro and sadly I think the most likely outcome is one where I skip it.

I have been in an epic battle to try and get a 30+TB SSD to work on the mac. ANY ONE. Just any 30TB SSD that would boot over in this thread. TLDR, none seem to work (and one I found kind of works but write speeds are not what they should be) mostly because Apple has continuously gutted important abilities in macOS (they REMOVED many U.2 drive support, prevent booting from RAIDS, etc etc--all in process of the iphone'ification of the mac).

As such, I think what's likely is the next Mac Pro will just be an:
-M3 Ultra with
-PCIe 4 and thunderbolt 4

That said, if ANY of the following are present, I would probably jump:
-Support for 30+TB SSDs
-M3Extreme
-PCIe5 and thunderbolt 5

Otherwise, I would probably hold on to the 7,1 (trying to solve the 30+TB bootable SSD problem) and wait for the M4 version of the Mac Pro, which I think is more likely to bring at least PCIe5 and thunderbolt 5.

The big problem with waiting for the M4 Mac Pro is there may be NO M4 Mac Pro. Apple has destroyed this platform to such an extent and is so untrustworthy, every Mac Pro they release could be the last. It's so sad.

But anyway, I'll take the positive speculation today as something to be optimistic about. It's nice to have that little bit of hope.

 

[ZombiePhysicist]

Likely not anymore than the rest of the M3 generation line up was. Probably very far off base if think the Mac Pro is going to get a solution for there that is any different to what has been rolled out for the M3/Pro/Max systems right now.

You probably know better, but I would guess the lead times on these chips is over a year, so unless a revision was somehow provably extremely easy to do and also would be known not to impact the rest of the chip, it seems like any variations of the M3 wouldn't get a "fix". Perhaps the only exception is if this is a microcode doable patch? Of course, I would love to be wrong and be surprised.

 

[impulse462]

You probably know better, but I would guess the lead times on these chips is over a year, so unless a revision was somehow provably extremely easy to do and also would be known not to impact the rest of the chip, it seems like any variations of the M3 wouldn't get a "fix". Perhaps the only exception is if this is a microcode doable patch? Of course, I would love to be wrong and be surprised.

security is not my area of expertise (if you can call it that) in computer architecture but the exploit is based off a hardware prefetcher which predicts memory addresses that will be used in the near future. this one in particular looks at the content of memory on order to determine what to fetch which means data and memory addresses get mixed thus i'm not sure if this can be easily fixed with microcode since those are usually instruction set level operations(although someone pls correct me if im wrong). it seems disabling it would probably lead to worse performance.

edit: looks like the authors disclosed their findings to apple dec 23 so they definitely know about it by now

 

[ZombiePhysicist]

https://twitter.com/i/web/status/1773416604988248545

I'm not even sure what the N3B vs N3E stuff is, but seems to make a difference for when the Extreme might come out.

 

[deconstruct60]

You probably know better, but I would guess the lead times on these chips is over a year, so unless a revision was somehow provably extremely easy to do and also would be known not to impact the rest of the chip, it seems like any variations of the M3 wouldn't get a "fix". Perhaps the only exception is if this is a microcode doable patch? Of course, I would love to be wrong and be surprised.

The M3 has a bit to turn the "guess at what is a pointer" subsystem off. The M3 also arrive pretty late ( which most have completely blamed TSMC for. Similar to how it often was "all Intel's fault" if Apple didn't ship a new Mac for some people. ) . The short term fix is only not to guess.

That could have been present all along ( a relatively dangerous thing to do to guess at pointers anyway. ) but only there for debug/testing (and then fused on permanently for shipping). That switch/bit may just have been reclassified as a end user 'feature'.

But this is all 'behind' the instruction decode; not microcode fix coming. ( Microcode probably wouldn't help anyway because it is too easy to fool the heuristic that is guessing at what is a pointer or not, so code limited to the same size of space likely going to be any 'smarter' in a very substantive way. )

 

[deconstruct60]

security is not my area of expertise (if you can call it that) in computer architecture but the exploit is based off a hardware prefetcher which predicts memory addresses that will be used in the near future.

It is a " data memory-dependent prefetchers" DMP. There are other prefecthers that look more at the past code and past data usage rather than speculate on the data values themselves. For an if/then/else branch prediciting going down the 'then' branch rather than the 'else' branch doesn't introduce a new path of the program that wasn't there.

This is more like "oh that piece of data might be an address" ... go off load that piece of data before even see any code that even asks for it.

edit: looks like the authors disclosed their findings to apple dec 23 so they definitely know about it by now

But they likely were asking pointed questions before that date. Plus this is only 'build on top' exploit that was already known in M1.

 

[Regulus67]

... Apple has continuously gutted important abilities in macOS (they REMOVED many U.2 drive support, prevent booting from RAIDS, etc etc--all in process of the iphone'ification of the mac).

As such, I think what's likely is the next Mac Pro will just be an:
-M3 Ultra with
-PCIe 4 and thunderbolt 4

If Apple has removed many drivers etc, and it also is an issue on the latest Mac Pro. Even if Apple promised that users could fill the PCIe slots with add in cards of different types.
What leads you to think that it will release a new updated version, with an updated chip?

I am very sceptical, and judging by past history. Apple stopped updating Mac Pro hardware since the Trashcan 6.1.
They will probably wait until they have a completely redesigned Mac Pro.

The present Mac Pro was just a compromise, right? To complete the transition from Intel. But it still has the now absolutely overpowered Power Supply.
Which does not in any way fit into the power efficiency picture that Apple prides itself of.

 

[ZombiePhysicist]

If Apple has removed many drivers etc, and it also is an issue on the latest Mac Pro. Even if Apple promised that users could fill the PCIe slots with add in cards of different types.
What leads you to think that it will release a new updated version, with an updated chip?

I'm not sure I fully understand your question. Are you saying since they removed support for many U.2 drives, what makes me think there will be more support? If so, that's a fair question. Nothing really. Although U.2 and EDSFF are becoming such a common form factor, maybe some idiot at apple could wake up and add some support back in and maybe add some space for that form factor in the Mac Pro itself. And if that magic were to happen, my guess is they are very unlikely to propagate that support back to the intel side of things.

If you are getting at that it's not a likely thing, sadly, I have to agree.

I am very sceptical, and judging by past history. Apple stopped updating Mac Pro hardware since the Trashcan 6.1.
They will probably wait until they have a completely redesigned Mac Pro.

The present Mac Pro was just a compromise, right? To complete the transition from Intel. But it still has the now absolutely overpowered Power Supply.
Which does not in any way fit into the power efficiency picture that Apple prides itself of.

You may be right. And there are so many technology trends they have been ignoring (like EDSFF). If you look at the SSD space it's completely revolutionized by EDSFF form factors. Huge speeds. Huge capacities bigger than mechanical disks (256TB drives are out now).

When do you guess they might do a complete redesign? My guess is an M4 with PCIe5 and thunderbolt 5 might be a good moment to look at storage developments and integrate something like EDSFF.

If I start really dreaming, if you put 2 or 4 such NVMe PCIe5 slots in a RAID, you get absurd throughput. Like 56GB/s and all the sudden you can have radical new architecture where your storage works nearly as fast as ram and you could get rid of the distinction and just have a singular huge address space with smart caching.

I'm not holding my breath for any of that.

 

[avro707]

They don't do any internal components, or the Mac Pro at all in the .au Amazon store.

We are geo-blocked from buying anything from the other stores, very annoying.

 

[Regulus67]

When do you guess they might do a complete redesign? My guess is an M4 with PCIe5 and thunderbolt 5 might be a good moment to look at storage developments and integrate something like EDSFF.

I have no idea when they might do a redesign. Because it has the same chips as the Mac Studio.

Perhaps the WWDC will give us a better picture of what Apple is planning. In terms of AI, 3D or spatial computing.
And if those two areas demand a different hardware capability, that exceeds the Studio's thermal limit.
Then yes, it could cause Apple to present a new Mac Pro sooner, rather than in 4-6 years.

 

[Regulus67]

We are geo-blocked from buying anything from the other stores, very annoying.

I used Amazon UK for several years. And I just checked to confirm I can make orders there. Seems fine.
So if you really need some parts, you could try to order from a shop that still have the part(s).
But there is the added custom tax. So I would only advice it, if the part has a much lower price.

When I ordered my Pro W6800x Duo, from Amazon SE, it did not specify that it was on sale. So if you don't know the regular price, you might not even notice that the price has dropped for a couple of days.

 

[Axonite71]

I would love a Mac Pro with the latest chip for colour grading.

Considering Flanders Scientific have recently released 3 4K UHD QD-OLED Monitors with HDR/Dolby Vision, with 4 12-G SDI inputs; I would require slots for cards like the Blackmagic 8K Decklink or Decklink 4K Extreme 12-G to connect to one of those monitors.

Perhaps now Apple have got to a stage where they can design M-Series chips that provide professional grunt.

 

[ZombiePhysicist]

Apple today:

 

[ZombiePhysicist]

No vision, too many bozos:

 

[mode11]

To be fair NeXT was too much vision, and not enough bozos. It was the sort of company that would make G4 Cubes and 2013 MPs all day, and sell hardly any computers. The experience was the making of Jobs; when he returned to Apple he was the full package, with both vision and business sense (occasional indulgence aside).

I always saw the iPod / iTunes as a bit of a Faustian pact. It took them from near-bankruptcy to worldwide success - ostensibly great for the future of the Mac - but started the shift in focus to the consumer electronics market. Having perpetually been the 'loser' in the PC war, it must been validating to finally achieve dominance with personal devices.

 

[mattspace]

I always saw the iPod / iTunes as a bit of a Faustian pact. It took them from near-bankruptcy to worldwide success - ostensibly great for the future of the Mac - but started the shift in focus to the consumer electronics market. Having perpetually been the 'loser' in the PC war, it must been validating to finally achieve dominance with personal devices.

unfortunately they lost the most important lesson of iTunes & iPod - it wasn’t that it was “simple” for the sake of (visual) simplicity, but rather that its complexity was ordered in hierarchies that made its overall function cognisible.

I think the biggest problem with Apple has been a tendency to try to break hierarchy, and make everything flat. One could speculate as to what sociopolitical beliefs drive that urge, but now that every device has to be a equal peer, they all have to do the full range of complexity of task, despite often lacking screen size or input methods that make seeing enough, but not too much of the big picture possible. Before it was easy to explain - the Mac is where to get music, and where you organise it, the iPod is where you listen to it. I could explain that to an elderly pensioner, and they could then be independent with the tech - set things had set functions, tasks were atomised and structured hierarchically and chronologically in the acquire / organise / listen experience.

That’s really not the case with Apple Music, and the podcasts app, etc it’s all at the front and all demanding equal attention. But then you look at the sheer mediocrity of UI design from Apple these days, it’s fairly clear a large proportion of what comes out of the company is made by people who know tools, but not theory.

 

[mode11]

now that every device has to be a equal peer, they all have to do the full range of complexity of task, despite often lacking screen size or input methods that make seeing enough, but not too much of the big picture possible.

Perhaps part of the issue is that the march of technology has enabled every device to be a fully-fledged computer. Even the Apple Watch has a dual core 1.8GHz CPU. The earlier iPods were dependent on a Mac / PC to manage their contents, whereas it would feel weirdly restrictive nowadays to insist people manage iPhones in the same way.

Personally, I've never even been keen on laptops to do 'real work'. More due to ergonomics than processing power - I'm just much more productive with multiple screens, a full size keyboard, a mouse etc.

 

[mattspace]

Perhaps part of the issue is that the march of technology has enabled every device to be a fully-fledged computer. Even the Apple Watch has a dual core 1.8GHz CPU. The earlier iPods were dependent on a Mac / PC to manage their contents, whereas it would feel weirdly restrictive nowadays to insist people manage iPhones in the same way.

The problem isn't the CPU thugh, it's that iTunes was designed to be on a big screen, to lay out information in a structured way. Having recently set up a new iPhone for the first time in ~7 years, it's just the most unimaginably painful process when compared to being able to set up all the screens and apps in iTunes.

Personally, I've never even been keen on laptops to do 'real work'. More due to ergonomics than processing power - I'm just much more productive with multiple screens, a full size keyboard, a mouse etc.

Well yeah, so map that onto managing your music on a Phone, rather than on the bigger screen and management-centric UI of iTunes on a laptop.

 

[mode11]

I'm agreeing with you - I prefer larger displays etc. for ergonomic reasons too. My point was that modern devices do everything because they can, not necessarily because they should. But if, say, the iPhone insisted you use a PC / Mac to set it up, and Android phones didn't, people would criticise Apple for being behind the times / overly restrictive.

 

[ZombiePhysicist]

To be fair NeXT was too much vision, and not enough bozos. It was the sort of company that would make G4 Cubes and 2013 MPs all day, and sell hardly any computers. The experience was the making of Jobs; when he returned to Apple he was the full package, with both vision and business sense (occasional indulgence aside).

I always saw the iPod / iTunes as a bit of a Faustian pact. It took them from near-bankruptcy to worldwide success - ostensibly great for the future of the Mac - but started the shift in focus to the consumer electronics market. Having perpetually been the 'loser' in the PC war, it must been validating to finally achieve dominance with personal devices.

Next was the outright result of too many bozos at Apple forcing jobs out—-it shows the danger of it. The bozos caused next’s existence, and he had to fight an imperfect fight out of necessity (going into education wasn’t the “free will” choice it seemed but a lock from litigation with the bozos wanting to tie his hands). The bozos were that existential, a cancer that kicked out the host and wanted the host to fail and die. So it certainly set jobs back on his heels, forcing him to start imperfectly with many forced disadvantages.

During that time he got Pixar public. Too much vision wasn’t nexts problem. And he had a lot of business sense and did a lot of things he didn’t want to make next profitable, including getting rid of hardware and going enterprise, and making WebObjects successful. He was about to release a openstep for windows layer for windows as well (it was super impressive and let you fat binary compile any next app so it ran like a native app on windows—I used it, very impressive—and something he didn’t want to do, but did because it was needed for profitability and reach).

Next was too late to be the 3rd standard in the short time he had wrt hardware, but he learned to run it profitably even though it hurt to do so in the final few years. Despite all those disadvantages, he still got the company profitable in the end. Next was profitable the last 2 years as opposed to Apple which was bleeding money And nearly died. Not to mention, it ended up acquiring Apple in the end. All the best next guys took over, gutted the bozos, and saved the company and made it thrive. We’re basically using nextstep today, that’s how good the vision is.

That said, I understand your view, we just disagree. FYI I worked at next so I obviously have my biases.

 

[ZombiePhysicist]

I'm agreeing with you - I prefer larger displays etc. for ergonomic reasons too. My point was that modern devices do everything because they can, not necessarily because they should. But if, say, the iPhone insisted you use a PC / Mac to set it up, and Android phones didn't, people would criticise Apple for being behind the times / overly restrictive.

I agree with both of you. In perhaps a simple way, one could argue that UI has just gone to hell at Apple. As bad as iTunes was in the end, it would be like getting ice water in hell to be able to use it now instead of the atrocious Music.app and what it bastardized into becoming from iTunes.

 

[mattspace]

I'm agreeing with you - I prefer larger displays etc. for ergonomic reasons too. My point was that modern devices do everything because they can, not necessarily because they should. But if, say, the iPhone insisted you use a PC / Mac to set it up, and Android phones didn't, people would criticise Apple for being behind the times / overly restrictive.

True. I still think it's a ideological thing as well - there's been a broader breakdown in hierarchy as a tool within computing - whether it's attempts to do away with filesystems, or the systemic violence against menu systems (that Apple is filled with people who don't understand why a notch in a menubar is deeply bad, the graphic designers not understanding that menus are supposed to be behind the proscenium arch of the menubar itself for structural / hierarchal / functional reasons, not just because it controls where the dropshadow falls etc.) to the extent that they made the iPad use a cellphone UI, despite all the use-cases being desktop computer based.

Modern Apple seems to really cargo cult the style and retro chic of old Apple without really caring about the logic behind the way it worked.

 

[mode11]

That said, I understand your view, we just disagree. FYI I worked at next so I obviously have my biases.

I don't have a hard and fast opinion on NeXT - you were there, and know a lot more about it than me. I think it's fair to say though that at least initially, Steve was off the chain in terms of being able to design his ultimate computer. I've read he agonised over the shade of blue that the factory robots were to be painted. The NeXT Cube in particular was a piece of computer art. It might even have been good value for money, but was priced so high that it was unlikely to find many buyers. It seemed like it was designed first, then a market sought afterwards - based primarily on who could afford it. Though you say NeXT had some kind of non-compete clause with Apple markets?

Steve swallowing such bitter pills as giving up on hardware and creating the OpenStep layer for Windows, were the kind of thing I was referring to. Learning these hard lessons ultimately made him a better leader. He had a bit of history with no-expense-spared designs - like the over-priced Lisa. He only took over the lower-cost Macintosh project from Jef Raskin when the Lisa bombed. But overall, he was clearly brilliant and achieved great things throughout his entire career (running Pixar as a side gig!). And although part of Apple's problem is they've become swollen on iPhone cash, losing Steve's vision was a big deal; I've barely watched a keynote since.

Modern Apple seems to really cargo cult the style and retro chic of old Apple without really caring about the logic behind the way it worked.

I think this is exactly true. In the early days UI principles were king. Now it does seem more about style. The annual OS releases are likely partly to blame, causing a constant 'need' for change. As is trying to create a UI that stylistically spans macOS and iOS.