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.
Concerns About an Apple Silicon Mac Pro
- Thread starter RedTheReader
- Start date
- Sort by reaction score
Some people (I have no idea where they are getting this from) are talking an upfront cost of $500million+ to design SOCs on TSMC 5nm.
No citation as to the numbers as far as I can see so not sure what “SOC” design is even being costed out but this is not the first time that I’ve heard that SOC design costs are increasing exponentially in addition to fabrication costs on newer nodes.
That makes no sense. What could that even be based on? Not engineering costs, for sure. And the fab charges you per-wafer start. You hand them a gdsII - it’s not like TSMC charges you to do engineering work. Looking at that graph, the biggest chunk is “software.” WTF? I know what cadence and synopsys licenses cost per seat, and those numbers are absurd. And what does “prototype” even mean?
I dunno seems dodgy to me too. The only thing consistent with it is that I’ve heard design costs are much higher on these lower nodes and are slated to keep going up even beyond fabrication but that was just a general statement and while I got the impression that people were grumbling about costs to develop getting higher, I didn’t get the impression that it was drug money laundering amounts. So yeah … no citations are a bit of a problem here.
Edit: The one thing that bothers me about a new die with 12 cores total is why not release it for the MacBook Pro? It could certainly handle it. And for a new dual die with 12 cores, even I think that’s cutting it down by a huge amount. I suppose they could be scraped together from dies that really didn’t pass but still …
Last edited:
On the iMac Pro front, I wonder what combination of CPU and GPU cores make would make it faster than the current iMac Pro? I would have thought that Apple would want the Apple Silicon version to be significantly faster than the current iMac Pro - from memory the single core already is, largely a matter of increasing the cores to beat the multicore score and enough gpu cores.
Yup, apart from part failure most studios just keep the config and then swap it out once it's outdated or give it to a department that doesn't need so much power (across the departments there is a fairly high variability in the amount of power required to work effectively). Old machine can always be retired to pipeline tasks (background transcoding etc). I feel a lot of the demand for replaceable parts comes more from freelancers and prosumers who are going to rely on machine as their sole machine. It's a valid use case, but not as universal as some people tend to make out.
As you said the single threaded score of the new M-series chips is much higher. The current M1 Max GPU should be faster in most scenarios than the Radeon Pro Vega 56 in iMac Pro without having to add more cores. In peak multicore, the Max CPU is almost as fast as the 18 core iMac Pro. So theoretically a 10+2 arrangement would be faster. Having said that, the iMac Pro was great for its time … 4 years ago. I’m hoping that they aim much higher for the top configuration of the new iMac. We’ll see.
From a marketing perspective - For the smaller Mac Pro, I'd have firstly ticked the expandability box.
And then, return on investment. Both of these success factors based on the buyer (who Apple are marketing to).
Currently a 14" Macbook Pro starts at $1,999, with not many cores, 16 MB RAM, & 512GB drive. That includes some fancy cooling, a very expensive display, a costly keyboard, a costly touchpad & a costly battery.
So there is no reason for Apple not having a low cost, entry level Mac Pro. Even with a faster processor. And I'd have designed it via the modular model. Just have a couple of the latest PCIe ports on the side, and offer a PCIe dock option. With its own power supply. Mechanical hard drives? Easy - use a doc, via the PCIe port.
If I was a unified memory Apple designer, no add in GPUs would work. But other stuff would.
Apple would have sold heaps IMO. Just make the docks stackable. Choose your Apple CPU, GPU & RAM configuration when you buy your new Mac Pro. That way Apple would get all the market, as they'd start with a low price, and they'd be offering expandability.
Instead, it seems we'll get a mini tower with hopefully a PCIe or two expansion availability. Who knows about mechanical drive expandability in such a machine. I suspect CPU, GPU & RAM will be an initial purchase decision, and hopefully, as broad a choice of those combinations as in the MacBook Pros.
There's lots of Classic Mac Pros types out there would be happy with a sharply price low configuration in M power with a couple of PCIe slots and disk choice availability. I'll be disappointed if Apple doesn't offer a low priced entry model like they have with the 14" Macbook Pro.
And then, return on investment. Both of these success factors based on the buyer (who Apple are marketing to).
Currently a 14" Macbook Pro starts at $1,999, with not many cores, 16 MB RAM, & 512GB drive. That includes some fancy cooling, a very expensive display, a costly keyboard, a costly touchpad & a costly battery.
So there is no reason for Apple not having a low cost, entry level Mac Pro. Even with a faster processor. And I'd have designed it via the modular model. Just have a couple of the latest PCIe ports on the side, and offer a PCIe dock option. With its own power supply. Mechanical hard drives? Easy - use a doc, via the PCIe port.
If I was a unified memory Apple designer, no add in GPUs would work. But other stuff would.
Apple would have sold heaps IMO. Just make the docks stackable. Choose your Apple CPU, GPU & RAM configuration when you buy your new Mac Pro. That way Apple would get all the market, as they'd start with a low price, and they'd be offering expandability.
Instead, it seems we'll get a mini tower with hopefully a PCIe or two expansion availability. Who knows about mechanical drive expandability in such a machine. I suspect CPU, GPU & RAM will be an initial purchase decision, and hopefully, as broad a choice of those combinations as in the MacBook Pros.
There's lots of Classic Mac Pros types out there would be happy with a sharply price low configuration in M power with a couple of PCIe slots and disk choice availability. I'll be disappointed if Apple doesn't offer a low priced entry model like they have with the 14" Macbook Pro.
Last edited:
Considering that the M1 Max is about 430 mm2 and a xeon, high end GPU chips are 600+ mm2, adding GPU and CPU cores seems possible. Expensive though.
So can we expect dual or quad of this new 12cpu M1 Max for the Mac pro? Or the standard M1 Max 10 cpu core?!
If the thermals in the bigger imac can allowed that the M1 Max 12 cpu could go even higher in clock speed that will be not just improved based on the nr of cores but also for the core speed increased
If the thermals in the bigger imac can allowed that the M1 Max 12 cpu could go even higher in clock speed that will be not just improved based on the nr of cores but also for the core speed increased
Apple can definitely afford to offer an "entry level" compact M1 Max Mac Pro for around $1999. Not sure that it would favourably compare to alternatives though. There is just not enough CPU/GPU power to rival desktop workstations. It would make sense as an ultracompact Mac Mini Pro though.
The iMac Pro has not been too bad a purchase (10 core, a Vega 64). Hardware wise it's showing it's age now, particularly in rendering (using a 32 core box at the studio and you can really see the difference). 10 performance cores is a good number; for a lot of simulation tasks that has been the sweet spot core wise. 10 cores that don't throttle with fast single core speed would probably be great for a lot of use cases (often responsiveness and fast feedback are more important than multicore scores).
Given the single core speed the new iMac shouldn't have that much trouble beating it. Going by geek bench scores, the current M1 max chip seems a bit light for a pro desktop and but the dual chip configuration would put it in a reasonable place. Might be a little on the pricy side though...
I think it was around 2TFLOPS FP32 per AMX core? At any rate far enough from Nvidia Tensor cores. The question of course is whether Apple would be willing to invest in the silicon area needed.
Didn’t look into it in detail. But from what I can see the basic design of all these matrix extensions (ARM, Intel, Apple) seems to be fairly similar.
All this means to me is that, as I read the rumors about Apple’s future desktop plans, those that vary significantly from Apple’s current lineup (which already takes into account a shrinking desktop market) feel like the longer shots.
The issue for Apple is whether Mac Pro desktops users switched to Windows, whether they'd also switch to Windows notebooks. And most would IMO. The Mac Pro facilitates Apple product cross selling. So Apple still needs to maintain a critical mass in the relatively shrinking desktop space. And revenues that surround that desktop space can be quite significant compared to a typical notebook. Even compared to a pricy Apple notebook....
All this means to me is that, as I read the rumors about Apple’s future desktop plans, those that vary significantly from Apple’s current lineup (which already takes into account a shrinking desktop market) feel like the longer shots.
And depending on cloud services is not a universal belief, despite the rental model's increasing popularity.
If we look at 2020, and see that Apple sold 20 million Macs. We know from Apple that 80% of those are laptops. Of the remaining 4 million the Mac Pro is likely a low single digit percentage of the 20 million, so let’s say 4% which would be 800,000.
Of the remaining 3.2 million, Apple’s indicated that after laptops, iMac is number 1 is iMacs iMac Pro is number two. Even if I’m generous and make all three roughly equal, that’s 1.06 Mac Mini’s plus 800,000 Mac Pros… 1,860,000 and that’s ONLY if I give the Mac Mini a number that’s most definitely well larger than it’s actual sales.
If Apple’s total revenue for 2020 was 28.4 billion, 1.9 million systems (again with an artificially boosted number for Mac mini sales) wouldn’t be that significant in the big Mac picture.
Well everyone can rest easy here. Nvidia has backed out of the deal.
Yeah under the circumstances. So many different regulators were against it. They weren’t going to fight them all.
Something I was unaware of until reading that story was that there was a March 2022 deadline. Softbank made a nice chunk of change with this non-deal!
EDIT: Ouch, just looked at how much they paid for ARM! Enter into 39 more deals that are bound to fail and they’ll make their money back! LOL
Yeah I’m not sure if that article is accurate - I think that was a soft deadline that could in principle be pushed back because other articles I read (even before the US sued to stop the deal) mentioned it might be if the regulators even just didn’t finish their reviews in time which was likely. That was when the deal was supposed to close though pending regulator approval which obviously they didn’t get.
What I was saying was that the Mac Pro facilitates cross selling of other Apple products and services - notebooks, phones, iPads, cloud, software and software commission revenues, etc.
I was also looking at whether Apple needs to have a good value Mac Pro (ie expandable tower Mac).
I'm guessing, but I suspect Apple makes more profit from its ear pods than it does from its Mac Pros. So I suspect the Mac Pro segment in Apple is a very weak profit sector, if it's judged just by itself.
By restricting the access to an affordable Mac Pro, those Mac users who want to control their data themselves and expand their desktop's capability relatively cheaply - who see a desktop computer facilitating that - will have to switch to desktop computers operating Windows. Of invest in costly and bulky cable add on hardware and buy a NAS.
I'm suggesting that those who switch from an Apple desktop tower computer to a Windows desktop computer will also result in cross selling switching - in notebooks, pads, phones, software, etc. So the loss of revenue to Apple is much more than just loosing a sale of a desktop tower computer.
However - the facts are now that if classic mac users (ie affordable Intel Mac Pro towers) did not buy a 2019 Mac Pro, then the "I like Apple because its easy to use" segment have now already left Apple's desktop towers, because the classic Mac Pro is no longer supported. For example: to be able to run today's software versions on Classic Mac machines now requires complexity (Opencore etc), so that the "Macs are easy to use" demographic has already left if they did not buy the costly 7,1 Mac Pro.
And I suspect many already use Windows as well as their desktop Macs - something likely forced onto them by the sustained lack of suitable Apple tower hardware.
So there is a business case that Apple doesn't see a need for a good value Mac Pro desktop - ie one with PCIe expansion and disk expansion built into a single form factor computer (a mini tower for instance).
We'll know soon enough. It wouldn't cost Apple much to try a value based M desktop computer though, and I think they'd gain a huge amount from doing so. But it would take some bravery I guess because the core of Apple's profit left the desktop years ago. The expandable Mac tower perhaps seems to most at Apple Corp as a historical monument.
Last edited:
That is likely correct. BUT, by the numbers, they could lose every desktop non-iMac user and that would probably be less than 1 million systems a year. Likely much less.
I, too, would actually be surprised if, after the last Mac Pro, the folks that were waiting for an affordable Mac Pro haven’t already moved on meaning that Apple wouldn’t even lose those million sales a year.
That Mac Pro units sales number is highly likely off by at least 4x. (even back in Mac Pro heyday it likely didn't crack 200K run rate) Pretty good chance that it is off by 8x . Mac Pro is probably in the >= 100K/units/yr range. It is no where near 1M. Not even ballpark distance. If look at workstation market estimation as to who the "players" are Apple is in the "other" category. It is relatively small.
Several years ago Apple said that Mac Pro was 'single digits'. I suspect they were being generous and rounding up to get to 1%. Several folks keep hand waving that 'single digits' must mean some number close to 5 since that is the 'middle of the range of single digital'. That isn't particularly rational.
Apple putting the Mac Pro last on the 6 year update plan and last in the transition is very indicative that the percent is quite low. That's why it has low priority for resources and effort. It makes for a nice "hobby product" but it isn't a driver of product line priorities.
Once haven't given away several 100K of units to the Mac Pro that it doesn't have that will leave a lot more for the Mini even if don't split them even with the iMac. When numbers were last out iMac was substantively over 50% of desktop.
Apple's total revenue isn't relevant. The aggregate Mac market makes enough money for itself to come up with its own priorities and resource allocations in terms of market segments that ware worth while to follow. It isn't big enough to 'fork off' and have its own completely independent CPU+GPU SoC though.
Similar for the Mac Pro. It is no where near big enough to significantly fork away from the rest of the Mac line up. Lots of the SoC will be stuff primarily made for the rest of the Mac (and iPad Pro) line up repurposed and augmented to fit the Mac Pro. That is just driven by the economics and economies of scale..
P.S. long , long ago the Mini numbers used to be pretty bad ( like lower than Mac Pro run rate). That shifted as mobile CPUs got better. The iGPU got out of the relatively really bad zone and iMacs shifted over to desktop processor ( which shrank the Mac Pro run rate substantively). When Apple only had one "dekstop" system that actually used a desktop processor that artificially floated the Mac Pro run rate up. ( Apple herded some folks to higher prices , but those folks left when there were more affordable options to upgrade to or buy into Mac market with. )
There is always a deadline. Even more so when using 'stock' for pay for a significant amount of the whole deal. That price of the stock being used to purchase isn't constant over a very long period of time. Take too long and may have to adjust the terms. There is/was likely a clause that they could get to a mutually agreed extension if it just took a little more time for the deal to be worked out or details settled.
Therin lies the core problem. Softbank isn't looking to sell Arm at a rational price. They primarily want to sell Arm to cover up several bonehead investments that Softbank made ( including significantly overpaying for Arm). They are looking more for a "bail out" than a business sale. Dumping a large pile of surcharged (inflated ) Nvidia stock on them would do that. It is somewhat doubtful that doing a paying an "even bigger fool" price for Arm would have helped rationalize Arm's basic business any better over the long term. And should have been a 'red flag' as to Nvidia having some kind of alternative motives for buying than they said.
Arm's primary business is being a contract R&D firm. It isn't a super duper high growth stock kind of company. To be successful long term that primarily need to plow money back into the business to stay out in front. Not dump gobs of money on stock buy backs and/or dividend disbursements.
While Nvidia's stock was shooting up very high it was largely just "Monopoly game" play money to pay most of the Arm costs with dilutive stock. Throwing Nvidia stock at Softbank which really wouldn't pass it along to Arm really won't fund much bleeding edge silicon design work.
Nvidia's breakup fee is probably enough money Softbank can funnel into Arm to keep it "growing" so maybe they can attract another consortium buyer or crank a hype train for an IPO (so Softbank can get most of its money back).
Agreed. I was swagging on the higher side with my numbers knowing that, even provided truly unrealistic numbers, those exaggerations would still not add up to a notable amount of revenue (as far as Apple is concerned, there are likely many companies that would like to have even that sliver of business
