The new Mac Pro is just around the corner. It's past time for speculation. When they release the Mac Pro, then it will be time for sales and design analysis.
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 Intel may be the only way to save the Mac Pro
- Thread starter Canubis
- Start date
- Sort by reaction score
That's not it, there was no mention of Intel there.
In the WWDC 2020 Keynote, Tim Cook said "In fact, we have some new Intel-based Macs in the pipeline." Hear it here:
After the WWDC in 2020, Apple introduced the 2020 27" iMac which was Intel based, and that's it.
Maybe got their wires crossed with what Tim Cook said at WWDC 2020 about there still being more Intel Macs in the pipeline?
Let me elaborate a bit more on that...
I am not the only one that notices Apple is trying to do everything with least amount of ingredients possible. You can see it in the Mac OS for example, it has so many assets just copy pasted from iOS and iPadOS.
With this observsation in mind, thinking that Apple is gonna keep maintaining x86 binaries just because of one niche device is silly, in my opinion.
Wow... just wow.... The amount of disconnect here is just.... wow
First, Apple publicly went on an apology tour and admitted that the 2013 trashcan was a failure, so they released the 7,1 to make up for that.
Furthermore, look what Timmy did in 2012:
Stop drinking the kool-aid
Aside from the fact macOS is my favorite OS and I don’t wanna deal with Windows anymore?
Plus I know a lot of the developers in my data center would love having macOS instead
That MacWorld author is utterly ignorant. If Apple were to make another x86 Pro (it won't), they'd be far better off with AMD Epycs. Intel is hardly their "only hope".
However all the repetitive talk about how AS' "architecture" doesn't permit expandable memory, or PCIe slots, or even GPUs is just ignorant. The fact that current chips don't support those things says nothing about Apple's ability to produce future chips that will. It's not even that challenging, as design work goes! Software integration, making the OS smart about NUMA characteristics of larger multichiplet designs, etc. - that's where the real work is likely to be. And of course the basics of scaling up the chip's uncore to meet requirements of many more cores is a challenge, though not a huge one - it's been solved in various ways by at least a half dozen other chip designers already.
BTW, there will never be a "dual Ultra" configuration. It's not feasible. You might however have something extremely similar - a 4x Max configuration with an I/O+switch chiplet in the middle, somewhat like AMD's design. They could also design the M3 Max chip to tile on two edges, though that might be problematic in that there may well not be enough shoreline on the other two edges to support enough I/O- all that PCIe and especially memory.
However all the repetitive talk about how AS' "architecture" doesn't permit expandable memory, or PCIe slots, or even GPUs is just ignorant. The fact that current chips don't support those things says nothing about Apple's ability to produce future chips that will. It's not even that challenging, as design work goes! Software integration, making the OS smart about NUMA characteristics of larger multichiplet designs, etc. - that's where the real work is likely to be. And of course the basics of scaling up the chip's uncore to meet requirements of many more cores is a challenge, though not a huge one - it's been solved in various ways by at least a half dozen other chip designers already.
BTW, there will never be a "dual Ultra" configuration. It's not feasible. You might however have something extremely similar - a 4x Max configuration with an I/O+switch chiplet in the middle, somewhat like AMD's design. They could also design the M3 Max chip to tile on two edges, though that might be problematic in that there may well not be enough shoreline on the other two edges to support enough I/O- all that PCIe and especially memory.
There's a far better argument against an AS Mac Pro, made far less often, that it's too niche a product to be worth Apple's time.
Perhaps. You can argue about the value of halo products endlessly - nobody really knows the truth, and nobody knows what Apple's thinking is on it.
But this argument forgets that the current product line is not the only gear Apple needs chips for. They also need to design for their AR/VR product, and wherever that's going in the future. And those products may simply demand too much GPU for monolithic chips even on the latest process nodes. In that case, Apple will have to disaggregate CPU and GPU anyway. That brings them much of the way towards what they need for the Mac Pro anyway.
Perhaps. You can argue about the value of halo products endlessly - nobody really knows the truth, and nobody knows what Apple's thinking is on it.
But this argument forgets that the current product line is not the only gear Apple needs chips for. They also need to design for their AR/VR product, and wherever that's going in the future. And those products may simply demand too much GPU for monolithic chips even on the latest process nodes. In that case, Apple will have to disaggregate CPU and GPU anyway. That brings them much of the way towards what they need for the Mac Pro anyway.
Yeah, there's no reason to put macOS in a data center.
This is what I keep coming back to. Sales of the Mac Pro have been a tiny fraction of the Macs Apple sells, and in many ways Apple has been its own enemy here. There were consistent updates to the Mac Pro from 2006-2010. Since then it's been a matter of uneven updates in both timing and quality, and it doesn't lead to potential buyers believing that it makes sense to even be in the Mac Pro market.
Sticking with the existing lineup minus the Mac Pro, keeping the Mac Studio updated on a regular basis, makes a lot more sense for a lot more users.
Apple can afford to hand off that super high end workstation niche to others–in many ways they did that years ago anyway.
Don’t forget that Mac Pro is for a VERY niche audience. It’s like a drop in an ocean from revenue stand point. It’s not for profit like iPhone, iPad, other Macs, etc. It’s more for showcasing what Apple is capable of. So catering towards Intel enthusiasts and losing face doesn’t make any sense at all imho.
Especially given the problems with Apple's Mac Pro line has been twofold: one, the question of what makes it worth buying, and secondly, the question of whether they're interested in supporting it. The GPU updates were welcome, but skipping two Intel CPUs while the new model is still off in the unknown future starts feeling like 2023 is the same as the last two Mac Pro iterations, where the lack of timely hardware updates was probably a bigger issue in killing the line than Apple's own missteps with the hardware design. (I was interested in a tube Mac Pro. I was going to get the rev 2 model once they ironed out the kinks of a new design... and then that machine never came out. So I got a 5,1 instead. And I'm sure lots of people just took Apple's radio silence on the machine as a reason to jump ship entirely, like my company at the time did.) I get that at one point, Apple probably thought by 2022 they'd have that new Apple Silicon machine out, but it doesn't really excuse leaving pro customers without good options... again.
Slap the new Intels in, produce a 7XXX AMD module, and call it a day. They'd earn a lot more goodwill than they would people upset they'd dare update an outdated lineup while people wait for the new machine. And it'd make sense to keep around as a legacy SKU for the people who want Intel compatibility even when the new Apple Silicon machine comes out (especially as it's almost certainly not going to be as flexible.)
Why can't macOS be in the data center? I work in a US federal government data center and we have a dedicated macOS team, our executives use macOS, and we are considering issuing out Apple Silicon Macs to our developers and end users since we already use iPhones as our government furnished work phones.
I think that's the point of AS though. To support those things would be antithesis to the AS philosophy.
The philosophy of AS is to cram as much as possible onto the die, and what cannot be crammed onto the die has to be as close to the SoC as possible, all to maximize channels/bandwidth and overall speed.
Unified memory is about putting the memory controller and memory allocation functionality on the SoC itself, with the actual DRAM chips directly next to the SoC. Likewise, the SoC includes the SSD controller so that the off-die SSD only needs to be storage NANDs without any controller.
To support DDR DIMMs or M.2 SSDs would be to surrender that control back to an off-die component, and thereby introduce the legacy bottlenecks and slowdowns. Apple won't do it.
All that is a long way to say, I think whatever they call the Mac Pro will look a lot more like the Mac Studio, and not like the Mac Pro we knew in the past.
The best predictor of what Apple will introduce is to simply look at what "I" bought most recently. Since I purchased a 2019 Intel Mac Pro, they will be forced to make "my" purchase obsolete ASAP!
It's not a matter of "can't," it's a matter of "there's no reason for it."
You quoted the first part of my post, where I raised that argument, but you avoided the second part, where I explained why that may not matter. If some of the major design goals of the Pro align with major design goals of the AR/VR device, then the incremental cost to design a big multichip AS Mac Pro may not be so high. Note that the two product lines clearly do NOT need the same chips - but they may have certain needs in common, like, separate GPUs.
I forgot to cover that part. I'd think that the power consumption/size/weight demands of an XR device would preclude being able to have separate chips for those purposes and that Apple's working toward a super integrated SoC lines up much better in that area.
Apple's only "philosophy" is to build better, more efficient devices. They will do whatever is expedient. In 2020 and 2021, that meant large monolithic chips, giant slow+wide cores, wide and sometimes fast memory busses, etc. In 2022 it expanded a bit to include two chips linked with InFO. They didn't make those choices because of some theoretical philosophy, they made them because they were smart and they *worked*.
Some of these decisions were less about technology and more about improving their profitability. Sometimes, as in the SSD controller incorporated into the M1/2, it was a mix.
In 2023, for some products, the numbers may well favor a different design. There is not likely even one person at Apple stupid enough to reject that just because it doesn't match some "philosophy".
You are wrong about unified memory. The heart of that technology is making the CPU and GPU cores coequal customers for memory access and bandwidth, and putting everything into one uniform address space. To "surrender control back to an off-die component" barely matters (and in fact, the RAM is always off-die anyway, just not always off-package). There's no doubt that a large memory system able to address, say, 4TB of RAM would have higher latency than current M1/2s do. But not likely enough to be significant. Going to a tiered system with some HBM is one way you could more than make up that difference.
Similarly, M.2 SSDs are not really "legacy" and would not introduce significant slowdowns. There are some PCIe4 SSDs out there with performance as good or better than Apple's, and Apple could most certainly support them, though I suspect that they'd still sell their own too (why not?).
That's a good argument. If everything stays in a head-worn device forever, then it's probably a correct argument. I'm not sure that's the case though. I can imagine a future with extremely light low-powered headsets and belt-worn or tabletop devices wirelessly connected (60Ghz, maybe). I don't think anyone knows enough to call that one yet.
Regardless - even if the end game is a single chip in a headset, they want way more processing power than they can have today. It may be that building something that's great for a Mac Pro is a good way to get started prototyping the xR future.
The more important indicator as it relates to GPU’s, though, is more about what direction Apple is providing to developers regarding “How to develop for Apple Silicon systems”. And, that direction has been “no GPU’s other than the integrated GPU”. A lot of what Metal on Apple Silicon is and does (and what Apple likely wants it to be) depends heavily on the GPU sharing memory with the CPU, which would not be likely with anything other than an SoC solution. Until Apple announces a change as to how they handle GPU’s (say, at a future WWDC), the only GPU any Apple Silicon system will have is the one on the SoC.
