Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Intel still has plenty of margin to buy the market with, and their own fabs. They can keep a good portion of the server market until someone makes the cost to switch low enough to be attractive.

That said, I think they blew it the the cellphone market, and now this has haunted them. They aren’t competitive in the broader market, and more the cost to switch from arm there is high. AMD I think is a near term issue in the server market, but ARM is the real issue.



Intel’s previous acquisitions have been pretty bad.

It doesn’t matter how deep your pockets are if you’re unable to successfully integrate or manage what you buy, and this speaks to management/organizational dysfunction, which money also doesn’t easily fix.


Re: ARM - spot on. They sold their XScale/StrongARM assets right around the same time Apple switched to them. They were probably riding high on that win.

However, market growth switched so quickly from PC to mobile/IoT, they couldn't get in. Instead, they just wasted resources on projects like Moblin which became boondoggles.


In general, the elephant in the room is: what if Apple outperforms both Intel and AMD with their Apple Silicon? ARM is an efficient architecture and the have a process node advantage for 2021. If they wanted, they probably could make very convincing server CPUs. If they want to move as announced the Mac Pro to arm, they would need server class CPUs anyway


> what if Apple outperforms both Intel and AMD with their Apple Silicon?

It wouldn't matter at all to either AMD or Intel until Apple starts supplying these chips to other OEM's. Till then, it will just be used in the walled garden of Apple products, which is the very thing that makes it very unattractive for many users.


The Mac isn't a "walled garden". MacOS is a Unix. Losing Apple as a CPU customer adds another dent into Intels revenue. On its own, it wouldn't be decisive, but in a time of sinking market share, you don't want this to happen.

Also, it will be interesting how Apples market share develops, as th AS could make the Macs much more attractive. Both from the performance and price and of course, because of the iOS compatibility.


>Also, it will be interesting how Apples market share develops, as th AS could make the Macs much more attractive. Both from the performance and price and of course, because of the iOS compatibility.

Apple's pricing strategy over the past few years has been to reduce the entry-level costs of their devices thus increasing their marketshare, ecosystem lock-in and services revenue. It's the strategy they've employed with the entry level iPad, iPhone SE and Apple Watch SE, and I fully expect them to bring to same strategy to the Mac, now that they no longer have to pay Intel's profit margins.

We'll very likely see a MacBook in the $700 price range in the upcoming months. That would put the price of Apple's entry-level MacBook right around the that of an average notebook computer in the United States, while providing far better performance than their competitors in that price range (see A14 benchmarks). This would naturally provide Apple with a big marketshare boost in the consumer notebook space.

Ultimately nobody but Apple knows exactly what their pricing strategy will be with Apple Silicon based notebooks. However, if I were competing with Apple in this space, I'd be tremendously concerned.


Were I in Apple's shoes, I'd be making a laptop that's cheap enough that everything cheaper is a bad computer. I'd want half of all freshmen to get that laptop. $700 sounds about right, $600 would nail it.


> The Mac isn't a "walled garden".

On the software side you have Gatekeeper and SIP. On the hardware side you have T2, soldered RAM, their weird custom SSDs, and their lawsuits against hackintosh manufacturers.

Yes, there are still ways around some of them, but it's pretty obvious where they want things to go.


Gatekeeper has been around for 10 years and every year people have predicted you couldn't install software outside of the app store. It never came by. As a development machine, it just wouldn't work.


The frog needs to be boiled slowly. Very slowly.


No. For a general purpose machine, especially for developers, you need the ability to access your files freely and to run any program. If that is no longer given, all developers and most users would drop the platform. Also, it completely doesn't make sense as long as you allow virtualisation. Which Apple not only does, they even demoed this capabiltiy in the keynote.


Lawsuits against hackintosh manufacturers seem fair enough to me.


I think that’s a short term take. Apple may be able to reduce the price of their laptops by hundreds of dollars, offer better performance, offer better battery life, offer more desirable form factors, and maintain their margins.

It seems unlikely to me that PC OEMs will just stand by and take it. Windows on ARM already exists and this may be the spark that really ignites that fire.


That is my hope too. So far ARM on the desktop/laptop lagged both because of the lack of suitable hardware and software. There are a few interesting Windows on ARM laptops around, I had been eyeing the Galacy Book S, but they lag behind as there is very few native software for them. Which is a chicken and an egg problem. As long as the number of ARM machines is low, software companies are saving money by ignoring them. And this means, the numbers will stay low.

With Apple Silicon, the game changes a bit. Not only is there the prospect of the AS being really fast, as Apple is going to make a complete transition, software companies have to support AS, if they want to continue to sell to Mac users. And suddenly, any Linux user also has a great ARM platform to work on, via a VM running on the Mac.

This might even have a direct impact on the Windows software offerings. So far, they could sell to Mac users running a x86-VM. This will be no longer possible. As much as Mac users contributed to Windows verndors revenue, they have to either give up the Mac market, make a native Mac application or hope that Windows on ARM becomes available in a VM on the Mac. But then, they at least need to fully support Windows on ARM.


I suspect it’s mostly price that makes it unattractive to users. Worrying about the walled garden is a fringe concern.


They have almost 0 computer market share in developing markets, nobody knows how to use apple, nobody knows their stack and their programming languages, all the infrastructure is built around windows.

With smartphones it's not such big a problem cause you don't do anything productive on smartphones anyway.

With PC/laptops it's a big disadvantage I don't see changing in a few years just because of small performance advantages, even if the prices weren't an issue.


This depends on the use case. For phones and consumer applications, the walled garden is often tolerable. But imagine trying to run a cloud service, a FinTech system, a factory, an ERP system, etc in the walled garden.


The Mac isn't a walled garden, why should servers with Apple Silicon be?


Not as long as Windows is not a default OS.


If, and it's a big if, Apple wanted to build "Darwin in the cloud", they could make the Big Three into the Big Four pretty quick.

Those enormous piles of cash could build out a world-class service that would be attractive both to developers servicing iOS applications, and just to developers, period, a lot of whom use Macs.

That would indirectly take a chunk out of Intel's server dominance, and might get the other players thinking about whether Intel is the best platform for them to stick with as well.


> > what if Apple outperforms both Intel and AMD with their Apple Silicon?

> It wouldn't matter at all to either AMD or Intel until Apple starts supplying these chips to other OEM's. Till then, it will just be used in the walled garden of Apple products, which is the very thing that makes it very unattractive for many users.

Well, there is another possible path: if Apple Silicon is seen as a PoC, someone else could decide to pursue the server market with an Apple Silicon inspired ARM-based offering. Perhaps aimed at FaaS deployments or something similar.


There are already ARM based servers available in AWS. The big problem is: almost no one has desktop machines, where the software development happens, with ARM. Linus once commented that ARM doesn't sell so well on the server, because developers are lacking desktop machines with that architectore. Apple Silicon could close that gap. You can develop with a Linux VM on AS, and deploy on ARM in the cloud.


Around the first .com wave, I used to develop on Windows (C and C++) and deploy on SPARC/PowerPC/PA-RISC servers.

The only x86 servers were running a mix of NT 4.0 and Windows 2000.

Apparently it worked.


Of course it works. Otherwise there would be no ARM on the servers. But it is second class to a setup, where the development and deployment happens on the same platform. Having good ARM machines available on the desktop will give ARM on the server a boost. As I wrote, don't take my word for it, listen to what Linus hat to say on that topic.


Ironically UNIX is exactly one kind of platform where remote development is first class.

One just needs to properly configure mount points and remote sessions.

There was hardly any difference between my X sessions and local development.


Sorry, in general that is not true. You need a very good network connection, both with bandwidth and latency, to make remote devlopment workable. Still, it never equals local development. If you follow the discussions here on hacker news, which terminal software has the smalles latency, remote development never can compete with that.


Apparently having UNIX servers on premises is a forgotten art.


It is. Even if companies have their own hardware, it is often in separate compute centers. And of course anything on AWS or Azure is not on premise.


Not everyone is FAANG, or pretends to be one.


Nobody cares about server market, literally if it runs on abacus fast enough it's ok.

Developers! DEvelopers! Developers!

ARM should be in consumer space, millions of laptops for developers first running Windows and Linux


With the cloud and managed runtimes (or stuff like POSIX), I don't care one second what CPUs are running there (beyond endiness and using the right widths), I never wrote hand Assembly for such kind of deployments and the optimizer was good enough for our workloads.

The only issue would be libraries available only in binary form for AOT compiled languages.


someone else could decide to pursue the server market with an Apple Silicon inspired ARM-based offering

In other words Nuvia.


I could see Apple come out with their own in-house Cloud Computing platform within the next 2 years.

  - They would be vertically integrated with Apple Silicon
  
  - Apple themselves are increasingly depending on cloud services

  - Async/Await in Swift will likely land next year, making Server-Side Swift much more appealing

  - Apple is greatly increasing cloud / Kubernetes hires

  - Could share a single Apple Silicon ARM architecture from client(ios) to dev(mac) to cloud


Unless something has changed radically since MacOS Server failed, Apple would either need to do massive software engineering of a sort that they are unfamiliar with or they would need to use a different OS entirely to have a credible cloud offering. Mac OS is unlikely to be able to compete with Linux or FreeBSD in the near future even if the CPU is magically better.


They could have a very bare bone macOS based server operating system specialised for running virtual machines. Dont they have their own xhyve virtualisation layer?


Federighi mentioned during some interviews after the Big Sur presentation, that macOS is a great hypervisor for vitualisation, when asked about Linux on AS Macs. Running Linux in the cloud would equally easy.


A lot of the prelim work is probably already being done with their docker optimizations.


Apple’s in-house cloud stuff is mostly Java on commodity Linux servers. I don’t think they are looking to have macOS compete with Linux.

Most of the special sauce from GCP and AWS is the integrated software services anyway.


> Apple’s in-house cloud stuff is mostly Java

Can u give some more details? I am not really familiar with Apple eco-system

But wouldn't most of it (storage, web, streaming) be based on commodity C/C++ projects ?


It wouldn’t have to be exclusively macOS. You could likely choose your OS or it could be some kind of containerization.


What happened when Apple outperformed Intel and AMD with PowerPC? People kept buying Windows PCs.

What if Apple outperforms both Intel and AMD with their Apple Silicon? People will keep buying Windows and ChromeOS PCs.

AWS is testing the server market right now with Graviton2, basically a drop-in replacement for x86 that's faster and cheaper. It's going to be fascinating to watch and I hope AWS reports architecture usage share over time.


Is graviton2 really faster? I dont remember the exact details but didnt they still have a (long) way until reaching intel/amd performance? If i remember correctly they are currently only interesting because of their price.



From running my asp .net core api on Intel vs Graviton 2, graviton turned out to be roughly 15% faster. But the SSR of vue turned out to be roughly 10% slower.


That is almost 20 years past, many things have changed. Most importantly, the iPhone has happened. I am not claiming, that Apple completely takes over the PC market, just that the market share could considerably rise. Having the better processors and iPhone compatibility could do the trick.


Apple didn't make PowerPC chips. That's a huge difference.

First, Apple wasn't a supplier. If Silicon takes off, they could become a supplier. Second, the reason they're switching to Silicon is the same reason they switched from IBM to Intel - the chip suppliers didn't react to the changing consumer landscape.

By making consumer devices, Apple is much more in tune with that landscape than IBM/Intel is. They can adapt chips much better and won't be handcuffed by chip suppliers in the future.


Moving to their own chips also allows for much more aggressive pricing.


The arm MacBooks will mean developing ARM code for your server is easier than x86.

I highly suspect we’ll see a switch from intel to arm in the server space in next 5-10 years.

The cloud vendors have so much to gain by building their own arm chips.


I wonder whether Intel/AMD will start offering x86_64 licensing if AArch64 takes off on servers in a major way. You'd think that they would prefer to have competitors on their own architecture/terms than something out of their control.


I wonder how much easier? Could devs use the Rosetta emulator to do x86 stuff on the ARM Macs anyway?


Nvidia bought ARM so it would be interesting to see where this goes.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: