back to article Qualcomm takes a swipe at Apple's build-not-buy culture (because it wants to sell stuff to Apple)

Qualcomm has distanced itself from Apple, and took a veiled shot at its rival as it tried to clear the air on its relationship with Google. The company's swipe at Cupertino came in the form of an assertion by CEO Cristiano Amon, during a speech at the Snapdragon Tech Summit in Hawaii, that the horizontal model of the chip …

  1. big_D Silver badge

    Cooperation makes for more innovation.

    That is why the Qualcomm chips are so much faster than Apple's and why their desktop chips run rings around the Apple M1 chips in laptops.

    Oh, wait...

    I'll be interested to see what the Snapdragon 8 Gen 1 brings to the mix, but even Google are looking at their own customized chips. Mediathek are also moving into the AI subsidised chipsets, nVidia ditto. I will be interested to see if Qualcomm actually have an answer, with Nuvia, they might have a chance in the desktop and server markets, moving forward, but they have a lot of catching up to do first.

    1. Charlie Clark Silver badge

      The M1 benchmarks flatter somewhat due to the use of shared memory. The M1 does, however, definitely benefit from the close coupling, which should make compiler optimisations for the hardware easier.

      In the phone world, it's clear that there's not much between the high end chips, and Google possibly does have the edge for ML work, having decided to go for more grunt and lower precision. Apple's move is really only about margins.

      1. Anonymous Coward
        Anonymous Coward

        I think Apple's chips are a lot faster...

        1. Charlie Clark Silver badge

          Do you now? At what are they faster?

  2. Anonymous Coward
    Anonymous Coward

    Aren't Qualcomm responsible for the terrible slow Android Wear chips which wrecked the ecosystem? TBH it's a bit embarrassing how much better Apple's mobile chips are

    1. Rahbut

      AIUI they simply rebadged older models and resold (basically) the same silicon.

      Having lost Kirin, it's good to see Apple doing their thing, Google shipping Tensor and MediaTeks Dimensity - we need more competition/innovation, something Qualcomm has stifled in the past.

      1. rcxb1

        <blockquote>we need more competition/innovation, something Qualcomm has stifled in the past.</blockquote>

        It's something Apple has stifled in the past, too.

        Wasn't long ago Apple forced Intel out of cellular modem biz, bought the division cheap, and resumed exactly what Intel was doing.

    2. Dave 126 Silver badge

      Qualcomm also had a deal with Microsoft to be exclusive suppliers of chips for Windows on ARM. This deal was secret until recently. It likely explains why MS have made no noises about ARM Windows on Apple M1 computers.

      1. Stuart Castle Silver badge

        The funny thing about that is that a lot of people assumed it was Apple. I've read that Apple have taken the position that they are happy to support Windows on the M1, they just need Microsoft to make it Windows for ARM available to end users.

  3. A Non e-mouse Silver badge

    Apple aren't a daft company: They know where to invest their money. If they thought Qualcom (or Intel) offered them a better solution rather than designing their own, Apple would take it. The fact that Apple have taken on the massive overhead of designing their own silicon and still come out with a design that it as good as, or better, than Qualcom/Intel speaks volumes.

    If Qualcom want Apple's business they should offer a solution that Apple couldn't resist. Actions speak louder than words. Qualcom are just throwing a tantrum as they no longer have Apple by the *ahem*...

    1. DS999 Silver badge

      Qualcomm's problem is that cellular modems are becoming a commodity. The quality of the modem matters less with each generation, so you see stupid stuff like Qualcomm claiming its X65 generation of modem is better than X60 because it supports 10 Gbps 5G instead of 7.5 Gbps 5G!

      Obviously none of us will ever see numbers close to that, just like we never did with the theoretical speeds of 3G and LTE before it, but we're at the point now where no cares about getting faster potential LTE or 5G speeds at this point. Any problems we have with cellular performance has to do with the carrier, and where they have sited their towers versus where we want to be when using cellular.

      I'm sure when Apple introduces phones using their own modems Qualcomm will try to paint that as bad for customers, since they won't be as good as Qualcomm's (oh no, maybe they'll only be capable of 5 Gbps!) Their problem is, no one will care. They also have formerly low/midrange only competitors like Mediatek competing with them toe to toe on the high end, Samsung looking to expand their use of Exynos beyond some of their own phones to third parties, etc.

      They know they need to come up with a compelling reason for smartphone OEMs to pay a premium for "Qualcomm inside", or they risk the 35% of their revenue that's from patent licensing becoming a majority, and eventually sliding into irrelevance as far as consumers are concerned as an IP only company.

      1. martinusher Silver badge

        I thought that the one of the reasons for the attack on Huawei was that they had scooped up about two thirds of the patents needed to implement 5G and that they had by far the best implementations of their technology.

      2. rcxb1

        <blockquote>The quality of the modem matters less with each generation, so you see stupid stuff like Qualcomm claiming its X65 generation of modem is better than X60 because it supports 10 Gbps 5G instead of 7.5 Gbps 5G!</blockquote>

        The quality of a cellular modem isn't in the theoretical maximum speed, but in how well it manages with a weak cellular signal and/or interference from lots of other phones in the area.

        Many people claimed the Qualcomm modem was better than the Intel modem in this respect in the last generation of iPhones. That was before Apple forced Intel's cellular modem division out of business and bought it up as their own.

        Will Apple improve on Intel's cellular modem design before they release the next version, or will Samsung be able to advertise that their phones get better reception than iPhones (thanks to their use of Qualcomm's modems)?

        1. DS999 Silver badge

          It is still improvement on the margins that don't matter to most people. I never noticed any difference in performance of my iPhone going from one with a Qualcomm modem to one with an Intel modem and eventually back to one with a Qualcomm modem.

          The differences may be there if you test with lab equipment, or you are unfortunate enough that your home or office is located on the edge of reception (and in those cases you use wifi assist so your cellular reception quality doesn't matter anyway)

          Samsung only uses Qualcomm modems in the US and a few other countries, in most of the world including the entire UK/EU region they've been using their Exynos SoC which contains their own modem for years, and their plan is to phase out their use of Qualcomm modems entirely and use Exynos SoCs worldwide. I think the only reason they were was their modem didn't implement Qualcomm's proprietary CDMA flavor of 3G which is primarily used in the US - but is being discontinued by its major carrier, Verizon, next year.

          It didn't affect Apple's sales at all because any difference was mostly theoretical and not something people could see - and when Qualcomm was slagging on Intel modems they were talking about them being able to do "only" 600 Mbps while Qualcomm modems could do 1 Gbps. They weren't talking about picking up weak signals because that's impossible to quantify. They like to advertise with concrete numbers which is why when they were talking about the X65 modem they were talking "10 Gbps" and if there was any improvement in its ability to pick up a weaker signal they didn't bother to mention it.

          1. rcxb1

            <blockquote>are unfortunate enough that your home or office is located on the edge of reception</blockquote>

            No, there are large swaths of the US where cellular reception is weak. Some very heavily populated areas, too. It affects people who live there, people driving through, people who travel to job sites, and more. It matters very, very much, and you can't always get good Wi-Fi calling where you have weak cellular coverage... . Verizon is able to charge premium prices and remain the top carrier precisely because they have the best coverage. Sprint is out of business because they had the poorest coverage.

            <blockquote>I never noticed</blockquote>

            Well then...

        2. A Non e-mouse Silver badge
          Joke

          The quality of a cellular modem isn't in the theoretical maximum speed, but in how well it manages with a weak cellular signal and/or interference from lots of other phones in the area.

          Apple have a whole lot of experience in antenna design ;)

    2. rcxb1

      <blockquote> If they thought Qualcom (or Intel) offered them a better solution rather than designing their own, Apple would take it.</blockquote>

      Funny, because Apple forced Intel's cellular modem division out of business, then bought it up, cheap.

      <blockquote>Apple have taken on the massive overhead of designing their own silicon and still come out with a design that it as good as, or better, than Qualcom/Intel speaks volumes.</blockquote>

      Actually, I'd say Apple used their cash and market size to get first call on TSMC's FAB, which is mostly responsible for their advantage. Would the M1 be any better than processors from Qualcomm and AMD if not for being a full node shrink ahead (and having the money to fab as many large cores as large as they wish)?

      1. DS999 Silver badge

        How did Apple "force" Intel's cellular division out of business? By making a deal with a Qualcomm? If your company has only one major customer, and they leave you for a competitor, does that equate to them "forcing you out of business" in your mind?

        And yes, M1 would still be better than Qualcomm's SoCs even without the node advantage - because the older Apple SoCs made on the same node were still significantly faster.

        1. rcxb1

          <blockquote>If your company has only one major customer, and they leave you for a competitor, does that equate to them "forcing you out of business"</blockquote>

          When both your "major customer" and also your biggest competitor are abusive monopolists, and they partner together with the intention of cutting you entirely out of the market, then yes.

          1. Chris--S

            Intel being schooled in abusive monopolist behaviour … how far it’s fortunes have fallen. No leadership in processors, no leadership in modems, no leadership in nodes AND no leadership in abusive monopolist behaviour. Put the Engineers back in charge!

    3. Charlie Clark Silver badge
      Stop

      Apple cares about margins and control. With the volumes it sells, it can easily offset development costs (chip design is much cheaper than building fabs, and this is only customisation of the ARM designs) and it can then enforce its walled garden in silicon. It's great for shareholders so it must be good, right?

      1. DS999 Silver badge

        No Apple is not "customizing" ARM designs. They are designing the ARM CPU cores from scratch entirely on their own, the only thing they get from ARM is the ISA - i.e. "opcode xxxx uses the following registers as inputs and outputs producing this defined result". Apple and other architectural licensees work with ARM on defining future extensions to the architecture, so they have some say on what gets added.

        If ARM went out of business tomorrow Apple wouldn't be affected other than no one continuing to iterate the architecture, but they can add opcodes to their own designs (and have, i.e. stuff like Apple's AMX extension) so they could carry on unbothered by ARM's absence.

        Companies who license cores from ARM, not so much.

    4. Stuart Castle Silver badge

      I think you are right. Apple wouldn't have gone to the effort of buying their own chip design business (PA Semi) unless they thought they would somehow profit. I've no doubt that if Qualcomm came up with something that suited their needs better, they'd go with Qualcomm like a shot.

      I think Apple buying their own semiconductor design company does give them a massive advantage. People have noted how well Apple Silicon performs (and it does, while I don't game on it, I have an Apple M1 Mac mini, and it is fast, as is the M1 iPad Pro I have).

      I don't think it's because Apple's silicon is inherently better than Intel's or Qualcomm's, but they have the advantage that because the teams designing the hardware, software and chipsets for their devices all work under the same roof. That makes it easy for each team to communicate their needs to the other teams, and to work together to ensure that they can adapt what they are doing to factor in the needs of the other teams.

  4. elsergiovolador Silver badge

    Chips on demand

    What we need is a foundry that could produce chips on demand and on a small scale e.g. 1-100 chips per order.

    We also need to teach at unis how to make design and make chips.

    There is a ton of people with innovative ideas, but they cannot afford to switch into the industry due to poor pay and you can't really do this part time.

    Only way is to become a slave of one of those big corporations, spend years as an employee and never achieve anything meaningful for yourself and your own family.

    We need an environment where anyone with a drive and idea could create their own corporation making chips.

    1. l8gravely

      Re: Chips on demand

      It's called FPGAs, and it's where alot of companies do initial proof of concepts. I've supported a group making ASIC chips for customers for many years now and it's not a solo project. You have layout, simulation, DFT (Design for Test) and packaging/thermal specialists who all work together.

      And then you have to take the logical design and tweak it to work in the actual process you're doing to use, which tells you how your transistors can be laid out and designed. Which influences your power, timing, heat and other factors of the design. It's all about tradeoffs.

      So if you want Uni students to learn chip design, then go buy them some FPGA boards and let them loose. They will learn the low level VHDL part of the design, and then work their way up the stack. But doing the foundry work means you're going to have to also get hold of the Cadence or Synopsys software, and those licenses are expensive, hugely so. Mind bogglingly expensive since it takes teams of engineers to write and validate this software.

      Oh yeah, I forgot about the team who does the design validation, making sure the customer's high level design actually works and passes the requirements. Fun fun fun!

      1. A Non e-mouse Silver badge

        Re: Chips on demand

        FPGAs are hugely powerful - completely different beasts to the PALs, etc many people may have started with. I've seen many startups begin with an FPGA version of the product until they get the money to upscale to silicon. (And many products stick with FPGAs through their life as they're easier to update with bug fixes through a software update)

      2. elsergiovolador Silver badge

        Re: Chips on demand

        Yes, so FPGA serves as a gatekeeping tool between commoners and people "holding the chips" if you will.

        Sure you can learn VHDL and whatnot, but that's like building a car from a LEGO versus actually building a car.

        Common people don't have access to these tools and some software that is essential for chip design is kept secret, so majority of people won't even be able to see it.

        So you get your idea running, provided you actually find a suitable FPGA platform, then you have to rely on big corporations and VCs and give up majority of your IP, if these almighty people find your design viable. Of course the costs are artificially high because of de facto monopoly.

        These technologies need to be democratised.

        1. Anonymous Coward
          Anonymous Coward

          Re: Chips on demand

          Oh for heaven's sake. What a load of unmitigated rubbish. The only gatekeeper for access to silicon is money. Not secrecy. And if you can't afford to buy whole wafers there are shuttle services available.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2022