Apple Transitioning Macs to Custom Chips With 'Incredible' Performance Starting Later This Year

[chucker23n1]

He literally said a 27“ 5K display had „a LOT more pixels“ than a 14“ 5K display - which is rubbish. Hence my reaction.

Yeah, but I think it's more of a mistake.

Here's the original statement:

"Apple has delivered phone GPUs - cool that they can drive a 14" screen in an iPad pro at 5K - but I haven't used a 14" screen on a computer since 1992."

The iPad Pro has nowhere near a 5K display, though. Maybe they misunderstand what the "K" nomenclature refers to (horizontal pixel count).

 

[Nütztjanix]

Yeah, but I think it's more of a mistake.

Here's the original statement:

"Apple has delivered phone GPUs - cool that they can drive a 14" screen in an iPad pro at 5K - but I haven't used a 14" screen on a computer since 1992."

The iPad Pro has nowhere near a 5K display, though. Maybe they misunderstand what the "K" nomenclature refers to (horizontal pixel count).

Which would be strange given that he said he‘s doing 3D art for years.
If I had to guess I‘d say he got carried away defending his point of view.

 

[chucker23n1]

If I had to guess I‘d say he got carried away defending his point of view.

It's been known to happen to me.

 

[SteveW928]

Take a look at the Levono TR Pro systems that ship in September - that is what the 7,1 should have been.

I don't even know what that is or what is special about it, but don't you think it is a bit odd to compare a machine released in 2019 to one coming in the future?

Now, you might have a point that Apple won't likely keep updating the Mac Pro at the rate of the rest of the industry, though we don't know that, it would just be based on past experience.

I get it, though, the 7,1 isn't for us. That's a bummer and I too wish they'd have something between that and my mini w/eGPU.

I don't think that Apple will deliver performance because of Tim Cook. Tim is a bean counter, and doesn't see the need for performance. If you look at his stewardship of Apple, the Mac division has at best, been neglected. 2,000+ days between the launch of the 6,1 to the 7,1. How many obsolete parts are still shipping with current macs?

I'm not sure if I ever expect Apple to be speedy to release every new thing, but they've done a way better job in the last couple of years. And, I think we've kind of answered that question... at least I hope. I too was saying either Apple had abandoned the Mac (and fearing that), or it was something like trying to 'hang on' due to rapid growth, and placing all their priorities on the primary product. I'm now pretty sure it was the latter.

The Mac mini is shipping with 8th Gen CPUs (6 cores/6 threads or 6 cores/12 threads, integrated graphics for all), with the iMacs, only the i9 and iMac "Pro" have multi-threaded CPUs. The rest of them are 6 core/6 thread budget i5 CPUs, with Polaris based video cards, although you can be only 1 generation back with a throttled Vega GPU for the iMac "Pro".

While I realize there are probably newer components, what should be in my mini? Would there be substantial gain?

The iMac absolutely needs a refresh, and probably the iMac Pro as well. Remember you can add eGPUs if you need other GPUs.

In spite of the current Mac lineup, the True Believers are convinced that Apple is suddenly going to design a series of CPUs that will out perform the top of the line CPUs from both Intel & AMD that Apple was too cheap to put in their own lineup. They also believe that Apple will suddenly develop GPUs that will outperform both AMD and Nvidia, when 2080ti performance will be the base line for performance in the next 6 months.

I suppose it is possible they aren't going crazy on the existing lineup because they have these new machines coming. But, I don't think they haven't updated because they are 'too cheap' to put a newer chip in.

As for GPUs, I'm not sure we're saying they'll have something better than 2080ti (or top of AMD's line), but that they'll have something better than Intel iGPUs. We'll probably add an AMD (or they will) if we need more.

I guess *some* are speculating Apple is going to be a major force in GPUs, but yeah, I wouldn't hold my breath on that one.

Apple has delivered phone CPUs, maxing out at 2 big cores and 2 little cores - Intel is delivering 28 core/56 thread CPUs today. AMD is delivering 16 core/32 thread CPUs at the consumer level, 64 core/128 thread CPUs at the HEDT level, and multiple socket 64c/128t CPUs at the server level - today.

You're allowed to be skeptical, but I certainly don't understand why you think they won't be able to scale up the phone/tablet tech to desktop application. I'm not a chip designer, so I can't weigh in too heavily. But, it seems actual chip-designers here and elsewhere don't see a big issue, so I'll probably go with their experience for now.

Apple has delivered phone GPUs - cool that they can drive a 14" screen in an iPad pro at 5K - but I haven't used a 14" screen on a computer since 1992. I have been using a dual screen setup for well over 15 years - I am looking at going to dual 4K monitors at Xmas. Neither you nor I know if Apple can/will support that - and keep in mind, 1440p 240Hz is mid-tier right now, never mind two years from now.

OK, I guess this has been picked apart a bit before I got to it, so I won't go there. But, again, why would you think Apple couldn't apply their GPU tech to dual displays or higher resolution ones? Even Intel iGPUs can run high-res displays to a point.

The next Mac Pro won't be competing with a TR3995XT based system - it will be competing with a Zen 4 or Zen 5 AMD (5nm) based system, or a Golden Cove (10nm - they will be there) based Xeon system.

Yes, Apple will have to update it some day if they intend to keep the Mac Pro competitive.

AMD based systems are now moving from DIY, enthusiasts, and boutique vendors to tier one OEMs. It isn't just those Levono's I mentioned earlier - AMD has an entire range of desktop APUs that for now, will only be available to OEMs (8 cores/16 threads with Vega 8 graphics @65watts would fit very nicely in a Mac Mini; take the same CPU and add a dedicated GPU and that iMac is now a workhorse.) This is why I say that the move to ARM is about control, not performance.

Yes, we'll have to see. Sure, I suppose they *could* no make a better Mac mini. But, I think you're missing a point here similar to your criticism of the base Mac Pro. The Intel iGPU/AMD 580 are perfectly competent GPUs for most people, which is why they are the base. You can easily add an eGPU, or GPUs into the Mac Pro if you need more. Someone buying the Mac Pro for scientific computing might not want more than a 580. Someone using the mini in a server rack won't want more than the Intel iGPU.

The iMac, I think everyone agrees, is well beyond due for an update. If they bump the mini, then great, but it isn't that far behind. What we both want is an xMac of sorts. You'd want it to be much more a mini-tower, and I'd just want better thermals. I doubt we'll get it... BUT there is the potential for a pretty cool Apple silicon machine at some point *IF* the software we need is available (that's the part that has me concerned).

 

[pasamio]

I don't think that Apple will deliver performance because of Tim Cook. Tim is a bean counter, and doesn't see the need for performance. If you look at his stewardship of Apple, the Mac division has at best, been neglected. 2,000+ days between the launch of the 6,1 to the 7,1. How many obsolete parts are still shipping with current macs?

Whilst I feel calling Tim a bean counter is perhaps a little harsh, he is indeed an operations focused individual and I can absolutely see him making the trade off that Apple isn't necessarily doing performance for the sake of performance. I can see some of the slowness to upgrade the Pro line as part of paying down a larger order of CPU and similar SKUs over a period of time to get a volume discount out of things. Does it need to be updated every year? Does it hurt them to not update every year? How many more Macs do they sell by doing a new release and does that outweigh the cost efficiency of selling something slightly longer?

Apple are capable of timely updates to their hardware, we see yearly to 18 month mobile device updates (though there are lines left behind for longer like iPad Mini or iPod Touch) but the desktop doesn't see this as much in recent years. The desktops used to be on a similar cycle but it slowed down. Part of me actually wonders if that is Tim's bean counting at work and they ran the numbers: it's cost effective for Apple to wait over two years to update these platforms.

Yes, we'll have to see. Sure, I suppose they *could* no make a better Mac mini. But, I think you're missing a point here similar to your criticism of the base Mac Pro. The Intel iGPU/AMD 580 are perfectly competent GPUs for most people, which is why they are the base. You can easily add an eGPU, or GPUs into the Mac Pro if you need more. Someone buying the Mac Pro for scientific computing might not want more than a 580. Someone using the mini in a server rack won't want more than the Intel iGPU.

I think this is in fact part of the challenge that Apple could make a better Mac Mini, they could have made a better Mac Pro as well however they made a product they knew they could sell to enough people to make it worth their while. To connect this to the earlier quote, the bean counters count the beans and if the beans don't add up then they don't do those things. It doesn't make as much fiscal sense for Apple to do what I think we'd all love which is upgradeable Macs available cheaper, though I'm sure they'd sell the question is would they sell enough.

Which brings us back to the Apple Silicon shift. It's going to impact upgradeability because I don't see Apple shifting on their SOC architecture. What makes my Mac Pro 5,1 still usable a decade later (with the Dashboard still) is that it was a solid machine at the time and it has an upgradeable architecture. I'm not sure we'll see that in a future Mac Pro which makes the 7,1 an interesting beast and maybe it'll be the end of an era. They already started with the Mac Pro 6,1 and iMac Pro trying to go down a path, maybe that will be it but in a Mac Pro 7,1 style case. Maybe Apple will sell logic board upgrades in the future and since it'll almost all be soldered on that will be the path but I don't think that'd be acceptable to folk. Time will tell what Apple have in mind for the Mac Pro and I can't help but think it won't be the cheaper version we all want.

 

[Nütztjanix]

In other news, Intel is (again) delaying chips due to unsatisfactory yields.

(Intel Delays 7-Nanometer Chips Until Late 2022 or Early 2023)

 

[chucker23n1]

Which brings us back to the Apple Silicon shift. It's going to impact upgradeability because I don't see Apple shifting on their SOC architecture. What makes my Mac Pro 5,1 still usable a decade later (with the Dashboard still) is that it was a solid machine at the time and it has an upgradeable architecture. I'm not sure we'll see that in a future Mac Pro which makes the 7,1 an interesting beast and maybe it'll be the end of an era. They already started with the Mac Pro 6,1 and iMac Pro trying to go down a path, maybe that will be it but in a Mac Pro 7,1 style case. Maybe Apple will sell logic board upgrades in the future and since it'll almost all be soldered on that will be the path but I don't think that'd be acceptable to folk. Time will tell what Apple have in mind for the Mac Pro and I can't help but think it won't be the cheaper version we all want.

I don't really see Apple making a flip-flop (fairly non-upgradeable 2013 Mac Pro, then a 2019 Mac Pro quite like the old 2003 one) only to go back to the 2013-like design again. Presumably, someone on the 2019 Mac Pro design team knew about the Apple Silicon transition (though not most of the team, I'm guessing, just enough to keep it in the back of their minds), and made sure they don't make any design decisions that don't make sense in the ARM future.

I therefore find many of these discussions a little silly. E.g., does anyone actually think Apple develops the MPX spec, and then immediately afterwards decides, oh hey, we don't need PCIe GPU cards at all?

I don't find that very plausible.

 

[chucker23n1]

In other news, Intel is (again) delaying chips due to unsatisfactory yields.

(Intel Delays 7-Nanometer Chips Until Late 2022 or Early 2023)

Complacency and hubris from the early 2010s, with long-term consequences.

 

[pasamio]

I don't really see Apple making a flip-flop (fairly non-upgradeable 2013 Mac Pro, then a 2019 Mac Pro quite like the old 2003 one) only to go back to the 2013-like design again. Presumably, someone on the 2019 Mac Pro design team knew about the Apple Silicon transition (though not most of the team, I'm guessing, just enough to keep it in the back of their minds), and made sure they don't make any design decisions that don't make sense in the ARM future.

I therefore find many of these discussions a little silly. E.g., does anyone actually think Apple develops the MPX spec, and then immediately afterwards decides, oh hey, we don't need PCIe GPU cards at all?

I don't find that very plausible.

I'm not saying they're going to throw away MPX, I think that will still exist what I'm saying is that I expect the CPU to be soldered onto the logic board like every other product as an SOC. I'm curious about how they'll handle memory in this architecture but I suspect that might be user upgradeable. They've already documented that they're changing the memory architecture for PCIe with the new chips so I think it's certain they'll be upgradeable on that angle even if the logic board ends up not being fully upgradeable (maybe memory upgrades but probably not CPU).

 

[chucker23n1]

I'm not saying they're going to throw away MPX, I think that will still exist what I'm saying is that I expect the CPU to be soldered onto the logic board like every other product as an SOC.

Right. I gave that as an example of "why make the 2019 Mac Pro fairly upgradeable only to backtrack on that one revision later".

The CPU might be soldered indeed. I don't know that Apple wants to go through the trouble of devision a socket or slot mechanism for a niche product, when that's rarely all that useful anyway. Do you want to swap for a newer CPU? Oops, that one requires a different socket. Oops, that one works with the socket, but really benefits from better RAM. Oh, and don't you want a new mainboard while you're at it? Do you want to swap for a higher-end CPU of the same generation? OK, but at the point you're realistically making that call, it's already quite an old CPU, and a poor bang for the buck.

Replacing the entire logic board fixes those issues.

I'm curious about how they'll handle memory in this architecture but I suspect that might be user upgradeable.

Yeah.

(Even the 2013 Mac Pro could have its RAM upgraded.)

 

[Nütztjanix]

Right. I gave that as an example of "why make the 2019 Mac Pro fairly upgradeable only to backtrack on that one revision later".

The CPU might be soldered indeed. I don't know that Apple wants to go through the trouble of devision a socket or slot mechanism for a niche product, when that's rarely all that useful anyway. Do you want to swap for a newer CPU? Oops, that one requires a different socket. Oops, that one works with the socket, but really benefits from better RAM. Oh, and don't you want a new mainboard while you're at it? Do you want to swap for a higher-end CPU of the same generation? OK, but at the point you're realistically making that call, it's already quite an old CPU, and a poor bang for the buck.

Replacing the entire logic board fixes those issues.



Yeah.

I honestly don't see Apple making their CPUs/SoCs available to purchase on a standalone basis - which would be necessary for DIY upgrades. I also don't think they are/will be made for any kind of socketed use, given Apple's aversion to socketed components (be it for the sake of thinness, reliability of connection or signal quality).

 

[chucker23n1]

I honestly don't see Apple making their CPUs/SoCs available to purchase on a standalone basis - which would be necessary for DIY upgrades. I also don't think they are/will be made for any kind of socketed use, given Apple's aversion to socketed components (be it for the sake of thinness, reliability of connection or signal quality).

Yup.

My point is even for machines that do let you swap the CPU later, it's not as great as it sounds.

I think you will get to choose different CPUs for the Mac Pro, but only at purchase.

 

[aid]

I'm not saying they're going to throw away MPX, I think that will still exist what I'm saying is that I expect the CPU to be soldered onto the logic board like every other product as an SOC.

I think for the MacPro; a more interesting solution would be the Apple Silicon SOC (CPU, GPU & RAM) to be hosted on a daughter card — with the MacPro able to accept multiple such daughter cards to increase performance.

 

[ThrowerGB]

I remember back to the transition from the G5 to Intel. In fact I still have a beautiful looking G5 next to my desk. The transition was amazingly smooth because of the way the operating system was architected. I expect the same thing this time. And even further back there was the smooth transition from Motorola chips to RISC. I loved getting my first PowerMac.
They said the developer machines they'll make available will be running Big Sur on an ARM chip. That's Big Sur folks! It's already out in Beta, so don't expect a giant catastrophe. They probably have a lot of middleware running in iPhoneOS and iPadOS that's shared or sharable with Big Sur.
Jobs would prioritize the user experience rather than the hardware technology. Apple didn't want to be in a commodity chip race with PC makers who also used Intel chips. Of course there's a lot of excitement, trepidation and anticipation about about transitioning Macs to ARM, but let's remember it's the user experience that will sell these products not necessarily the hardware in the box.

 

[pdoherty]

I'm not saying they're going to throw away MPX, I think that will still exist what I'm saying is that I expect the CPU to be soldered onto the logic board like every other product as an SOC. I'm curious about how they'll handle memory in this architecture but I suspect that might be user upgradeable. They've already documented that they're changing the memory architecture for PCIe with the new chips so I think it's certain they'll be upgradeable on that angle even if the logic board ends up not being fully upgradeable (maybe memory upgrades but probably not CPU).

Why the soldering of CPUs? Are CPU sockets really so pricey? What does it add, 50 cents, to the cost of the Mac? Then people can upgrade CPUs and extend the life... oh wait, I get it.

 

[SteveW928]

Whilst I feel calling Tim a bean counter is perhaps a little harsh, he is indeed an operations focused individual and I can absolutely see him making the trade off that Apple isn't necessarily doing performance for the sake of performance. ...

I imagine they calculate everything out pretty carefully. I think the main problem, though, has been a lack of resources as they rapidly grew and focused on other things.

My concern about the 'bean counter' aspect, though, is that much like car companies and racing, bean counters aren't often great at seeing benefits that don't fit neatly into the spreadsheets.

It's kind of like going to the moon. These things often are money-losers with HUGE long-term payoffs. Hopefully Apple still recognizes that, because Jobs certainly did. Apple made massive long-term payoff efforts in the past.

Which brings us back to the Apple Silicon shift. It's going to impact upgradeability because I don't see Apple shifting on their SOC architecture. What makes my Mac Pro 5,1 still usable a decade later (with the Dashboard still) is that it was a solid machine at the time and it has an upgradeable architecture. I'm not sure we'll see that in a future Mac Pro which makes the 7,1 an interesting beast and maybe it'll be the end of an era. They already started with the Mac Pro 6,1 and iMac Pro trying to go down a path, maybe that will be it but in a Mac Pro 7,1 style case. Maybe Apple will sell logic board upgrades in the future and since it'll almost all be soldered on that will be the path but I don't think that'd be acceptable to folk. Time will tell what Apple have in mind for the Mac Pro and I can't help but think it won't be the cheaper version we all want.

I guess my question, though, is if Apple had done a good job of keeping current machines in the pipeline, how many go after CPU upgrades or that kind of thing, vs just buying the next model and selling or passing-down the existing one?

Upgrading RAM is one thing. Adding a GPU, another. But, beyond that, I think there is little benefit outside certain high-end segments, or hobbyists. I don't think Apple really sees themselves in serving the 'machine builder' type hobbyists at all.

Yes, maybe there will be some kind of CPU-card or logic-board swaps. I doubt they will make a big enough line of CPUs for that kind of swap to even make much sense. Remember that the existing lineup of Intel CPUs comes from Intel having to serve a vast array of markets. Apple will probably just make enough different CPU variants to serve their product lineup. (ie. maybe a few different core variants, and then a few speed variants in the high-end due to 'binning').

I therefore find many of these discussions a little silly. E.g., does anyone actually think Apple develops the MPX spec, and then immediately afterwards decides, oh hey, we don't need PCIe GPU cards at all?

I don't find that very plausible.

I'm not even sure where that idea came from in the first place. I think there were some points made about Apple's GPUs outperforming the iGPU options, meaning Apple would include their own GPUs in the new machines. And, then somehow that turned into Apple not using dGPUs anymore.

I've seen some speculation that Apple might be able to be competitive with Nvidia or AMD in GPUs at some point, but I think that is just pure speculation.

Why the soldering of CPUs? Are CPU sockets really so pricey? What does it add, 50 cents, to the cost of the Mac? Then people can upgrade CPUs and extend the life... oh wait, I get it.

It probably makes things more reliable and cheapens the product process (not so much component costs, true). Also, as I was saying above... how many people do that? And, will Apple even create a very big lineup of CPU options? I kind of doubt both.

 

[chucker23n1]

I'm not even sure where that idea came from in the first place. I think there were some points made about Apple's GPUs outperforming the iGPU options, meaning Apple would include their own GPUs in the new machines. And, then somehow that turned into Apple not using dGPUs anymore.

I've seen some speculation that Apple might be able to be competitive with Nvidia or AMD in GPUs at some point, but I think that is just pure speculation.

Right, I think the speculation in that area went into overdrive. Part of it is slides from WWDC that seemed to suggest that ARM Macs don't support heterogenous memory. If true, I indeed don't see how most discrete GPUs from third parties would work.

But I'm guessing it was poorly phrased and/or taken out of context.

 

[Nütztjanix]

Why the soldering of CPUs? Are CPU sockets really so pricey? What does it add, 50 cents, to the cost of the Mac? Then people can upgrade CPUs and extend the life... oh wait, I get it.

It's not so much about the component cost of a socket, but more about production cost (board complexity, added components to route, solder) and integrity of signal. A few years ago, I read a piece of an electronics engineer who showed that, by soldering the RAM directly onto the logic board, the Retina MacBook Pro was able to operate that RAM within an extremely minimal margin from the theoretical maximum of speed/timing/throughput. If you eliminate non-fixed connections, you're much more likely to get optimal signal quality.

 

[ericwn]

macbook is dead to me, i dont want an enlarged ipad

This post aged extremely well, the product is basically dead now…

 

[ericwn]

Well, this is my last Mac. It's been a long time coming, but this is the final straw for me.

Did you buy that 14” MAX by now?

 

[venom600]

Did you buy that 14” MAX by now?

I loathe you so, so much. It’s literally a 14” Max and it arrives in mid December. ?

I was hoping that comment would be forgotten. There are a lot of reasons that I changed my mind, but the biggest one is that bootcamp has no future. Windows 10 has three years of support left and you can’t run Windows 11 in bootcamp without hacks. So I’m left with a virtual machine as the supported option, and Windows on ARM is WAY better than I thought it’d be.

Combine that with the 16” Macbooks overheating when you attach a monitor, the offer of $1500 I got from Apple for a trade in on my two year old computer, and being blown away by the battery life I got on the Macbook Air I got my mom, and I was sold.

 

[ericwn]

I loathe you so, so much. It’s literally a 14” Max and it arrives in mid December.

I was hoping that comment would be forgotten. There are a lot of reasons that I changed my mind, but the biggest one is that bootcamp has no future. Windows 10 has three years of support left and you can’t run Windows 11 in bootcamp without hacks. So I’m left with a virtual machine as the supported option, and Windows on ARM is WAY better than I thought it’d be.

Combine that with the 16” Macbooks overheating when you attach a monitor, the offer of $1500 I got from Apple for a trade in on my two year old computer, and being blown away by the battery life I got on the Macbook Air I got my mom, and I was sold.

Thanks for the context, I came across your old post from browsing the thread and then just decided to look up your more recent posts. All the best with the new machine - I’m sure it will be pretty awesome!

 

[Christopher Kim]

I loathe you so, so much. It’s literally a 14” Max and it arrives in mid December. ?

I was hoping that comment would be forgotten. There are a lot of reasons that I changed my mind, but the biggest one is that bootcamp has no future. Windows 10 has three years of support left and you can’t run Windows 11 in bootcamp without hacks. So I’m left with a virtual machine as the supported option, and Windows on ARM is WAY better than I thought it’d be.

Combine that with the 16” Macbooks overheating when you attach a monitor, the offer of $1500 I got from Apple for a trade in on my two year old computer, and being blown away by the battery life I got on the Macbook Air I got my mom, and I was sold.

Lol, just want to say good on you for being willing to pivot and change your position after additional info was out that changed the initial decision making process.

Your rationale and line of reasoning totally makes sense. I'm loving my 14" M1 Pro that just arrived last week!

 

[alien3dx]

It's not so much about the component cost of a socket, but more about production cost (board complexity, added components to route, solder) and integrity of signal. A few years ago, I read a piece of an electronics engineer who showed that, by soldering the RAM directly onto the logic board, the Retina MacBook Pro was able to operate that RAM within an extremely minimal margin from the theoretical maximum of speed/timing/throughput. If you eliminate non-fixed connections, you're much more likely to get optimal signal quality.

if production cost , should be reduce price but apple got awesome price hehe

 

[venom600]

Lol, just want to say good on you for being willing to pivot and change your position after additional info was out that changed the initial decision making process.

Your rationale and line of reasoning totally makes sense. I'm loving my 14" M1 Pro that just arrived last week!

I ended up cancelling the order because I found an identically specced 16" at the Apple Store. So far it has blown away my expectations.