ARM in PCs

[INST2011]

Hello,can anyone give me a balanced view on ARM's chances in PC's over the next five years?

 

[iam2thecrowe]

i got my arm stuck in a PC once. I wouldnt recommend it, chances of surviving without consequences are low, you can get some nasy cuts and scratches.

 

[math1337]

when there ae wall plugs, why not use a power guzzzling, fast x86 chip

 

[loneninja]

I don't think they'll have much of a chance, even with Windows 8 supporting the ARM platform there will still be software compatibility problems. Currently ARMs strong point is very low power consumption which makes it great for mobile devices, but they're lacking in raw power.

 

[MU_Engineer]

Hello,can anyone give me a balanced view on ARM's chances in PC's over the next five years?

Sure.

- Windows 8 for ARM will be a netbook/mobile device OS because full-power desktops/laptops will need to run x86/amd64-only programs. Full-on CPU emulation will be required to run those programs on ARM CPUs, and ARM CPUs aren't anywhere near fast enough to run x86/amd64 programs in CPU emulation mode. A current top-of-the-line 1 GHz ARM9 CPU running in x86 emulation mode is about as fast as a Pentium 100. You'd need about 20 times the performance to be able to get acceptable performance in emulating x86/amd64 programs. That's just not going to happen in the next five years. Getting rid of old legacy x86/amd64 Windows programs in favor of native ARM programs also isn't going to happen in five years either. Just look at how many businesses still have Web applications than *only* run on IE6, despite IE6 having been replaced by IE7 for about five years. Rewriting a Web application to work with a newer browser is arguably easier than porting a compiled application as well.

- ARM will not seriously break into high-power devices due to a complete and utter lack of a >32-bit ISA, and pretty much every full-performance unit today has 4+ GB of RAM and needs more than 32 bits of address space. ARM will probably debut a 64-bit version in the next few years, but it would have to make up the 10 to 20+ year head start that the current high-performance ISAs currently have on ARM in the >32-bit memory addressing world. MIPS64 debuted in 1991, SPARC64 in 1995, Intel's 36-bit PAE address mode in 1995, POWER64 in 1998, and full-on 64-bit x86 with amd64 in 2003.

- ARM-based units will nonetheless likely become the most-common "computers" out there as people use very low-power mobile devices like smartphones and netbooks in place of real computers due to portability reasons. Serious users will continue to use conventional high-powered CPU ISAs like amd64 and POWER and not ARM-based units.

The only real non-x86-based computers I can see being developed as real competitor to current high-performance CPUs are units like China's Godson/Loongson units. Those are MIPS-based, not ARM-based, and MIPS has a much better chance of being a high-performance ISA than ARM since MIPS was once a high-power supercomputer ISA. ARM at beast was a midrange PC ISA back in the early '80s with the Acorn PC, then it became solely a low-power ISA. MIPS was actually the first common 64-bit ISA out there as MIPS64 debuted in 1991. MIPS64-based SGI workstations and servers were quite potent in the '90s, until SGI drank the Intel kool-aid and spun off MIPS and adopted Itanics in their product line and then went bankrupt as Itanic stank.

So all in all, I don't predict ARM will be a real threat to high-power amd64/Intel 64 or POWER-based machines. It will nip at the heels of low-power embeddable PowerPC and Bobcat/Atom x86, but it won't do much more. MIPS64 as pushed by Chinese firms like Loongson has a much larger likelihood of challenging x86/amd64/POWER, but until China decides to put money into developing their own tech instead of ripping ours off, amd64/POWER will reign supreme as high-power ISAs. I hope China does get serious with pushing the snot out of MIPS64 as a high-power MIPS64 unit would really give AMD, Intel, and IBM a run for their money and be good for the market. However, China's track record isn't so great with developing truly novel technology- they have a long history of preferring to rip off "Western" technology and manufacture parts for cheap using it rather than brew their own tech from scratch to make truly innovative goods. China certainly has the capability of being the undisputed tech powerhouse of the world due to their massive population, financial resources, a good work ethic, and an education system heavily biased towards STEM disciplines. They just have yet to really tap those resources.

 

[Supermuncher85]

The adoption will heavily rely on future software design. Right now the majority market is for x86. To optimize for ARM most code has to rewritten completely since I there no real efficient re compiler. Therefore the question becomes, is it worth it for companies? If you are a consumer or a business you may want/need that Microsoft office, or that CS5. ARM's success will depend heavily on the software developers shoulders. It will either make or break them (good JDKs from their side will make things much easier).

However with smart phones adoption becoming more and more universal, there is already a great amount of software available and the faster their chips get, the more likely they are to be useful in desktop situations. Imagine being able to buy word for android and use it on both your phone and ARM computer.

Also windows 8 is supposed to have arm support, and Apple is rumored to be switching to arm in 2015 when their A15 design is planned to be released. From a company standpoint this makes a lot of sense for Apple. This means they can then use the same software branching over from one device to the other, only needing minor fixing or scaling issues. Currently 40% of their quarterly profits comes from Cellphone sales and will most likely continue to be their cash cow for the foreseeable future.

Will ARM completely replace x86? Unlikely. I see them evolving in parallel until quantum computing arrives. There will still be many cases where one or the other is better (be it power, speed, cost, or something else), so I can well imagine when going into a store in 2016 that there will be a wide variety of machines around to chose from depending on what use you need it for. Ultralow power devices for home automation, or powerful number crunchers for data-mining, CAD or GPGPU computations.

Welcome to the post PC era!

And before anyone starts commenting on that last sentence my quick definition for post PC: a time where not all digital actions are solely based on the use of personal computer but can be used with other devices. Or simply put, 15 years ago when you wanted to check mails, check the web, or edit a spreadsheet you had to use a PC. Nowadays you can use your cellphone, a tablet or other device to achieve the same goals without the need to own a PC, therefore post PC.

 

[MU_Engineer]

The adoption will heavily rely on future software design. Right now the majority market is for x86. To optimize for ARM most code has to rewritten completely since I there no real efficient re compiler.

Also, decompilation is prohibited by almost every proprietary software license out there as well, so even if there was a good decompilation/recompilation system, it would probably get sued out of existence. About the best you could do is to do "dynamic translation" a la the Transmeta Crusoe/Efficieon, where program binary's amd64/x86 instructions get fed directly to the CPU and the CPU translates from x86/amd64 into native ISA code. That was a lot better than full-on software CPU emulation as the translation got roughly 80% of native CPU speed compared to 10% or so for purely CPU emulation, but it's still notably slower and more power-hungry than native code. I believe it also added quite a bit of complexity to the CPU, so you further reduce theoretical CPU performance by devoting needed transistors and watts to translating from x86/amd64 to native compared to just executing native code.

Therefore the question becomes, is it worth it for companies? If you are a consumer or a business you may want/need that Microsoft office, or that CS5. ARM's success will depend heavily on the software developers shoulders. It will either make or break them (good JDKs from their side will make things much easier).

History would say the question to "is it worth it?" is a resounding "no." Like I said before, there are TONS of currently-used business applications that refuse to run on anything newer than 32-bit Windows XP + IE6. It took 6-7 years after amd64 CPUs debuted for Windows to considered usable on that ISA, and much of that usability is based on the ability of am64 CPUs being able to directly execute old 32-bit x86 code directly. The only people who run exclusively or nearly exclusively 64-bit programs on amd64 machines are running Linux or another open-source UNIX-based OS, not MacOS or Windows.

However with smart phones adoption becoming more and more universal, there is already a great amount of software available and the faster their chips get, the more likely they are to be useful in desktop situations. Imagine being able to buy word for android and use it on both your phone and ARM computer.

The problem with smartphones isn't processor speed, it's the limitations of form factor of the device. A smartphone with a 4" screen and maybe a thumb keyboard (or worse, an on-screen keyboard) isn't going to be a very pleasant machine on which to run a word processor and type something. Netbooks with their 90%-sized keyboards and 10" screens are painful to use for using a word processor; a smartphone would be absolutely awful. There is a reason why normal-sized 14-16" laptops and desktops with 20+" screens still exist- they are simply better machines to work with.

Also windows 8 is supposed to have arm support, and Apple is rumored to be switching to arm in 2015 when their A15 design is planned to be released. From a company standpoint this makes a lot of sense for Apple. This means they can then use the same software branching over from one device to the other, only needing minor fixing or scaling issues. Currently 40% of their quarterly profits comes from Cellphone sales and will most likely continue to be their cash cow for the foreseeable future.

Not surprising, Apple fans will buy anything with the half-eaten fruit logo. The highly impractical and extremely expensive MacBook Air, the overheating Intel-powered laptops, and the similarly-overheating and notoriously fragile iPhones not driving Apple into Chapter 11 is testament to that fact. Jobs could probably put an apple logo on a silver-colored bag of horse manure and sell it for $200 and still find buyers.

Will ARM completely replace x86? Unlikely. I see them evolving in parallel until quantum computing arrives. There will still be many cases where one or the other is better (be it power, speed, cost, or something else), so I can well imagine when going into a store in 2016 that there will be a wide variety of machines around to chose from depending on what use you need it for. Ultralow power devices for home automation, or powerful number crunchers for data-mining, CAD or GPGPU computations.

Welcome to the post PC era!

And before anyone starts commenting on that last sentence my quick definition for post PC: a time where not all digital actions are solely based on the use of personal computer but can be used with other devices. Or simply put, 15 years ago when you wanted to check mails, check the web, or edit a spreadsheet you had to use a PC. Nowadays you can use your cellphone, a tablet or other device to achieve the same goals without the need to own a PC, therefore post PC.[/quotemsg]

Most people just want to do stupid social-y stuff with computers, which they did with telephones before computers were around, they did that with computers when the Internet was developed since it was more convenient than telephones, and then with smartphones since they were more convenient yet. They don't do much for real work with them. Real work will continue to be done on suitable machines, and suitable real computers will continue to be made. Sure, they might be more expensive and harder to find, but they'll still be around. Compare it to being able to buy good-quality non-powered tools once power tools became ubiquitous. You can still buy good crosscut saws, axes, and reel lawnmowers despite various powered saws, chainsaws, and engine-powered mowers becoming ubiquitous. They just are special-order items and cost more.

 

[Supermuncher85]

Also, decompilation is prohibited by almost every proprietary software license out there as well, so even if there was a good decompilation/recompilation system, it would probably get sued out of existence. About the best you could do is to do "dynamic translation" a la the Transmeta Crusoe/Efficieon, where program binary's amd64/x86 instructions get fed directly to the CPU and the CPU translates from x86/amd64 into native ISA code. That was a lot better than full-on software CPU emulation as the translation got roughly 80% of native CPU speed compared to 10% or so for purely CPU emulation, but it's still notably slower and more power-hungry than native code. I believe it also added quite a bit of complexity to the CPU, so you further reduce theoretical CPU performance by devoting needed transistors and watts to translating from x86/amd64 to native compared to just executing native code.

Right and we know how well emulation works. It sucks. Heck people complain when a program is using the JVM and not written straight out in c++. Emulation didn't work with PPC and it certainly won't work in the future either. Although there is what I would call an abundance of processing power in today's desktops, I doubt that an ARM CPU will be anywhere near the performance levels of what we have today.

History would say the question to "is it worth it?" is a resounding "no." Like I said before, there are TONS of currently-used business applications that refuse to run on anything newer than 32-bit Windows XP + IE6. It took 6-7 years after amd64 CPUs debuted for Windows to considered usable on that ISA, and much of that usability is based on the ability of am64 CPUs being able to directly execute old 32-bit x86 code directly. The only people who run exclusively or nearly exclusively 64-bit programs on amd64 machines are running Linux or another open-source UNIX-based OS, not MacOS or Windows.

This is unfortunate indeed. For that very reason I still run several machines with XP, since some of the software I just can't get to work via virtualization, or is not available anymore. Some might now so, oh just buy an updated version! Well a "new" version was never made and having a custom program written to replace what is working now is extremely expensive and simply not worth it. If it ain't broken, don't fix it.

But here I have to raise the point though, I don't think that ARM is necessarily trying to muscle itself into this niche in the first place. From a design point of view they are low power devices and would expect them excel at just that. When looking at lower power devices, I often think of Point-of-sale systems, used in retail stores and fast food chains. Many of these run on standard x86hardware. Why, when you could equally well use ARM for cheaper and less power consumption, from the ground up designed to work with touch. McDonals spent a fortune to streamline and standardize it's Point of sale systems, and I could very well see ARM being able to take a significant chunk here.

The problem with smartphones isn't processor speed, it's the limitations of form factor of the device. A smartphone with a 4" screen and maybe a thumb keyboard (or worse, an on-screen keyboard) isn't going to be a very pleasant machine on which to run a word processor and type something. Netbooks with their 90%-sized keyboards and 10" screens are painful to use for using a word processor; a smartphone would be absolutely awful. There is a reason why normal-sized 14-16" laptops and desktops with 20+" screens still exist- they are simply better machines to work with.

Well yes and no, in a sense speed was a major hindrance. It's not been till recently that mobile processors have made many modern application uses possible. When I look at the iPad I still marvel at the technology. It is truly mesmerizing at what it can do. I would never write a paper on one, but I must admit nowadays when I just want to quickly check an email, or toms I grab the tablet instead of powering up my PC. Yes size is an issue but to a large part this has been circumvented by the use of touch input. By removing the physical need for touch input, the screen estate has increased substantially in recent years. Remember were still talking about mobile devices that are meant to not replace but augment PCs. But your right I will never trash my desktop. I can't live without multiple screens and there have been multiple occasions while using excel that I just wanted to take a dump on my netbook. Saying that though, I have written a 30 page document on my netbook once I hooked up an external monitor and keyboard.

Going back to my last post as to what I would expect of android in 5 years time: I would not at all be surprised to being able to buy, much like today a sound-dock for your music player, a Screen-dock for my cellphone. A dock that will allow you to use your phone with the convenience and productivity benefits of a large screen or even multiple screens. Although I find the Motorola atrix itself a failure (the netbook cellphonedok http://www.motorola.com/Consumers/US-EN/Consumer-Product-and-Services/Mobile-Phones/Motorola-ATRIX-US-EN) because of the poor implementation, it is a technical demonstration of what is possible with android.

Most people just want to do stupid social-y stuff with computers, which they did with telephones before computers were around, they did that with computers when the Internet was developed since it was more convenient than telephones, and then with smartphones since they were more convenient yet. They don't do much for real work with them. Real work will continue to be done on suitable machines, and suitable real computers will continue to be made. Sure, they might be more expensive and harder to find, but they'll still be around. Compare it to being able to buy good-quality non-powered tools once power tools became ubiquitous. You can still buy good crosscut saws, axes, and reel lawnmowers despite various powered saws, chainsaws, and engine-powered mowers becoming ubiquitous. They just are special-order items and cost more.

In a sense I am hoping that ARM will become a big player in the future. Intel has enjoyed total market dominance in the last few years and a bit of competition is direly needed to keep technological progress going, not to mention price competition, which I as a consumer really want.

 

[INST2011]

Best answer selected by INST2011.

 

[INST2011]

SuperMuncher and MU_Engineer,

Thanks for your very complete answers!..now I just gotta get up to speed on quantum computing and 64-bit architecture...and how to change my handle

Thank again!

Alex (INST2011)

 

[Mousemonkey]

This topic has been closed by Mousemonkey