They should trick competitors to make their base storage 1/4 speed so Apple's 1/2 speed base storage will seem faster.
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
Data centers have already started a switch over to ARM chips:
https://aws.amazon.com/ec2/graviton/
https://cloud.google.com/compute/docs/instances/arm-on-compute
Why can't it be both? Also, why do people act like if a company doesn't immediately do everything possible to help reduce environmental waste, that the steps they have taken so far are somehow void/meaningless/BS?
I know. And that’s why it would never happen. But for gaming I prefer my Switch over my iDevices so my comment was entirely based on my preferences. One can dream…Yeah, that’s an iPhone or an iPad
Apple and Nintendo have always reminded me of each other. Rather tight knit and occasionally quirky companies that prefer to control the entire experience that they offer.
All Apple needs is a joycon-like attachment for the iPhone and a dock and one is good to go. I think there was even a patent for that?
Thats just not true, yes there is an advantage on being on the latest node, but Apples performance per watt is NOT the outcome of just the node.
The secret sauce is the secret sauce, it’s fantastic chip design. Stating that they can only do this because of TSMC is just not accurate.
That would just be a dumb idea… totally distracting from the focus of Apple products only… nuff said
If they were to do this, the open market 3rd-party AS devices would be different from M-series SoCs, probably with a modified GPU, subtly gimped cores (like smaller ROBs, different branch predictors, etc) and several specific-logic units missing. Their performance would never be equal to Apple product SoCs but would be much better than the nearest competitor, and would likely include a proprietary feature that would make it difficult to go back to a non-Apple ARM SoC design.
Inasmuch as Apple is not ever going to be straight selling M or A series SoCs to anyone else, a 3rd-party product would have to be a completely different tape-out and impression. I cannot see them going to that much trouble. The market will eventually take care of itself.
Inasmuch as Apple is not ever going to be straight selling M or A series SoCs to anyone else, a 3rd-party product would have to be a completely different tape-out and impression. I cannot see them going to that much trouble. The market will eventually take care of itself.
You will want to go back to Apple 90s. Make Apple thin and vulnerable.
Keep Apple M-series for mobile and desktop and laptop; this is their secret sauce.
But that would just mean Apple would be using MORE raw materials to make chips for all these people. No, the ONLY proper solution for an Apple that SAYS they care so much about the environment is to simply shut down. Stop making products, let go all their employees that are using vast sums of energy to work there, and turn off all the lights in all the buildings they’re overseeing. If Apple was truly concerned about saving the planet, they would simply not exist… not make MORE stuff!
It’s time for Apple to put their money where their environment is.
Apple’s definition of being green is from the perspective of them making more money (green).
Apple is all about vertical integration. They even start making their own modem chips. The days of Apple licensing their software/hardware to others was already dead the moment Jobs pulled the plug on Macintosh clones. It will never happen again.
Naah. Apple just wants to be a lifestyle company.
No need to give Pat a heart attack.
No need to give Pat a heart attack.
Agreed. Memory not so good anymore (kind of like Apples current base offering on Airs) but I think it was Scully(?) that did that. When the big guy came back he was not pleased with how some of the licensees used less than ideal parts and pieces and he was not happy with the loss of control over the user experience.
There is hugely flawed presumption here that most of the large players have not already made moves here. Apple adds about nothing because there are already multiple SoC vendors moving in this space already. Arm itself provise Neoverse baseline designs for the server space that are competitive. Apple has ZERO "first mover" play here in the server focused realm. They don't have a lots of the necessary pieces to compete in general server workloads.
" ...
Several major cloud providers now on the Arm bandwagon
With the latest introduction of Arm-based cloud instances, the British chip designer's ISA is now supported by six of the world's largest cloud service providers: Amazon Web Services, Microsoft Azure, Google Cloud, Alibaba Cloud, Tencent Cloud, and Oracle Cloud. Other cloud providers are getting behind Arm too, such as JD Cloud, UCloud, and Equinix Metal. ..."https://www.theregister.com/2022/07/14/arm_cloud_trend/
So which major Cloud services player is missing? None of the large ones.
Between Amazon Graviton 2 and 3 and upcoming 4 . Ampere Computing's Altra 80-120 core models and upcoming next gen late '22 - early '23 . Nvidia's upcoming Grace and Grace-Hopper SoCs. And a couple of other in-house projects. Apple isn't bring anything real for mid-large scale servers to the table that the major cloud services vendors can't buy off the shelf right now.
By the end of the year, Smaller businesses can go to HP and buy an "off the shelf" Ampere Altra unit.
HPE Is The First Big OEM To Adopt Ampere Computing Arm Chips
Hewlett Packard Enterprise has been an early and enthusiastic supporter of alternate processor architectures outside of the standard Xeon X86 CPUs that
www.nextplatform.com
[ So far it is has mainly been 'white box' , 'generic box' vendors that just offer hardware for "do it yourself" operators (like large scale cloud services that pragmatically mostly build and support their own configs. ]
Where to Buy - Cloud Service Providers, Systems Partners, Distributors
Ampere Computing® platforms are available for purchase through our ecosystem of partners including cloud service providers, systems partners, distributors, integrators
amperecomputing.com
e.g. Inspur , Supermicro , Gigabyte , etc. ]
In the narrow space where just want to run a macOS instances Apple has traction. But multi-tennat , dense hosted services? No.
Apple is missing major 'pieces' for the server market.
1. No ECC RAM. Don't need it to run some neighborhood mom-and-pop store website . However, Major corporate financials app where bit errors means money goes 'missing'. No ECC = Nope. Long running computations that need high accuracy? Nope; doesn't met project requirements.
Apple's RAM packages are also semi-custom and therefore not standardized , multiple source components.
2. An Soc optimized for less than 64 threads. Apple's super-duper , Extreme SoC is suppose to be a Wow factor with 40 cores. That doesn't make you a server player in 2022. Maybe back 2015 or so , but not now.
The design focus of Apple's mid-large SoC is to allocating most of the die area over to GPU cores and video specialized processing. That does a whole lot of nothing for CPU focused loads.
3. Apple's design philosophy for PCI-e v4 was to enable fewer lane provisioning out of the SoC. There dies max out at x4 PCI-e v4 lanes ( and some lanes buried inside of Thunderbolt controllers. ). In 2022 , the baseline for server processors is > 64 PCI-e v4 lanes; if not > 64 PCI-e v5 lanes with CXL 1.1 (or better) support. Apple has zero of that. Not even a member of CXL. Not even ability to put put a single x16 lane bundle.
Pretty good chance that Apple caps out at something like 32 PCI-e v4 lanes with there largest model SoC. (i.e., backsliding from what the Mac Pro 2019's W-3200 provides in terms of lane bundle count. )
4. Apple provides zero technical support for booting anything other than macOS. (macOS is capped at 64 threads so not surprising that #2 constraint exists). You can hack around with something other than macOS , but Apple is not going to provide tier 1 support for that. No type 1 Hypervisor support either.
No UEFI native boot support. ( so plug in cards can be initialized early in boot process ) . Apple has completely forked off the UEFI path. Can 'hack' back around that but that a 'hacked' boot process is not what folks who have 24/7/365 five 9's service level agreements generally look for in terms of implementation.
In contrast, Linux runs just fine on Ampere , Graviton , and other Neoverse implementations. Apple is way , way , way behind here. There was a 3-4 year period were Linux on Arm was "just around the corner" for being viable. It took several vendors doing lots of work over a time period longer than many expected to get things to the smooth state at now. ( lots of different low level management utilities and features that several server operators want/need to plug into the rest of their operations. )
5. No RAS ( reliability , availability , serviceability ) features either. Apple isn't even remotely trying to compete here.
6. Apple's SoC is highly skewed to single user ( high interactive , low single user latencies ) memory access deisgn. The P core complexed have capped bandwidth ( in part to not block the GPU memory throughput and latencies). The E cores don't provide the same service levels performance as the P cores.
So an M-series will do better on a single thread drag race workload, but not necessarily better running 20 different apps for 20 different people on the same SoC.
On major part of Apple's "secret sauce" on iPhone A-series and Mac M-series is that they dump features that don't fit their products. In the phone space there were the first to jump on Arm64. Not because it got them more memory access ( not being used on Phone products) or a bigger feature checklist status. They went quickly most so they could dump the legacy 32bit and (lower) stuff more quickly. They got to implement a cleaner instruction set and devote more design elements into better "new" code as opposed to holding onto legacy code the longest. It wasn't to ease the path to large memory servers... it was the path to faster small memory phone processors.
Apple spends more time "polishing" a smaller set of SoC features that matter more to just their products. That is a dual edged sword. It makes their SoCs better for Apple products, but also often makes them work for non-Apple products. Apple just swings the sword in the 'good outcome' direction.
Same thing in the M-series. The SoC are relatively anemic is talking to non Apple discrete components, because the primary target here is laptops and "iPad thin" iMacs. That is most of what Apple sells so they have SoC baselines designs that are highly skewed toward that. They are not trying to make everything for everybody. Qualcomm has more SoC designs per generation than Apple has products. Similar with Intel/AMD... more CPU processor SKU than Apple would ever use per generation.
What Apple has largely left out is server focused elements.
The Premise is fundamentally flawed so this "So"/"Therefore" goes down the drain . There is zero obvious leap here for Apple. Even if they were not strategically averse to "cloning" efforts.
Apple does not have an appropriate SoC product to fill this general space. So there is no "do the right thing" here for them to do. It is also not an largely unmet need for server companies. Apple would only be joining a party already supplying solution; not blocking the data center business from leveraging it.
If Ampere imploded and Arm largely abandoned their multigeneration Neoverse efforts that would be a problem. Arm is only going to be viable long term if they don't start collecting more $/units built ( higher average selling cost for Arm SoCs. ). So the chance of Arm falling out is zero. Ampere could shoot themselves in the foot, but for now they should have a very steady revenue source. Apple isn't competing with them. Amazon has zero incentive to sell their solution to others. And the rest are similarly constrained.
Apple can "help" green things up by rolling out suitable replacements for the many tens of thousands of Intel Minis and Mac Pros that are out there in places like MacStadium and other coloc/services vendors. Lowering the power consumption ( and/or raising the Performance/Watt ) of those will saves a sizable amount of kilowatt hours. That means a Mini Pro ( with a "Pro sized class" SoC) , appropriate Mac Studio updates, and likely a "rackable" Mac Pro with Apple silicon. IF Apple went to a "half sized" Mac Pro then the rack version could possible double the compute density of the 5U space it soaks up now. That would help in displacing the denser packed Mac Pro 2013 that are deployed. There is enough local and remote macs deployed as servers to upgrade over next 5 years that can make a incremental difference. Apple doesn't have to chase into enitrely new markets to do something positive "green" wise.
Last edited:
That is a bit of revisionist history.
Cloning came well after Scully. a bit of second "half" of Spindler and definitely with Amelio
Apple didn't transition to the PPC until 1994. They were not trying to do cloning start at the same time as the transition ( although preparations were ramping to try ). However, the PPC alliance was suppose to be about growing the base of PPC systems. Expanding along the macOS line was an obvious good vector that could possibly accomplish that. Apple though was mainly about ROM dongle (and I/O chipset limited ) approach to holding control of the Macs.
It was little to do about parts. Motorola wasn't building "junk" (and had leverage because made the CPUs). Neither was PowerComputing ( who was moving faster to get a G3 out). What Apple wanted to do is send the cloners off into sub markets that Apple didn't want to do themselves. Cloners were suppose to go and expand the market share of MacOS in the broader general PC market. What more so happened is that went for the more juicy spots that Apple wanted for themselves. They mainly sold back into the same submarket with Apple. Even more boxes with slots didn't do much.
There was a standard presumption in the "clone strategy" approach in that if macOS would just expand to 20+% of the market there would be space for Apple and others to sit in. However, if take the view that the classic PC war was over then macOS share wasn't going to go past 10% (and perhaps lower) and there is little upside in smaller shares of a smaller pie. It is not a viable enough niche for multiple players at comfortable margins. [ Apple's somewhat imploding balance sheet was ample evidence that smaller share of smaller pie didn't work. ]
That Apple "had to" buy NeXT to enable a more to a long term viable Operating system was another problem with the presumption of a quick rocket launch into 20+% market share. (and was another expensive load on the weakened balance sheet) Apple didn't have the operating system pieces either.
Apple's general PC market shrank during the clone era phase (not increased at all ) .
".. .
Apple sold 4.5 million Macs in 1995, a level it wouldn’t reach again for a decade. With the clone program in place and competition from Windows 95, that dropped to 4.0 million in 1996 and 2.8 million in 1997. That was the year the word beleaguered became widely attached to the Apple name. ...
"
After cloning was over , Apple killed off more than a few Apple Mac products also before the balance sheet turned around. That whole "bigger than Dell or HP " market share or bust illusion mainly ended in 'bust'. ( or to be more Microsoft than Microsoft). Apple was neither of those kinds of companies.
Last edited:
I was not insinuating that at all. You did not understand my post. I was saying that another company can reverse Apple Silicon and make iterative changes to comply with applicable IP laws.
Yeah, I addressed that with an apology for my misunderstanding.
Just 18 months ago, people were saying “there‘s no way a phone chip is going to compete with Intel….a phone chip can’t handle PC workloads!”
And here we are….most mainstream websites and video channels comparing a phone chip against i9’s and RTX30xx series - hilarious.
Tech changes; remind of the PPC era and early iPhone days. Even Chromebook kills some low-end and education laptop market.
The performance advantage is largely due to the tight, low level integration between the hardware and the software. Unless Apple also starts making firmware and drivers that offer the same level of compatibility to non-MacOS operating systems and/or hypervisors then Apple Silicone is probably not going to be different to commodity ARM64 CPU's.
The issue with this take is that you seem to be assuming Apple Silicon is off-the-shelf ARM64 design. It's not --- they license the ISA, but the microarch is completely Apple-designed. Apple owning the whole stack is of course an advantage, but their microarch is objectively better than competing designs.
I'm not assuming that at all, if I thought it was off the shelf then why would I be suggesting Apple needs to write stuff for it?
You are making an assumption though, that Apple's microarch is better than anyone else's. It's not open source hardware so unless you happen to work in a very specific part of Apple there's no way you could know this. It almost certainly is at least for the time being, but good luck verifying that.