pastrychef
macrumors 601
Do you have a link and quote to this BS that you speak of?
I don't recall them ever saying they're not in business to make money.
Last edited:
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.
Apple should start selling their chips to 3rd parties
- Thread starter Zest28
- Start date
- Sort by reaction score
Do you have a link and quote to this BS that you speak of?
I don't recall them ever saying they're not in business to make money.
The huge problem with that analogy is that Apple had a Personal Computer product and Personal Computer operating system. Apple was in that market, but just using Intel only a the CPU package.
Apple killed off their specific server hardware long ago. The 2019 Mac Pro rackmount is just an inexpensive 90 rotation of the PC tower box they sell. It is not rack hostile ( handles and feet sitting way out when in the horizontal position), but it is not server density optimized at all. For what it is ( number of internal drives and CPU core count) it is relatively huge.
Apple has been choking "macOS Server" to death for years on the software side.
If if Apple has zero interest in this space for their own products, how are they going to come out with an amazingly competitive Server SoC for other people????? That is the huge gap here.
Apple was interested in dumping Intel. Apple was Intel only buyer and Intel was stumbling badly. Why wouldn't they be interested in getting off? In 2016-2018 AMD wouldn't have looked like an option. ( maybe they are stopping shooting themselves in the foot or maybe not).
The plain M1 ( and others with suffixes) are not a phone chip. If put an M1 in a phone it would fail. Likewise if put an M1 in a PC tower ... it would also be a fail ( 3-4 x1 PCI-e v4 lanes isn't going to cut it).
The laptop i9's and RTX 30xx are the main comparisons.
Macs sold by Apple are 70+ % laptops and that's mainly what done with the M1-M2 series; build laptop competitive SoCs.
Where someone has been using Intel Minis to do small-business/home/department-level server duties then the M1/M2 Mini (and hopefully M2 Pro) will do just fine. However, hat is really mutating the connotation of "server" this thread started off with.
Apple has said several times where they think they can do a much better job than the current component makers then they'll consider replacing it with something of their own. In the "big iron" server SoC space where is the huge fail going on if put all of the options on the table? Apple's data center is chock full of Linux servers that run just fine on top of AMD or Ampere Compute SoCs also if Apple wanted to go that way. The Mac SoC are missing gobs of features necessary (so not current vendors are not messing up there either). So how is Apple doing a much better job than the alternatives?
Xcode Cloud is not the more inexpensively priced possible cloud service either. At the price point Apple is using it can be run just fine using off the shelf Macs as the system components. The bulk of the rest of their set up is not macOS at all so what does a macOS SoC buy there? Not much.
Apple has largely saddled the currently transitioned Mac desktops with laptop SoCs : mini - M1 , iMac - M1 , Studio (iMac 27" replacement) - Max. The Ultra takes the highly constrained but far cheaper to make path of just being two Max dies glue together. So the dual edged sword on that is that is assumes much of the laptop design constraints also. The Mini should get the M2 Pro as an alternative.
So Apple hasn't really done anything even PC desktop/workstation specific yet. The Mac Pro SoC is missing, but if it just incrementally moves the constraint needle to just mainstream "desktop" features , then there still won't be a true medium-large scale server offering.
If Apple took the MP 2019 logic board and made changes to drop in a AMD Threadripper WX 5000 and previous out the I/O from that it would be an across the board uplift in CPU performance and I/O . Not sure where the component vendors are 'failing' them there. ( Yes , pragmatically stuck with only Intel since transitioning away from x86-64 ). The WX5000 outstrips macOS thread limit. The PCI-e I/O is better match to current and next years GPUs . RAM capacity limits are cheaper to get to. etc. etc.
That is probably not where the next Mac Pro is going. It is probably going to be something with constraints that Apple throws at the Mac Pro their SoC will be the "better fit" SoC. More Mac specific constraints (as opposed to server oriented ones). Some of the features of the Intel Xeon W-3200 and WX 5000 are because of silicon reuse that Intel/AMD weaved into the product for maximized server die reuse . Some features tricked into the MP 2019 was highly coupled to the W-3200 and because Intel thought to include them. Apple leverage the features in the marketing of the product that used that CPU package, but were not necessary inherently huge fans of all the features. ( Intel laptop and desktop CPU provision 8-20 PCI-e lanes and Apple does x4 (and a niche kind of x8 they don't fully use). )
Apple has done backsliding on RAM capacity , I/O, number of displays driven , etc. on the current set of M-series offerings. With respect to those characteristics the M1 series is much closer to being a "phone chip". There is more to generally good system design than Geekbench and synthetic offscreen GPU scores. Even more so in the server space.
Yebubbleman
macrumors 603
Apple designs their chips for their software and they design their software for their chips. It's a decent part of why their performance is so great.
Sydde
macrumors 68030
Although, AIUI, other stuff, like linusOS, runs better on Mx than other ARM SoCs. This might be in part due to IceStorm efficiency cores, which are apparently much better than even the 7xx mid-peformance cores others use in their typical 4+3+1 layouts.
tomO2013
macrumors member
Nail on the head!
This is why I struggle to understand why folks come onto this forum and quote latest desktop scalder-lake, bunson-burner lake processors, cherry picked fish fingers chess benchmarks and non-optimized 3d workloads in blender (relative to x86). It demonstrates (to me anyway) a lack of understanding of Apples business model and very future focussed view of silicon / software interplay in terms of heterogenous computing and experiences.
Apple is focussed on heterogeneous ‘experiences’ - I say experience rather than just silicon design as Intel/AMD are trying to go in the same heterogeneous direction but they don’t have full control of the OS, the development toolchain, software products using the equivalent of the accelerators and co-processors that Apple bakes into their silicon to support. This forces a very general purpose approach to their forward looking heterogenous silicon designs where the balance of how much silicon real estate to donate to a particular accelerator is more challenging given the obvious backdrop of ‘who will use this?’ ‘ how quickly can they optimize their solution around the latest accelerator/co-processor’, ’where is the requirement for X new accelerator coming from?’.
Apple was way ahead at the races even predating the M1, in the iPhone era. That same approach allowed Apple to use smaller batteries but claim similar battery life to Galaxy phones with twice the miliamp hour capacity. It allowed an M1 macbook air to playback Canon Raw footage from a Canon R3 or R5 with smooth playback. It allowed thin and light designs in iMac and macbook air etc..
In short Apple sells an experience and that heterogenous silicon design - in my humble opinion only - needs to be viewed through the lens of a heterogeneous experience to the end user. Laser targeted silicon accelerating laser focussed features in software that Apple pushes out and controls through their ‘soup to nuts’ toolchain.
In effect, this is a different business model entirely to that of Intel, AMD, Samsung, Qualcomm etc..
Parking the existing supply chain issues aside for a moment (where Apple can bare feed the demand for it’s own products), it represents such a fundamental shift in approach to all parts of the tool chain that if Apple WERE to sell it’s silicon to third parties, it would no longer be Apple. It would loose its secret sauce.
Coochie Boogs
macrumors regular
Thankfully, this will never happen. The 3rd parties would just have junky software that won't match the efficiency of Apple's software.
1BadManVan
macrumors 68040
I completely agree. Apples software integration with their hardware is just as important if not more important then the hardware itself. That's why And and intel can have insane specs on paper but still fall short in a ton of areas to apples hardware, that on paper, should be quite a bit slower.
Same reason android has to cram massive batteries in their phones to try and compete with iPhone battery life and still fall short on performance, even with twice the amount of ram, higher clocked cores, more cores etc. Gotta give apple credit in this space, they are definitely far ahead of the competition with performance and efficiency
Sydde
macrumors 68030
Well, looks like I was not quite correct, as M-series processors seem to have forced the fixing of a kernel design flaw that only manifests in "ridiculously out-of-order" processors. Everything was just fine and then this upstart machine comes along and forces us to make our code robust.
ericwn
macrumors G5
So any possible action for the planet has to be taken by Apple like providing one of their easiest to use differentiator asset to competing companies? Interesting take.