Flexing financial muscles, Arm aims to elbow into Windows PC market Chip designer Arm predicts that PCs based on its architecture will account for a significant share of the Windows market within three years as the company claims record revenues for the quarter just ended. The biz is pinning growth prospects on more than just AI, despite riding the coattails of the industry craze for anything …
I shrieked with laughter today when I read that Oracle had renamed its database from Oracle 23c to Oracle 23ai.
Everybody's trying to get onto the AI bandwagon!
https://www.infoworld.com/article/3715303/oracle-renames-database-23c-to-23ai-makes-it-generally-available.html
I am running the ARM version of Windows11 on a M1 Airbook (assisted by the Parallels VM software). It works really well. There may be a slight delay when an app provided for the X64 architecture is used for the first time and translated on the fly into ARM instructions. I will be happy to buy an ARM desktop PC to run Windows when there are a few more choices available.. Laptops and PCs that don't need fans are the way to go.
The article mentions laptops specifically as where they think they will take significant market share, so businesses would seem to be the main target, but for those home users in the market for a laptop, a power-efficient ARM machine may well be a tempting option, if enough native Windows apps are available.
"if enough native Windows apps are available"
This is going to be the problem. I assume that x86 apps will introduce a fairly hefty performance hit (just a guess, never having tried Windows on arm, but it seems reasonable to me), so you're going to need native apps to take advantage of the alleged benefits.
Either the lack of apps will kill, or at least wound, the platform, as with Windows phone, or demand for the platform will have developers scrambling to make native apps. I'm guessing the former is more likely.
How easy is it to port an x86 app to Windows on arm, though? I assume it's not as simple as just recompiling for the new target? Is it something any old Windows programmer can do, or are there new tricks to learn? All of that is going to influence how many apps get ported - the more work is involved, the less likely a developer is to spend time porting to an unproven platform.
It'll depend on how Microsoft implements it, but when Apple did Rosetta 2 it was done with a pretty minimal performance hit, though they're vertically integrated from the silicon to the API, so they could custom tailor things in ways MS probably can't. For the most part, it should just be a simple recompile assuming developer stuck to published APIs and any supporting libs have ARM versions. If it's a .NET app it probably doesn't require anything on the part of the developer at all. I guess if you use SMID stuff you'll have to rewrite all that, but it should be pretty quick and relatively painless for most apps.
I dunno.
I have a colleague with an Apple laptop based on Arm. He has no trouble running Diablo IV at full resolution (not sure, but at least 4K HD). And there is no fan noise to trouble your ears.
I think that is pretty impressive, so much that I am looking forward to Windows on Arm. I wouldn't mind looking into that as an upgrade in the next few years (aka when I retire).
I have an Arm M2 based Mac mini and that is also silent apart from when I go max tabbage in Safari, when sometimes a very faint fan comes on briefly - oh .. it boots from cold in 10 seconds as well. .Of course it has a power efficient 5nm Apple M2 chip inside, and of course is running a Mach inspired BSD variant, not Windows. So Windows on ARM has a couple of headwinds to conquer apart from critical mass/software - process geometry and OS efficiency. Then there is the whole issue of defining a "standard" ARM PC hardware architecture that multiple vendors can target.. not just a sequence of one off and incompatible PC/laptop designs.
(boo hiss).
Moving from an intel MacBook to an ARM powered one was almost transparent. The apps get translated once and it is 'job done'.
I have an M1 based MacBook Pro and while there are a few issues I won't go back to intel anything.
The fact that I can work all day disconnected from power is a big plus. My main regret was not getting more RAM. 16Gb is not enough for my workflow. I'll be keeping it until the end of the year when I'll get an M4 based device.
I'm sure that MS will find a way to shackle their OS on the latest ARM kit. The WINTEL alliance needs to keep on working until Intel can match ARM performance including power consumption.
The benchmark laid down by Apple with their SOC style packaging has set a pretty high bad for others to get over. I'm sure that Qualcomm and Broadcomm will get there in time. Intel? all bets are off IMHO.
Whereas I do... people who are avid gamers won't switch, but they'll often be using a desktop anyway - and I don't actually think they're a particularly large slice of the home market (a very vocal and engaged slice, but I don't think they are a large slice)
People will value the ability to use a laptop for tens of hours at a time without carrying a suitcased sized UPS.
and it didn't work out then. So, what do you [Arm] think has changed this time?
I keep telling everyone that the very existence, benefit, and market share of Windows depends upon application compatibility. People don't necessary "care" about the OS, Windows, itself - they / we use Windows because it runs the applications we use to actually get things done.
Last time, x86 compatibility on Arm depended upon an emulation layer. What has changed this time? Windows on Arm was a huge disappointment because native application development fell flat on its face compared to x86 app availability. Arm will quote
"primarily because of the level of experience that we've seen in the other ecosystem, the fantastic performance, the great battery life, the fact that you can build a high-performance machine minus a fan."
but all that takes a huge hit when you run your apps inside an emulation layer. Battery life? Fantastic performance? Only if you are running native apps...and will native apps for Windows on Arm actually be there for the market volume you hope to attain?
-------------------------------
I feel this is just another CEO / MBA pipe dream. You know, like Broadcom's. These fools keep thinking that "Because we proclaim it, it will be True".
Through an emulation layer. And why should Joe Average move to, and learn how to manage, an incompatible OS just to be forced to run an emulation layer (and learn and deal with that, too) in order to get their old programs back??
It makes absolutely no sense, yet Linux fans keep up the dream.
I wonder how many Apps do most people actually use. I've just looked at what I have installed.
In the last month I've used: FireFox, Chrome, Thunderbird, LibreOffice, Acrobat Reader, Word & Excel viewers
I would not expect any problem running those. Also Basic PAYE tools which I'd hope would run.
In the last year I've also used:
Garmin Connect. My new Garmin watch now uploads via my phone so no longer needed
Andica Tax Self Assesment for Partnerships. I'd expect it to run o.k. under emulation
File Zilla FTP. If it had a problem I'm sure another FTP client would be available
Datapower Reader. Gives me access to various old databases it would be a pain if it didn't run o.k. under emulation but that is the only program that might be significant
I have a few other Apps installed that I've not used in the last year.
The ones I anticipate to need again are:
Irfanview & Zoom
I think my wife only ever uses:
FireFox, Chrome, Thunderbird, LibreOffice, Acrobat Reader & Word viewer.
I think there is a bigger market than many IT professionals think.
Not sure why ARM stock price responds as it does (better now than at the 09/23 IPO, but not as good as this past March). Long term, I'd expect it to at least follow a similar trend to that of other IP vendors (esp. Synopsys and Cadence) that went up by factors of 3x in the past 5 years (not necessarily as good as Supermicro, Nvidia, or AMD, but better than Apple, Microsoft, and IBM).
The challenge with "mass" adoption of ARM probably lies with the many bugs in Windows that x86 willfully adapted itself to, over the years, IMHO (Apple's OSes, and Android, had no such issues).
ARM looks really good for the next year or two as it will benefit from the AI bubble without being hit by regulation - it does silicon not software.
I would prefer to see a cheap retail Pi (or other SBC) system with a works package, browser and media player. Boxed, just open and use, sold on the high street and online, backed by a brand. But not Microsoft, Apple, Google or Amazon.
Cheap SBCs are now powerful enough for this. Code can be produced and rolled out quickly. An internet capable revamp of the 80s. Let's go back to the future and do it all over again, but without the limitations and restrictions imposed by GAFA. Distributed versions of internet services as standard.
One of the drivers for x86 adoption was a royalty free PC hardware standard that picked up where the IBM PC-AT spec left off. Various vendors defined vendor neutral standards for BIOS/UEFI boot ROM, ACPI power, PCIe bus, PIC/APIC interrupt mappings, standard I/O locations, RVI/EPT/AMD-V virtualisation facilities, 3D gaming etc etc..
Since the Compaq 386 broke away from the PC-AT reference manual "standard" in 1986 the industry, rather than IBM, have had a vendor neutral target architecture. This, in my mind at least, has fuelled adoption, with successive higher performance implementations from multiple vendors driving price/performance to an unprecedented degree, while maintaining backward compatibiltiy.
We obviously have a few existing ARM "PC" hardware standards although I am not sure any of them are vendor neutral? Or designed for enhancement with backward compatibility ?
- Raspberry PI 5
- Apple Mac Mini/MacBook
- Microsoft Project Volterra
- Microsoft ECS LIVA Mini Box QC710 Desktop (Snapdragon)
Do any of The Register readership know if there is a "de-facto" standard for ARM "PC"s ? If I have this all wrong and there is some sort of agreed hardware reference standard for ARM "PC"s, (or some sort of effort to define one), then I would be keen to understand what it is ? Hopefully it's not something Microsoft has locked down and demands royalties for ? That would be doomed to failure in my mind ... :-)
Windows will always be tied to x86. There's literally no incentive for Microsoft to migrate Windows towards ARM since it would lose the ability to run more than 30 years or legacy software. ARM may be somewhat more power frugal, but that doesn't make up for the loss in software. Not by a long shot even.
And we've been here before, with Windows RT about 5 years ago. Little has changed since then The drawbacks are still there.
I am not sure that windows will never be on ARM.
From reports, Windows 11 has copied the Mac in regards to the central taskbar etc ?
If so, then the long battery life requirement for laptops will be a must have. I don't see Intel or AMD achieving that with their current processor offerings, so Windows on ARM may be more probable than not.
Intel and AMD may be able to make their offerings more competitive by dropping the 16/32 bit features of their processors in favor of pure x64 bit software. IOW, follow Apple's lead when 'Catalina' would only run x64 opcode. This made Apple's transition to ARM much easier, but an x64 only processor would have fixed length instructions making decode simpler and faster.
The current Intel instruction set has remnants of the 8008 instruction set, which I was first exposed to 51 years ago.
All bootloaders assume legacy non-protected mode. Yes, you could rewrite bootloaders to support 64-bit protected mode on startup but that would leave a large compatibility gap since only newer operating systems would support that feature.
It means you can't run older operating systems on your x64 processor and that would be extremely problematic.
I have to say this but not being able to run older operating systems may be extremely problematic for you and some others but for the majority of PC users it wont be something that have ever even thought about. To most people the hardware and the OS are the same thing so as long as the PC is cheap enough and does what they need it to do they wont notice if the CPU an ant farm
We are talking about PCs not servers, and being honest the biggest market for PCs is probably business and they are mostly laptops or small form factor PCs (a laptop without the battery, screen.....). Most software on a business PC is Windows, MS Office/Google Docs and a browser, Microsoft make all of them and can make ARM native versions with out any legacy stuff in there, for the small percentage of business users who need to do thing that cant be done in native ARM versions there can be a translation layer and the software will run a bit slowly, and for the people what cant have things running a bit slowly they probably already have not standard PCs so they can continue to have none standard PCs with a different x64 CPU as required.
For Personal use, most people use a phone or if they need to type something they will have a laptop or maybe a tablet with a keyboard, and they will probably use MS Office/Google docs
Gamers as has been said before a a very small but vocal and engaged group, and they will be using desktops and different CPUs but even then they wont care too much a but the CPU as long as it can work with their graphics card of choice and run the latest games with a framerates of x with all the details max.
The really really tiny group, like us who read the register, will probably have an old PC to run older operating systems on, or have some form of server to play with and this is about PCs not servers so they are excluded from this.
I can assure you that companies like HP or Dell will not take the risk. If only 10% of those customers decides to install a server operating system or older Windows version the returns and support calls alone could wipe out their ENTIRE profit.
And for what? Slightly better battery life to compete with ARM?
> "I can assure you that companies like HP or Dell will not take the risk."
Huh? It's not like they have to replace their whole lineup. They may well not be rushing to be first into the market, but if ARM laptops running Windows start to account for even a tiny fraction of total laptop sales, they will happily add a SKU or two to the mix. They can't afford to get caught with their pants down.
I know Windows RT seems like it was only about 5 years ago, but its approaching its 12th birthday this year, it was released along side Windows 8 and it also had the awful 'metro' interface which made it look like a laptop from the Fisher Price preschool range. And in those 12 years the performance of ARM SOCs has got way better.
I think for some business Windows ARM laptops maybe useful, Office 356 and Chrome/Edge have native versions and the X86-64 to ARM translation layer is probably good enough for running most business apps that don't have a ARM native version. As AFIAK it caches the translated ARM instructions once you have run the program for the first time, meaning on subsequent usage the performance improves as it can run the cached ARM instructions directly without needing to translate.
... is less-conplete than you think. MS keeps changing their programming interfaces, data access methods, and libraries. Much old Windows software will never be rewritten for newer versions of Windows, because the companies which originated the software have gone out of business, or been bought up, then axed.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
