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 7,1 What does Apple's switch to ARM based processors mean for the future of the Mac Pro
- Thread starter Korican100
- Start date
- Sort by reaction score
Point is they showed the desktop versions running on the Macs natively. Just because they have iOS versions means nothing as the apps are completely different.
[automerge]1592852756[/automerge]
In my opinion they will use an SoC for the Mac Pro but in the same way that AMD do on EPYC. No separate chipset. Storage, Audio, USB and so on all within the SoC. Graphics will be handled by dedicated cards. This is all how it's done in EPYC.
The big question is will they use chiplets like AMD has on Epyc, Threadripper and Ryzen or will they use a monolithic die like Intel? - I suspect for the Mac Pro they will go with a chiplet approach.
Hey look they showed an iPad CPU, running 3D rendering software built for _x86_ performing well and the CPU didn't burst into flames.
Guess ARM is ready for real work after all.
Guess ARM is ready for real work after all.
Given the 2-year timeframe and the complexities involved, we should not be surprised that the Mac Pro will likely be the last to be replaced. If you bought it as a work tool, by then it will be nearly three years old, and not only should it be fully depreciated (at least in the US), you also should have gotten your money's worth of work out of it. I never keep most computers past three years, and I didn't expect to keep my Mac Pro's longer than that--things advance too much and for work I generally need the highest power I can get.
If Apple can match or exceed the performance of the Intel chips and my primary apps will work on ARM, then I'm all good with it, and that's a reasonable timeframe as well. Certainly not the super-fast switch some here were claiming Apple would have to pull off. It's not necessary like it was in the PPC -> Intel days and therefore why push it.
Apple's really been on this push for a few years anyway and first iPadOS and then Catalyst were just steps on the path. Now they feel confident enough in their silicon development and roadmap to make the full leap. Interesting times ahead!
If Apple can match or exceed the performance of the Intel chips and my primary apps will work on ARM, then I'm all good with it, and that's a reasonable timeframe as well. Certainly not the super-fast switch some here were claiming Apple would have to pull off. It's not necessary like it was in the PPC -> Intel days and therefore why push it.
Apple's really been on this push for a few years anyway and first iPadOS and then Catalyst were just steps on the path. Now they feel confident enough in their silicon development and roadmap to make the full leap. Interesting times ahead!
Perhaps more importantly, Aperture being around kept Adobe from making Lightroom subscription only for years.
I hadn’t thought about it until their presentation, but they currently have the same core processor architecture scaled from a watch up to an iPad Pro. I wouldn’t doubt that the engineering effort is already done. It’s not like they just decided to do this yesterday and have nothing to show for their work. They just showed a new version of macOS running on Apple Silicon with apps already compiled for Apple Silicon. From the demo’s, I wouldn’t have known it was Apple Silicon unless they said.
I thought the Calm app was an excellent example. You can take the app you already own, install it on your Mac and have the same experience there as you’d have on your iOS device. It’s only audio, so it doesn’t really matter what it looks like.
There's gonna be software that's not likely to get dedicated Mac versions, even if it's easier. Overcast is one example, and it'll serve fine as a unmodified app on a Mac, better than a web player.
I wish Aperture was still around. I have a feeling it would smoke Lightroom on Intel Mac's today and on the new silicon, even better. Maybe in a dream world Apple will revive it? Not likely, alas.
It looks like OpenGL/OpenCL has survived onto ARM Macs. So apps that are still OpenGL/OpenCL should be able to make the jump without being held up by a Metal port.
I'm sure for some things it will be fine, but my point was that the majority of iPadOS/iOS apps aren't really desktop replacements for any desktop-type apps. They will be great as little applets like the widgets, but I've personally never really used many of those on my desktop Macs anyway. I realize they have appeal for some though.
Definitely, but I think "widgets+" is how they should probably be thought of, and certainly how a lot will be used.
I guess the question now is if we get another iteration of the Mac Pro on Intel. We got the quad-processor PowerMacs after the announcement so a substantial refresh isn't out of the question, although I think Apple surprised itself by their speed.
New -SP chips are coming from Intel, but the 2100 -Ws are pretty new, and I have no idea about the replacements for the -Ws in the Mac Pro.
I suspect at the end of this year or early next year you might see ones with a speed bump; I seem to recall that there was an expected second round of the current chips from Intel, but that ultimately the socket for this version of the Xeon-W is a dead-end. So if I am wrong about a speed bump on the existing chip, then my prediction is no upgrade or update at all on the Intel Mac Pro, and the next Mac Pro will be an Apple processor version. They say two years, but it took longer than that for this Mac Pro, so grain of salt, announced in two years, shipping in 2.5 years.
That's because it was running on the A12Z the same processor found in the iPad Pro's. Of course they didn't just start now. But with all the work they have to do leaving the system that requires the processor which is the most different to their other systems last is just obvious.
They don't use the same SoC on everything and the cores in the Apple Watch are not the same as the cores in the iPhone processors. They have less cache, smaller cores and so forth. Similar to how Intels ATOM processors are not comparable to their i series and XEON processors. The architectures are similar but not like-for-like.
[automerge]1592857021[/automerge]
I don't think we'll see that. With work really focusing on the GPU and accelerator cards more than the CPU. Just my opinion though.
I wouldn't really expect to see it. Part of the performance improvement on the A series is that the GPU and CPU share a single address space which means they can access the same information directly. No copying into VRAM. Multiple CPUs would probably break that arrangement and create the potential of information having to be copied around.
I still expect the A series Macs will support traditional discrete GPUs though. And those GPUs will still have to copy data into VRAM. But having a backup GPU on a Mac Pro with direct access to CPU memory space will be helpful for compute tasks that like to stream data.
Totally agree. The Afterburner card is a perfect example of the things you can do with a different approach to the processor vs. specialized add-on silicon. To me that's what's most exciting about this going forward, specialized silicon deployed to speed up specific tasks vs. the monolithic x86 approach where it's mostly been about cranking clock speeds and throwing more cores at problems. We started the move away from that when GPU's got so powerful software started being written to use them to accelerate compute tasks; this move by Apple has the potential to open that up much further.
I respect that. At this point I have a couple of mission critical iOS apps which will be nice to run on my Mac, but I admit those are the exception.
Post-keynote, we know that Adobe and MS are already working on the transition. A big factor will be how well Rosetta 2 works and how long it will stick around. I agree, it was/is the small software (sometimes critical) that really hurts/or adds up. The move to Catalina was a pain for me for that reason. However, software that made it to Catalina is most likely to make to just to ARM (fairly) smoothly. Although there are one or two programs I use that I'm curious to see if/when they make the leap because they depend heavily on legacy/cross-platform libraries.
I'm still optimistic. It sounds like you've already made your decision, and I can't fault you for it, sounds like it made a lot of sense for you. But I do think that's biasing your perspective.
If we can assume most elements of the Mac Pro will be the same, I can see another GPU refresh in that time frame and maybe slight CPU price drops. Hopefully the 580 doesn't remain the stock GPU, and presumably MXM modules can migrate to the new machine.
The point I was making is that if they’re talking about being done in 24 months, the engineering work on the CPU is likely already done, they’re now well into designing the system that will carry that solution. I just wouldn’t say “most engineering effort on top of what they’ve done” because that effort’s likely already done.
Mmm I don't agree. Having been involved with chip designing and knowing the lead times and things it's more probable they're not even in the sample stage from TSMC yet for a Mac Pro chip if it's coming Dec 2021. Of course I don't know how Apple actually works but we usually got back final silicon without errata only 6 months before a product launch. We would get basic samples 12-8 months before product launch with only 2 to 4 months of rework for errata.
These were ASIC's with only a few billion transistors. Apple's up to like 7 Billion? NVIDIA is up-to like 21 Billion.
Hey folks, just had a beautiful Pro delivered (my first) for demanding photo work. I can’t help but feel like I am second guessing my decision now with all the ARM developments given I was hoping to continue to upgrade this machine and make it last at least 5 years and hopefully 8. What do you guys think?
What were those ASIC’s meant to go into? My assumption is based on the fact that “what goes in” for Apple is solidified well in advance (the iPhone 2 years out already has it’s CPU designed) so they can focus on the integrating all the components without having those components going through additional iterations. Anything going into a Pro has to be tested with a variety of configurations that wouldn’t exist in a mobile system. For them to get working CPU’s six months from shipping their Apple Silicon Mac Pro is cutting it WAY too close. Granted, they won’t have to make very many of them...
[automerge]1592862168[/automerge]
Still, though, 3 4k streams on a architecture as powerful and power efficient as a 12z is impressive.
We don't really know what Apple's 2-year timeframe is though. Two years from today is a reasonable assumption, but they didn't say that they would have every model replaced in two years. They also said they would be shipping Intel Macs for some time going forward. My guess is that two years from now they will have every model with a roadmap, but whether they will have shipped every model is another question. I can see everything *except* the Mac Pro being somewhat easily shipped in the two-year time frame because let's face it, most of what they sell are the low-power, lower-performance mobile-type Intel chips. They've nailed that chip type for years now in the iPhone and iPads, and scaling that is probably a very minor job relatively speaking. The full-sized desktops (iMacs) are a small fraction of their market, and the few desktop machines with the full-fledged full-power Intel chips are only a fraction of that. The Mac Pro? A tiny fraction.
My guess is that similar to the original MP the next generation will be identical in the mechanicals with a new motherboard. I can't see tossing the MPX design so soon, so they will probably have MPX modules in their ARM Pro.
Were I taking bets I'd say you'll be able to buy an ARM Mac Pro no earlier than December 2022, and quite possibly more like June-December 2023 given their track record in the "Pro" space. From Apple's perspective that may be perfectly fine anyway, as the software will have more time to be mature and the Pro users will want mature solid software.
[automerge]1592862821[/automerge]
I think five years will be easy. Whether you get 8 years will depend on how long Apple supports updates to the x86 version of MacOS. If you're using it for work, then anything past three years should be gravy. If not, raise your fees a tiny bit now and by the time you can get a ARM Mac Pro you'll have saved enough to buy one.
That's the lead time for NVIDIA too by the way. They'll get final silicon only a few months before release and they go into professional machines and servers.
The chips I worked on ranged from PCIe switch chips to network SoC's for network switches and firewalls (think 200Gb/s-1Tb/s+ internal switching capable networking devices), PCIe based device emulator chips (which consisted of over 50 chips combined in a single working system to emulate other PCIe based ASIC's which are in the prototyping stage).
Pretty much all the time before we reached final product it's only about 6 months that we've had final chips and in some projects we would have less than that if we found some error in the chip and needed a re-spin though due to chip fabrication lead times about 3 months out is the absolute best you can get, it literally takes that long to make one chip with all the steps involved in the manufacturing process.
Sample silicon though where you can start designing the product you'll have that 18 months out for sure. Very rarely though did we finish a design of the whole system/product more than a few months out, sometimes tweaks to the systems within the last month like changing the gauge of a wire or the source of a plastic washer as example of the more minor things.
But again I don't know how Apple does things. Clearly it took them ages to do this current Mac Pro. You may remember they brought in a bunch of journalists and well-known Mac enthusiasts to their design lab about two years before the new Mac Pro was shown off and told them they were planning a new Mac Pro. And in that case they weren't even making the CPU it uses.