Power draw matters for desktop computers as well. It's important for every computer ever made, from entry level notebooks, to multi million dollar super computers.
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.
Qualcomm Expecting to Supply Just 20% of 2023 iPhone Modems As Apple Prepares to Launch Its Own Chips
- Thread starter MacRumors
- Start date
- Sort by reaction score
The only OEM most serious tech businesses consider in the PC/Linux world consider as an alternative to Apple is Lenovo, which admittedly sells great laptops when on sale. Although, Apple is way better as client machines as long as you don't need CUDA.
Can't wait for Apple to add this revolutionary feature and charge an extra whopping $199 for it...and it'll probably be eSim.
Apple's GPUs are nowhere near NVIDIA's best. They are significantly better than what Apple was using before, which is what is important, but they are currently quite behind. And in terms of power efficiency, Apple is two nodes ahead. TSMC 5 vs Samsung 8. That's a huge difference. Apple is using many more transistors than NVIDIA's best to compete with their mid-range- NVIDIA is not the one that looks bad here.
Last edited:
Sure, but that's not what was said. It doesn't matter on a total production level. On the desktop, people will go to great lengths to squeeze every drop of performance from their hardware at any power expense. They will go so far as to soak the entire rig in mineral oil and attach two PSU's. Power is relevant, but only because it's a limitation not a concern. So yeah, on a performance per watt metric, Apple is killing it. On a total performance metric, they are way, way behind.
Last edited:
Some people will do this because they want the most raw performance from their hardware. That is fine and their choice. However it does limit future innovation as you can only push more and more of the same thing before other physical and scientific limitations get in the way. Eventually someone has to think differently to come up with a new way of doing things so even more performance can be found. Not initially of cause, it will take the new way of doing things a while to match and then exceed the old way but every new idea has to start somewhere.
That is a very subjective statement.
Based on pure performance, NVIDIA wins and looks good. But based of any performance per power ratio NVIVIA looks rather poor in comparison to Apple. It all boils down to what you want from your hardware and what compromises you want. Neither side is perfect and both have significant pros and cons.
*****
Also as an aside Apple have not released a desktop class desktop machine with desktop class Apple Silicon yet. Traditionally the only Mac to have this is the Mac Pro. iMacs traditionally are mostly (better) notebook hardware in a desktop form factor. Potentially with the improvements Apple Silicon has brought to the table, Apple might be able to put desktop class versions of Apple Silicon into the iMac. That is a huge maybe though. We shall see when the prosumer and pro level Apple Silicon iMacs are released.
People are saying hardware in a notebook has less raw performance than hardware in a high end desktop. Well duh! Thanks captain obvious. Comparing apples to apples, is there any non Apple notebooks out there with vastly better performance than M1 Pro / M1 Max? If so what are their power draw values and thermals like?
Finally raw power is nice, but what really matters is the performance in the tasks you do every day. The Apple SoC is more than just a GPU so again that is not really an Apples to Apples comparison.
Yep, you both are correct. Also: bigger battery in iPads. So they can take that risk of elevated power draw at first before honing it down for iPhone, then again shortly afterwards, for the Watch.
Did they ever do anything with that, other than make a SIM tray? (I honestly don't know)I think you missed that headline when Apple bought the patents and exclusive rights for Liquid Metal
It is indeed an odd statement. There could be some power play behind it, similar to Broadcom WiFi chip business.
The new iPhone doesn't represent 80% of their sales. Which means there is no way a new iPhone launched with Apple Modem Qualcomm would only left with 20% business in a single year. For their Fiscal 2023 to have less iPhone Modem, it would mean 2022 iPhone will start using their own Modem. If Apple only start using their own modem in 2023 September / October, in Qualcomm's view that is 2024 already. Quote from Reuters.
But then if Apple were to use their own modem in 2022 iPhone, Qualcomm would have already know by now. Since Final Engineering validation ended in July. And supply for 2022 iPhone Modem has already been secured. And they do have an obligation to disclose it to their investors an update their fiscal 2022 forecast. Both of which they haven't done and they went on to clarify there is no concrete data suggest Apple will stop using their modem in 2023.
My theory is that I trust what Cristiano Amon is saying and trying to put the message forward.
Qualcomm is basically updating all their future forecast without the inclusion of Apple. i.e All forecast are provided as worst case scenario.
And yes, I still doubt Apple will launch their own modem by 2023. And Oh I remember 90% of MR said Apple will launch their modem in 2020, 2021, and 2022. Where are they now?
But they have released a desktop-class chip, it's just that they decided to release it exclusively on laptops. Until they release a desktop computer their laptop will be compared to desktops because one could always just plug it in and use it as a desktop machine.
It's not because it's in a notebook. It's got less raw performance because it's not a dGPU. The fact that it's a laptop means little because it's being judged against desktops. Until they provide alternative hardware to be compared with that what it's up against.
This doesn't make sense. The tasks I do every day are what the performance measures. If it unzips files 1-second faster that's a LOT of time over the life of the device. Real-world the 3090 renders a video in 2 minutes instead of 10 minutes that is a huge deal and can play a game at 8K at 90 fps instead of 4K at 50 fps. Those are major daily impacts.
One can always take it on the go and use it untethered to a power point as well. Something a pure desktop computer can not do. Just because Apple have not released any professional level Apple silicon desktops yet does not mean it should be compared with something it's not. (Though people are doing this and it's makes for some very entertaining reading) The fact it is being compared with something many classes higher than it is while still maintaining a low power draw is quite the achievement. How often are people rushing to compare desktops to notebooks in this way? Not that often, but Apple Silicon is a totally different SoC.
They? What? There exists a whole ecosystem of non Apple notebook computers for the Apple Silicon MacBooks/Macbook Pros to be compared with. People just choose to compare the Apple Silicon notebooks to desktops because the non Apple notebooks just can not compete on any level at all.
It makes a lot of sense. Your comment here just proved it. If you are willing to accept the extra power draw and heat and lack of portability to achieve the gains you mentioned above, that's great. All the better for you. Pure raw performance has its place. But thinking differently also has its place. In the end it is all about what each person needs and each person understanding the limitations of what machine they have chosen to get their tasks done.
I’m not the one who made the initial comparison. I responded to someone who claimed apples chip was comparable to Nvidia. It’s not. That’s all I’m saying.
F qualcom with a big F. I am an adroid user, i have an s20+ and very happy about it but qualcom are arseholes. they purposefully stop supporting chipsets after under 3 years to to make you buy a new device. expensive and powerfull devices rendered scrap by their greed.
20% of the unit volume of phones or 20% of the revenue? Part of the reason why Apple "hates" Qualcomm so much is that the licensing fees scale with the price of the resulting overall device. A $300 phone has one price for a Qualcomm modem+RF chipset and a. $1,200 phone has another. The Qualcomm modem in an entry level iPad versus top of the line iPad Pro has same issue. [ reportedly this is to help shift the cost of R&D from the "low margin" phones to the high margin phones. Apple isn't a fan of the progressive tax model. ]
In 2023, would the unit volume drop that low? Probably not. Would the "using Qualcomm" units have very largely shifted to below the $599 price point ? Probably yes. That would be a substantive hit to Qualcomm's revenues and that is the primary message that is being here ( and that they need to find more even set of revenue sources. )
Even if Apple didn't wrangle some "no scale with price" agreement, then it doesn't hurt to give a conservative estimate on business impact. State that it will be 20% and if 30% it just means more money. Qualcomm won't get sued for making more money than they projected. ( the opposite isn't true. Say expect 30% and make 25% and someone will threaten to sue if thought you knew it would be 25%. There is about zero incentive to try to be "aggressive" on the business retention estimate here. )
Apple would have to estimate what the non 2022 iPhone flow would be also. Apple doesn't do "fixed" modem buys. They have "just in time" contracts with some guarantee minimal amount and a "up to" amount that they just put a placeholder on ( with no guarantee to buy ). If Qualcomm is doing conservative estimates of the unit business then prudently they should just use the "minimal guarantee" amount and not the "up to" amount as a long term estimate. That is the only thing they would have a firm commitment to. Little good reason in a finance project to weave in a large portion of the "up to" number when know that Apple might flip-flop on some (or most) of that if it turned out that the 2022 iPhone did better than expected. Qualcomm really shouldn't be guessing how much market share shifting the new phone will do. If it turns out fewer folks don't like the upgrades and more 2021 phones get bought , then it would be a shift up in buy. If the upgrade is some quantum leap in better, then this years projections for "last year's" and "even one year older" is going to be substantively off. Apple largely doesn't care because they have "flex buy" contracts so the risk is largely dumped off onto their suppliers if Apple gets the estimates wrong.
When Qualcomm has 2-3 generations of iPhones covered, then the shift from old/new doesn't mean as much in overall revenue roll up . (and even less on overall units sold ).
Pretty good chance also that Apple is trying to inject a wider. "minimal" and "up to" variability gap number on Qualcomm contracts also. If their modem works really will with fewer hiccups than expected they could redo a "last year" or "even one year older" phone with their own modem. For example a. iPhone SE upgrade that is usually a 'paid for' hand-me-down design. Flip to a less expensive modem on a lower cost phone could help goose the margins higher for Apple. But won't necessarily tell that to Qualcomm ahead of time, but could build in a lower minimum if was going to try to ramp a replacement after the initial 2023 iPhone models' demand bubble subsided and freed up more modems . Not like Apple hasn't sold same basic iPhone modem with two different modems in them before.
Pretty good chance Apple will have some backup Apple product to "soak up" the extra Apple modems if iPhone 2023 projections are off, but also don't want to take the full risk of attaching that product to the new modems.
2023 is about right if they are shooting for a "better than Intel/Infineon" but "not quite as much coverage as Qualcomm" modem. I don't expect them to 'beat' or even match Qualcomm in some edge cases, but it will probably be something that is "good enough". 2023 is plenty of time to get the basics of the 5G modem they had decoupled from Intel fab design rules and into something else and working against 2021 era 5G base stations.
Apple says that their rule of thumb is that they will do something in-house primarily if they can do a much better job than can be done by external suppliers. I don't think that is the primary driver here. Lowering the component costs is probably an equal (or even higher priority ) driver here. [even more so with the looming threat of decreased services revenue with apps/subscriptions from outside paystores. ] If Apple has something that is just "good enough" (not necessarily better in every way) they'll just do it. Even more so if it is still a discrete chip with a common interface that the alternative also uses.
That's partially Google's fault on Android design also. Samsung phones on Exynos don't get substantially lower support from Android either and have an entire stack, SoC -> Android -> Phone , there with Samsung. It is the combinatorics problem of matching Android variants to hardware. Samsung is doing something like 30 phones designs in the next year or so. Apple isn't going to do 30 phone designs in 5-10 years. The super broad diversity of Android is a dual edged sword.
There is better support for decoupling the SoC specific layer from the generics of Android now. But even Pixel 6 isn't all that much longer. The full breath is security upgrades not OS upgrades. ( OS upgrades to 2024 ; so 3. and security to 2026 which is more so the five which is closer to what Apple does )
Not sure why would expecting results like that. Apple software dealing with Apple hardware is one thing. To be a highly effective modem Apple's modem has to work with a wide variety of cell phone towers/base stations that run software Apple does not make and run hardware that Apple didn't make either.
Second, Analog-to-digital doesn't necessarily benefit much by using a process node that is newer than your competitors. Bleeding edge digital stuff does better on bleeding edge fab process. Analog doesn't necessarily match up that way. The first level RF management chips aren't likely to be deeply integrated into the Apple SoC .... so not much 'magic' there by Apple SoC. The modem itself has decent change of not making the main SoC also. If still a discrete modem then have inter-chip communication wattage issues to deal with. ( where Qualcomm makes different area and connectivity trade offs to do more monolithic designs. For example, Qualcomm goes with smaller GPU and smaller caches to get the digital modem stuff on board. )
Unified Memory here isn't going to be a huge win. The overwhelming vast majority of the data processed by the modem is only needed by the modem. Apple's oversized caches and GPU performance tweaks buy a whole lot of nothing when it comes to the grunt work the modem has to do.
Third, Qualcomm really isn't all that horrible when it comes to performance and/or power consumption. Folks complain about the price, but not about the results. Their modems costs more, but generally work better ( over a broader spectrum of operating contexts ( service providers , software , and base stations ) ) than the competing stuff. The modem maker that was stumbling on performance was the one that Apple bought. They needed to get substantively better just to get close to where was Qualcomm was. Apple was primarily using them because they were cheaper and "good enough". Not that they were better.
Fourth, Apple can not do "faster than 5G" just on one side of the connection. 5G is a standard they don't solely own and for their modem to work at all it has to talk to something else. In the performance/watt ratio, they primarily can only lower the "watt" component of that equation; not the performance. That kind of graph where score more "tech porn" points (vertical axis ) than the other guys isn't the primary metric of "success" here.
Doesn't matter unit or revenue, there is no way it is down to 20% within "one year". And this is specific to Modem, not patents. Apple will pay Qualcomm Patents regardless of using their modem or not. So changing modem doesn't impact their patent license agreement. But yes they are simply downplaying the effect of a no apple as customer future.
Except that is not how it works. That is not how Just in time works, nor how up to or guarantees works. The flexility of contract only works when you have a basket of choice in supplier, like DRAM or NAND. Not to chips with single vendor, and especially for leading edge node fab capacity.
Last edited:
You are presuming that Apple doesn't "backport" their modem to a subset of the current products. It won't be that hard. The iPhone 13 for instance. ( iPhone 13 Pro teardown pics below step 8 ).
iPhone 13 Pro Teardown
The iPhone 13 might look like an incremental update from the outside, but we’ve got a feeling there’s more lurking beneath the surface. We took a...
www.ifixit.com
There is a 'sandwich' of two boards there. One is the radio stuff and the other is the "computer" (SoC , storage ) parts. The 'tops' of the two boards decoupled.
The Orange and yellow outlined chips are the Qualcomm ones. If Apple simply makes another board with a Apple modem and RF module that use the same (or less) power, dimensions, and the same SoC to Modem interface standards, then they could make a Qualcomm-less version of that daugtherboard. Couple that to the other board and ta-da .... lower Qualcomm sales. This isn't some giant moon-landing complexity project. Especially, if they give themselves a couple of years to do it.
The logic board assembly in the current ( designed in the last year or so after Apple bought the modem division ) have modular "radio" daughterboards. Modular means replaceable.
Apple could straightforwardly have an Apple modem chipset in hundreds/thousands of iPhone 13's in Early 22 as a test mule harness to validate the chip in the field. ( doesn't necessarily need to wait for 2022 or 2023 designs to finish before starting. )
' Back of the envelope numbers' just to put it into a more grounded context. If Qualcomm is charging $130 for their two chips and Apple can make their "clone replacements" at $100 that is a net saving of $30. Say legacy phone make the switch for sells 4M. So that is $120M. Even if the R&D for the board and recertification cost Apple $25M (which likely no where near) that would mean Apple would net $95M with the switch. Your hypothesis is that Apple would pass up double digit millions in net revenue to give Qualcomm more money. I wouldn't be on that. Neither is Qualcomm.
Even if the gap between the two is only $8 , that is still $16M which is likely close enough to breakeven.
This primary issue is that if Apple plans ahead to replace Qualcomm, then it won't be that hard. If planning ahead it also wouldn't be hard to design an Apple modem that can be 'dialed back' to 2021 feature set even if trying to do something that will compete fully with Qualcomm's solutions in 2023. They could make a 'mode' that matches the iPhone 13 so there is no significant specificaiton change. Just another SKU. ( which if they are at a transition to a "gets cheaper because another year older" phase change ... not that hard to introduce a new SKU).
the patents don't come with the "% of device" costs. Qualcomm's margins are the bigger issue.
Apple doesn't really doesn't much care about the Qualcomm patent revenue payments they'll still have to make , Apple will just make the end customers pay. Slapping a 20+% markup on on higher BOM costs just makes them more money anyway.
No. Just in time has a major component in inventory control. That is a different dimension that trying to pit multiple suppliers against each other. For example.
"... Samsung is the sole supplier of the 5.4-inch panel used in the iPhone 12 mini, the smallest model of Apple's 5G-supporting iPhone 12 series. ... "
Apple 'Likely' to Reimburse Samsung for Missing OLED Display Purchase Targets Due to Poor iPhone 12 Mini Sales
Apple is likely to reimburse Samsung Display over a shortfall in OLED panel orders owing to lower-than-expected demand for iPhone 12 mini models,...
www.macrumors.com
"... Apple has agreements with many of its suppliers to make a minimum purchase, likely to achieve lower prices per unit. If Apple fails to hit that minimum purchase, contracts usually stipulate that a penalty will apply. This year, Apple’s supply chain and demand were both affected by COVID-19, which could explain what led the company to order fewer OLED panels than initially expected. ... "
Report: Apple faces nearly $1 billion penalty from Samsung for ordering too few OLED screens - 9to5Mac
Apple has reportedly made another payment to Samsung for selling too few iPhones. According to a new report from Display...
9to5mac.com
Companies that do not have Apple's "800 lbs gorilla" status may also use just-in-time to play off competiting suppliers against each other. Apple is just far more direct in doing risk shifting. Pointing that the opportunity to make lots of money if the volume is high ( " we're Apple , hordes of fanboys stand in the rain for our products... we will sell alot" ) , but deeply Scrooge McDuck at the same time. Discounts and more than willing to stick the supplier with hefty problems if things go wrong.
Apple withheld payments and sued Qualcomm. Pretty doubtful they are out to end the supplier relationship without a departing "FU" on their way out the door. Apple will likely pull out the full "800 lbs Gorilla" when doing to the "ending" contract terms.
Every Monday morning Apple looks at sales of all the products. If there is an identifiable trend of substantive mismatch between what they have ordered and what they are selling.... Apple bumps the brakes on buying. That is how their contracts are set up. Apple spends tons of daily/weekly effort to keep from buying "too much". The whole infrastructure, including their supplier contracts, is set up that way.
Last edited:
Yes because that would require a whole new validation process? And new FCC process. New Antenna testing. it isn't just the modem. It is the whole stack from RF Front End to Filtering.
Qualcomm doesn't charge $130 for Modem. Not even $100.
Either part of the new agreement and settlement with Qualcomm. Or the norm would be percentage per devices cost, just the same as old times and industry wide. Not sure why you are getting the idea it is not percentage just because they are not using the modem.
I know how it works. I used to work in CE supply chain.
Basically you are reading too much Internet stuff, I even commented in one of your quoted Samsung OLED article. They are all horse ****.
I believe Asus or Acer also had sim slots on some of their offerings but you're right, Lenovo is the king when it comes to PC laptops.
I've had a few Lenovo Thinkpads over the years while working in large scale ITAD for a while. We used to work with big corporations down to little mom-n-pop operations. The only Apples we ever got came from select govt agencies, education, and the occasional micro-business.
Apple cannot compete with the OEM PC market in business particularly since they don't even bother trying. I saw a web ad for "Apple for Business" and got a good chuckle out of that.
They can always go back making Eudora. I loved that email software! Am I the only one that remembers that?