Who will win the ARM race?

36

I think we would agree that ARM is the future of PC laptops, workstations, and servers.

Who will be the eventual non-Apple ARM winner? Qualcomm, AMD, Intel, Nvidia, Microsoft (make their own chips ala Apple?), or someone else?

Comments (38)

38 responses to “Who will win the ARM race?”

  1. shameer_mulji

    It's too late for MS to make their own ARM SoC's. They would have to spend billions just to acquire talent and to top it off, it isn't like there's an abundance of A-class chip designers to go around. And assuming, they do acquire the talent, it'll be 3 to 5 years, at least, before their first SoC gets out the door. As far as Intel goes, I'm not aware that they are designing any ARM SoC's. Right now, money is on Qualcomm and Nvidia. AMD is the wild card.

  2. scj123

    It seems to me that everybody was sort of happy with Intel / AMD and as soon as Apple releases a version of Mac running on ARM (Apple Silicon) this is the future of everything.


    People seem to be forgetting most people can pick up a laptop that does everything they want and has (for them) great battery life for around £500. If I was to say to my parents they could buy a new laptop with a better battery to them this would mean they would have to charge it once every 2 weeks instead of once a week.


    Yes tech enthusiasts will always want faster and longer battery but the majority of people are just happy when they work

  3. longhorn

    ARM might be a bad proposition for Windows developers. Are they going to make one binary for Qualcomm, another for Samsung and a third for AMD?


    Maybe UWP is supposed to be the new format for cross SoC compatibility, but it hasn't exactly taken off. Windows wasn't designed to run everything in a virtual machine like Android.


    ARM might be one nail in the coffin for Windows, but there is still AMD64 and knowing what kind of company Apple is I wouldn't be surprised if they pull the handbrake if they reach more than 15 % global desktop/laptop market share. If everyone uses Apple then the products are clearly too inexpensive and important margins are wasted.


    If you can become the world's most valuable company by targeting 10 - 20 % of the population and evade monopolization charges because you are so small, why change that concept?


    One of the biggest threats to Apple as a (united) company might be increased market share. Large profits are OK as long as you are a niche player.


    • bkkcanuck

      In reply to longhorn:

      When you don't have to give up the margin on CPU components to Intel, and you don't have to give up that same cut to Qualcomm for modem chips (which are a percentage of the device price)... you can keep the same margin on a little lower priced device which in turn can make it more competitive in the overall market. Apple usually aims for a specific range, and I don't see that changing but I also see more affordable options in the future like with the iPad or iPhone (as they have done). A higher percentage of the market for devices, increases services revenue while not increasing as much the cost of providing those services... So they won't subsidize hardware, but they also are not going to purposely reduce their potential marketshare and their services revenue from that.

    • wright_is

      In reply to longhorn:
      Are they going to make one binary for Qualcomm, another for Samsung and a third for AMD?

      That is how Apple does it, although just 2 binaries. The two binaries are moulded into one executable file and the relevant part is executed, so it is transparent for the user. But the developer still needs to compile and test both versions.

      • bkkcanuck

        In reply to wright_is:

        The ARM instruction set itself (for an architectural license) must contain the full instruction set. GPU functionality built in to the SoC would be a different driver (as it is for Intel GPUs). For further things like neural engine functionality etc. this would likely be additional drivers. It really is no different than supporting different chipsets from Marvel etc. that are built into the motherboard ... but now some of this is right on a single package. As far as running an app, you would distribute the app for Windows (1 binary) and use the OS provided API to access specific functionality within the chips. Sorry, but I don't see what the big deal is. What is needed is someone to write the specific drivers for the additional functionality.

  4. anoldamigauser

    I do not necessarily agree that ARM is the future of PCs and Servers, though the power and thermals are sort of compelling from the perspective of a data center.

    Most normal people (and if you are on this site, you are not normal, at least from a technology perspective) do not give a thought to the processor in their PC. Does the PC run the applications they want? Yes, then great. No, it is an expensive doorstop. All of this stuff is literally magic to them. They do not want to understand, they just want it work for what they need.

    Apple's great achievement here is that normal people won't really notice a difference between an Intel Mac and an ARM Mac. So if a Mac works for you, then great, you are all set. If a Mac does not work for you though, well then, it still will not work for you. There is a reason that Mac has only about 10% market share, and it has nothing to do with the wow factor of the hardware.

    • bkkcanuck

      In reply to AnOldAmigaUser:

      Laptops make up 2/3rds of the market - if they provide a compelling product that ran ARM (big if the way things are now with Qualcomm chips). As laptops go the rest of the industry for general computing would follow (assuming quality high end chips were available). Everything rests on the laptop market.

      • anoldamigauser

        In reply to bkkcanuck:

        Really? Everything rests on the laptop market? Surely you jest.

        ARM has significant advantages of power consumption and thermals for the datacenter (servers). ARM powers virtually every smartphone, a market that dwarfs the PC business. Laptops are a segment of the smallest of the three markets you mention; everything does not rest on the laptop.


        • bkkcanuck

          In reply to AnOldAmigaUser:

          No, I don't jest. The datacenter uptake is slowed by not having development platforms with the same processor at reasonable prices. If you cannot develop, test and then deploy on the same architecture then it acts as a deterrent to adoption. You can get an ARM based instances from cloud providers for half the price of Intel with the same performance... yet there is not this massive wave of movement because until you go through full development cycles and certify it on that architecture - you tend to deploy on the same architecture as you develop on (it is safer).


          While there are ARM based Linux kernels, not all of the linux distribution is fully available -- as the teams doing development are using Intel PCs running Linux.


          As laptops make up the greatest segment of the market -- it is the market that can provide the greatest chances of pulling the rest of the market with it. If you have more adoption with ARM based laptops because it has much more battery, more features, more performance within the thermal envelopes - it is the market from the non-server (also potential development platforms). If you can drive adoption in the laptop market that makes up 2/3rds of the users, which then provides a large enough market to get developers to transition their software to ARM native... and removing potentially 2/3 of the x86 chip market -- then the cost to develop each successive generation of x86 more expensive and that along with high performance and app availability would likely lead to manufactures to make sure there complete line is on the same architecture. Once you have all these machines available to load Linux on for development, the development of server application can have their complete cycle of development on ARM architecture - then it becomes much more likely the deployment on ARM at lower cost would be a no-brainer.


          x86 would not be this big in the datacenter if it was not for the it being the easiest to do based on all development being done on x86 development machines.

          • anoldamigauser

            In reply to bkkcanuck:

            And there you come to the crux of the issue. People do not buy a computer because of the processor (OK, maybe those of us here do), they buy a computer to run software. If the programs they want to run, do not run on the Operating System or Processor, then they will not buy that computer.

          • illuminated

            In reply to bkkcanuck:

            You can develop .net core services on windows and deploy to linux servers in the cloud. No need to have the same exact OS or even CPU.

            • bkkcanuck

              In reply to illuminated:

              That said, cross platform is a nice theory, but not in practice. The underlying binary that your code is relying on is different than developed on. No matter how thorough testing is, it never covers all cases that may be technical issues - but development would likely have already hit it and noticed it even if not part of the QA process. As such having development with the same architecture and binary is essential to ensure that the delivered product does not have surprises. It is the same reason we don't develop using Java 7 and then deploy on Java 8 based runtimes - in theory everything should run fine but that is theory. There are of course additional small assumptions developers can overlook when it comes to things - like characters (line separator different on Windows and Linux - a developer could use '/' or '' or the cross platform version. Differences between windows and linux file naming (some characters are invalid on one and not on another)... cross platform is nice in theory and may make transition to a new platform easier -- but having the whole development on the architecture and binary it is deployed on when millions or more dollars on the line... is important.

              • illuminated

                In reply to illuminated:


                You would have been right like 5-10 years ago. There are services in production that are developed on windows and run on Linux in the cloud. It is not a joke and it is not magic and it is not that difficult.


                .net core is mature platform now. If you can use Java you can use .net core.


            • bkkcanuck

              In reply to illuminated:

              I would never use .net on the backend of any system I would implement (market: banks, brokers, etc.)

          • wright_is

            In reply to bkkcanuck:

            You develop directly on the ARM servers. They are more powerful than a desktop or laptop anyway.

            That is how I started programming back in the 80s and how I've always done big Web development projects as well.

            At the last 2 software companies I worked at, they used thin clients and developed on terminal servers and the software was compiled on a staging server..

            The other thing is that ARM isn't ARM, the SoC is unique, even if the processors instruction set is the same, everything around it is different.

    • basic sandbox

      In reply to AnOldAmigaUser: I agree that normies don't care or understand processors but plenty of normies wish their laptops weren't hot, didn't have loud fans, and had better battery life.


  5. Alastair Cooper

    Apple set the standard in 1984 with the Macintosh and still lost in terms of market share to the PC once it caught up. It's very possible history will repeat itself.

  6. dftf

    Apple clearly when it comes to Apple devices, as they will only company allowed to go into them.


    For "Windows on ARM" devices, I'd imagine either Qualcomm or AMD. Though I still wonder if "Windows on ARM" actually is the future, as you suggest, or whether Windows simply dropping 32-bit app support (Intel32) and supporting only AMD64 apps would work too. Then, 32-bit code could be removed from Windows 10, making the OS lighter, and having a smaller install-size, and AMD CPUs could drop Intel32 code and would become more power-efficient, secure and faster as a result.


    I'm sure they would still be a gap between an Apple ARM processor and a future 64-bit only AMD64 CPU, but removing Intel32 code should narrow-the-gap. (Though what would that mean for Intel, I wonder: currently AMD and Intel have an agreement where AMD can implement Intel32, and Intel can implement AMD64. So if Intel32 could be removed from AMD's chips, would that mean Intel could not produce AMD64-only chips? Maybe Intel would become an ARM licencee?)

  7. wright_is

    Fujitsu are winning it at the moment. There latest supercomputer is around 3 times as fast as the AMD machine in second place, istr.

  8. shark47

    My money is on AMD. Qualcomm doesn't care about the PC market. I think AMD has the most to lose if they lose gaming.

    • bkkcanuck

      In reply to shark47:

      If AMD comes out with an ARM processor (now that they have a little more healthy financial situation), I would put part my money on AMD.... the other one I would not rule out is nVidia (depends on what there focus is). I do wonder however at what price point either of these companies will enter the market at - is going to be similar to their other offerings and charge a premium and focus only on thermals.... or are they going to compete to reduce the overall cost of devices.

      • jackwagon

        In reply to bkkcanuck:

        I know AMD once upon a time had ARM-based Opteron chips. As far as I know, they haven't ruled out using ARM in future architectures, but right now they're back on AMD64.

        • bkkcanuck

          In reply to Jackwagon:

          They shelved their ARM project (but not completely - it was still ongoing apparently) while they worked to save the company and that meant focusing on those that could return the company to health (Ryzen first, GPU second) [limited R&D budget]. They have returned the company to health and rumours are surfacing that it is being fully reactived (makes sense to cover all the basis). They obviously have a good team in place for the silicon design, so it is not a big leap that they could do a reasonably first version (they still apparently have an active ARM Architectural license).

    • Truffles

      In reply to shark47:

      I have trouble imagining game devs creating a port of their AAA-titles just for Windows on ARM. Consequently I have trouble imagining AMD spending a few billion dollars developing ARM chipsets just for the gamer market.

      • bkkcanuck

        In reply to Truffles:

        It really depends on the size of the market. Consoles vs PCs, PCs make up 30% of the game market so consoles make up maybe 70% for both X-Box and PS... so Maybe PS made up 40% overall when it came to PS3 (which would have been a smaller market back then). The PS3 ran basically on modified PowerPC chips. Games are mostly written in C++ and as long as the OS APIs are basically the same, the cost to port would be much much less than 'a billion dollars'.

        The gamer market is still a niche within the PC market. If a manufacturer came out with an ARM process that was substantially better for consoles - I could see them being incorporated into nextgen consoles. (consoles are more thermally restricted than PCs can be since they tend to be focused on a smaller size - even if larger this time).

  9. ngc224

    Apple has such a lead, I don’t think they’ll be displaced anytime soon.


  10. illuminated

    How much better is Apple ARM CPU compared to Intel or AMD CPUs at the same energy and performance levels? So far I have seen only hype without numbers.

    • bkkcanuck

      In reply to illuminated:

      I have seen lots of numbers and use cases from different reviewers (both those that are Mac oriented and some that is Windows oriented) ... Just searching youtube and watching them - I think you will find that performance and thermals are far outstripping any Intel processor within the same thermals. Also performance is challenging the top of the laptop lineup (current generation running Intel)... even on the fanless one (there will be a downshifting of around 20% or so after 10 minutes of sustained high performance usage -- on the fanless MacBook Air). These are only the entry level processor, there is likely 3 or more still to come in this family for higher thermal laptops and desktops...

    • Greg Green

      In reply to illuminated:

      There’s plenty of hard info out there, you just have to be willing to look for it. I’ve seen a dozen videos on YouTube where they compare arm macs with intel macs on real world applications. Then there’s the websites like MacWorld, 9 to 5 Mac, etc. and it was done on 20w arm systems against 200w Mac systems


      Its only hype if you’re afraid to look for hard info.

  11. peterc

    I think people are missing the issue of scale. Mobile arm chip shipments dwarf total laptop/pc sales by a significant factor. I doubt it’s financially viable for the chip manufacturers you mention to create a chipset for a niche segment of the total pc sales business only. It’s not enough numbers.


    Don’t forget apple have created a silicon series for their complete iOS and Mac OS product line. They have the benefit of manufacturing and investing at scale. However if enough OEMs and Microsoft are prepared to order and pay for a chip series they can all use and are committed to for say 5 years then it might happen.


    It maybe that Samsung is the company who can manufacture at scale the chipsets required. Combine Samsung and Microsoft and you might get an interesting product offering, maybe...

    • Truffles

      In reply to peterc:

      I wonder if they're tempted to enter into new market segments?? They've got a few years lead on their competitors, they've invested billions refining their chip designs, and higher production will take up otherwise free 5nm fabs, thereby making it more expensive and slower for competitors.


      Just winging it here, but I can see them introducing a fast but very low power server-farm style device that's just a CPU in a box, stripped of all the bells and whistles seen on consumer devices such as GPU, ML etc.


      Less crazy, how about an XBox / PS5 competitor? The Apple TV surely needs a change of focus since it was first released.

      • bkkcanuck

        In reply to Truffles:

        I don't think Apple TV needs a change of 'focus' at this point, what they need is a compelling gaming performance for 80% of the gamers out there. The focus can safely remain on this is your device for your living room entertainment, and BTW, you can play your favourite games on this device.

        Just scaling taking one of the higher end SoCs (with more graphics on package) would likely be sufficient to having enough performance for 80% of the market. (maybe more). [maybe multiple Apple TV devices - an Apple TV Mini for TV mostly, and an Apple TV Max with more performance for games.

Leave a Reply