ARM Going 64-Bit To Compete In High-End Desktop Market

[Fourstaff]

ARM does not push out as much FLOPS as Intel/AMD, so other than your usual computers for entertainment, the current x86-64 will still dominate work computers.

 

[qubit]

Oh my! I wonder how it will hold up against x86-64 processors in performance, cost, and power consumption. The latter two are likely very well, but I'm not sure about performance.

They might be a greater threat to Via than AMD/Intel.

Oh, an ARM chip certainly can have the performance. It's all about how you engineer it. At the moment, they're tuned for low power use, with some performance. They're roughly equivalent to the Intel Atom in this respect: an x86 CPU built to sip power, which of course compromises computing power greatly.

However, tune the designs for out and out performance the way current desktop x86 chips are and you have a monster! So yes, having 64-bit chips out and running Windows natively is indeed a threat to Intel. As ARMs an efficient and streamlined RISC load/store architecture, I reckon that for the same number of transistor, heat and power use you would get more performance than x86. You just need to build them with the same high technology that currently goes into an x86 CPU to realise this.

But then you make a valid point here:

I wonder if applications coded for x86-64 will run on the ARM version of Windows. IA-64 applications, for example, don't work on x86-64 nor x86. IA-64 runs x86 code in emulation and generally doesn't do a very good job at it.

Put bluntly, if ARM can't run the plethora of x86/x86-64 Windows software out there, it's dead before it starts. The ARM machines would be relegated to very specific tasks (e.g. OS on point of sale machines).


Edit: But wait, there's .NET. Microsoft is likely to adapt .NET for Windows 8 ARM processors which means all .NET applications compiled for "Any CPU" will run without modification on ARM.

No, x86 apps most certainly do not run on it and that is a major catch 22. ARM & Microsoft need to work together to make the mother of all emulators, so that x86 performance is reasonable, while ARM performance is blisteringly fast and hence an attractive target to code for. It's a tough call and will require significant market penetration of ARM processors to pull off, but this is the world arena and Microsoft at least, has the resources to pour into it so they could help ARM out here in some sort of partnership. What we need is a WOW emulator that works seamlessly as the current one and with high performance - a tough call indeed. It's well known that there's no love lost between Microsoft and Intel, so this is quite a realistic prospect.

 

[xBruce88x]

If the guys that handle DOSBox can do it i'm sure ARM and Microsoft can...

http://androiddosbox.appspot.com/

 

[Syborfical]

AMD holds the x64 license,

AMD holds the X86 64 bit license. Otherwise DEC alpha , sun, HP and everone else who has bought out an 64 bit processor would owe them money.
AMD owns the license x86 64bit license. you could bring out any other processor thats no x86 and AMD can't do a thing...

I would be nice to see a non x86 desktop cpu. ...
I can't see arm doing any worse than AMDs latest offerings

Bring it on

 

[FordGT90Concept]

Oh, an ARM chip certainly can have the performance. It's all about how you engineer it. At the moment, they're tuned for low power use, with some performance. They're roughly equivalent to the Intel Atom in this respect: an x86 CPU built to sip power, which of course compromises computing power greatly.

However, tune the designs for out and out performance the way current desktop x86 chips are and you have a monster! So yes, having 64-bit chips out and running Windows natively is indeed a threat to Intel. As ARMs an efficient and streamlined RISC load/store architecture, I reckon that for the same number of transistor, heat and power use you would get more performance than x86. You just need to build them with the same high technology that currently goes into an x86 CPU to realise this.

But x86 has been extended to accelerate virtually every task a personal computer sees and that includes insanely high precision (128-bit or more) floating point operations. These are things ARM is not intended to do. x86 creates a very generalized processor whereas ARM makes a very specific processor. Which continues into the next point...

No, x86 apps most certainly do not run on it and that is a major catch 22. ARM & Microsoft need to work together to make the mother of all emulators, so that x86 performance is reasonable, while ARM performance is blisteringly fast and hence an attractive target to code for. It's a tough call and will require significant market penetration of ARM processors to pull off, but this is the world arena and Microsoft at least, has the resources to pour into it so they could help ARM out here in some sort of partnership. What we need is a WOW emulator that works seamlessly as the current one and with high performance - a tough call indeed. It's well known that there's no love lost between Microsoft and Intel, so this is quite a realistic prospect.

Can you say "Windows Phone 7?" It hasn't got significant market penetration yet but it will in 5 years. The thing is, Windows Phone 7 is all about the floating point operations. It is practically a GPU with some arithmetic operations slapped on for the more mundane stuff. The most demanding thing the phone does is play back video and perform animations and it does a very good job at it. Expect it to search a large database though and it expect it to be slow as hell.

In short, I think ARM can make a good processor for the everyman (like AMD APUs and Via platforms) but don't expect it to be good for anything more demanding than video playback. For example, it wouldn't be able to run games, compile anything, encode video, edit massive pictures, etc. worth a shit.

I think what Microsoft is aiming for is something that can compete with the Google Chrome Book by substantially bringing the costs of basic internet computers down and at the same time, provide a lot more flexibility to OEMs to turn it into whatever they want.

As I think about it whilst typing this, it sounds like Microsoft wants to pull and Apple and make Windows-powered disposable appliances available rather than something that is maintained like the Windows world currently is (less Phone).

 

[qubit]

But x86 has been extended to accelerate virtually every task a personal computer sees and that includes insanely high precision (128-bit or more) floating point operations. These are things ARM is not intended to do. x86 creates a very generalized processor whereas ARM makes a very specific processor. Which continues into the next point...

Yes, that's true now, as they were aiming for the embedded, low power market. However, the design is very modular, so the extra functions will be added to make it a viable desktop competitor. Yes, it's a tall order, but these aren't small-fry companies we're talking about and certainly not with Microsoft as a partner.

Can you say "Windows Phone 7?" It hasn't got significant market penetration yet but it will in 5 years. The thing is, Windows Phone 7 is all about the floating point operations. It is practically a GPU with some arithmetic operations slapped on for the more mundane stuff. The most demanding thing the phone does is play back video and perform animations and it does a very good job at it. Expect it to search a large database though and it expect it to be slow as hell.

In short, I think ARM can make a good processor for the everyman (like AMD APUs and Via platforms) but don't expect it to be good for anything more demanding than video playback. For example, it wouldn't be able to run games, compile anything, encode video, edit massive pictures, etc. worth a shit.

I think what Microsoft is aiming for is something that can compete with the Google Chrome Book by substantially bringing the costs of basic internet computers down and at the same time, provide a lot more flexibility to OEMs to turn it into whatever they want.

As I think about it whilst typing this, it sounds like Microsoft wants to pull and Apple and make Windows-powered disposable appliances available rather than something that is maintained like the Windows world currently is (less Phone).

You might well be right, which would change the picture again, so we'll just have to wait and see - think the current fashion for "cloud" computing which basically puts a low powered terminal in your hand and gives all your data and control to some third party running big-iron servers. However, I don't think Microsoft really wants to have to develop and maintain two versions of the OS in the long run, as that really is expensive. It looks more like they want phones, tablets and PCs to all work the same way, hence the Metro interface on Windows 8. Therefore, I believe they will try to transition Windows to ARM. Microsoft perhaps aren't happy being stuck in the x86 world and want something better to support, I don't know. I can't see any other reason for starting support for an alternative processor with Windows 8.

But this is all wonderful speculation, which is why I ended my article with "The next decade will tell."

 

[orionbg]

if i remember properly

ARM licenses some of there tech from AMD

AMD holds the x64 license,

ARM does good AMD does good as AMD receives money as well do to those licenses,

ARM entering the x64 market means AMD makes more money from ARM and of course gets money from Intel,

Another thing to consider is ARM entering the desktop and mobile sector means while they may not be as powerful as there competitor tablets and netbooks using ARM cpu could improve battery life even further without sacrificing features. forcing Intel and AMD to release even more competitive products,

right now AMD can get down to 9w -18w on a single - dual core low power APU in the 1-1.4ghz range

ARM has quadcore chips already that can fit in smartphone power requirments,

now they might not support all features other CPUs do, the key thing to remember here is games, movies, music, web browsing dont really need all those instructions set, meaning the typical end user in the near future could end up with an ARM quadcore that uses less then 10w but feels and functions the same as an Intel or AMD dual core- triple core in day to day tasks. as ARM is already capable of creating octo core RISC cpu.

so as it stands ARM can already products 8 core mobile chips that are low power, they can also already adress up to 1TB of memory if needed, alot of there tech is lic direct from AMD, so alot of AMD tech ARM has complete access to, should be fairly interesting to see what happens.

ARM will not be paying anything to AMD because there 64 chips will not have anything to do with x64! Their 64 bit architecture is ARM based and not x86!

 

[AsRock]

A massive wakeup call it is but i doubt ARM has the money to buy out AMD and its Patents. Not for at least for 20-30years.

ARM are still a small company, but they will expand slowly and take the place of AMD (hopefully) buying patents are a different story though....

IMO AMD should spin off their graphics department so we can have ATi again

It's a British company so if it gets to big some one will buy arm LTD out lol.


Would be nice to see 3 players in the field once again and i am sure some of you remember the old ibm \ win chips and what happened to them.

 

[Shihab]

Wouldn't hold my breath for their chips. But competition is competition. My wallet's gonna loooove this~

 

[TheoneandonlyMrK]

an interesting aside might be the use of 64bit Arm cores in Nvidias project denver or its next gen more likely, a gpu with four arm 64bit cores for phones tabs and more mmm that might offer an APU to compete with AMDs

 

[FreedomEclipse]

It's a British company so if it gets to big some one will buy arm LTD out lol.

I get a feeling GOOGLE have had their eye on ARM for a while. If they bought ARM, that would give them a serious huge advantage in the mobile/pad/Ultra Portable market as well as anything beyond that as ARM are on a serious roll.

ARM will open so many doors for google if they were aquired by them. not just in the hardware sector, but in the software sector also. ARM+Andriod OS = serious business.

 

[xBruce88x]

hmm... to make up for the lack of raw power per core they could adopt the "module" stratagy that amd used... say 8 ARM cores per module and have the OS see each module as a "core" with each module having its own set of cache, etc. well maybe L3 shared among all the modules. And clock each module's core at 1.5ghz, heh you got some real power there. So, lets say a Quad module marketed as a Quad Core Arm cpu. or maybe even an 8 module for a total of 64cores haha.

edit: Oh, and they could work with nvidia to have a Tegra on steroids and slap that in the package, and that would be ARM's answer to the APU or intel's cpu's with integrated graphics.

 

[AsRock]

I get a feeling GOOGLE have had their eye on ARM for a while. If they bought ARM, that would give them a serious huge advantage in the mobile/pad/Ultra Portable market as well as anything beyond that as ARM are on a serious roll.

ARM will open so many doors for google if they were aquired by them. not just in the hardware sector, but in the software sector also. ARM+Andriod OS = serious business.

Not a google fan here, IF they were took over i be more for IBM\Samsung or even MS to take it over.. But better still no one and their stock holders and who ever else stick to their guns and keep it British..

 

[FordGT90Concept]

You might well be right, which would change the picture again, so we'll just have to wait and see - think the current fashion for "cloud" computing which basically puts a low powered terminal in your hand and gives all your data and control to some third party running big-iron servers. However, I don't think Microsoft really wants to have to develop and maintain two versions of the OS in the long run, as that really is expensive. It looks more like they want phones, tablets and PCs to all work the same way, hence the Metro interface on Windows 8. Therefore, I believe they will try to transition Windows to ARM. Microsoft perhaps aren't happy being stuck in the x86 world and want something better to support, I don't know. I can't see any other reason for starting support for an alternative processor with Windows 8.

It just struck me that what I described in paragraphs you bolded is essentially a tablet computer. Proprietary, mostly unserviceable, cheap, and effectively just an extended Windows Phone without the guaranteed chains of a "provider" (e.g. AT&T, Verizon, T-Mobile, Virgin Mobile, etc.). It fits the mold of what we know perfectly. In other words, the odds of a "desktop" version of Windows with an ARM processor are virtually none. They're all mobile, but more flexible (not limited to just Silverlight/WPF like Windows Phone 7 is). This is also why Microsoft doesn't care about backwards compatibility (read: virtually no chance of an emulator). All the applications offered for these ARM processors are likely to be coded either natively for the platform or on .NET which will presumably support it.

In short, ARM won't compete with the desktop variants of Intel and AMD processors. Instead, it will focus on AMD Geode, Intel Atom, and Via Nano products.

 

[qubit]

It just struck me that what I described in paragraphs you bolded is essentially a tablet computer. Proprietary, mostly unserviceable, cheap, and effectively just an extended Windows Phone without the guaranteed chains of a "provider" (e.g. AT&T, Verizon, T-Mobile, Virgin Mobile, etc.). It fits the mold of what we know perfectly. In other words, the odds of a "desktop" version of Windows with an ARM processor are virtually none. They're all mobile, but more flexible (not limited to just Silverlight/WPF like Windows Phone 7 is). This is also why Microsoft doesn't care about backwards compatibility (read: virtually no chance of an emulator). All the applications offered for these ARM processors are likely to be coded either natively for the platform or on .NET which will presumably support it.

In short, ARM won't compete with the desktop variants of Intel and AMD processors. Instead, it will focus on AMD Geode, Intel Atom, and Via Nano products.

Hmmm, depressing if true. It's certainly plausible.

 

[FreedomEclipse]

Not a google fan here, IF they were took over i be more for IBM\Samsung or even MS to take it over.. But better still no one and their stock holders and who ever else stick to their guns and keep it British..

It makes a lot of sense for google to aquire a growing company like ARM.

Google has Android OS and just recently aquired Motorola's 'Mobility' side of the business - thats the division that has EVERYTHING to do with Motorla mobiles and pads.

and that gives google even more power to take on the pad and somewhat the mobile handset market as most of the critical components can be sourced from 'in house' instead of from external sources which cuts overheads and puts them out of the line of fire from bullies like APPLE who will probably start finger pointing and screaming out 'patent infringement' as soon as google gets its first google made pad out onto the market.

Google can also SELL ARM processors which rakes them even more revenue in both sales and royalties.

It makes perfect sense. whether or not Google will take that step is another story but i wouldnt put it past them. with the way technology market and ARM has grown in the past few years. ARM must look very tasty to any tech company that has money to throw around.

IBM is a small possibility

Samsung is probably one of the companies that can stop google getting ARM. they have a huge market share and probably enough cash to throw around to counter any offer Google has depending on how desperate google are.

As for M$ - It doesnt make a lot of sense for them to aquire it - they have no use for it other then just holding onto the patents and licenses. They are a software company not hardware.

 

[Kreij]

AppliedMicro Revitalizes the Cloud with World's First 64-bit ARM Architecture Compliant Processor Converging Compute Performance with Power Efficiency

With architectural scalability up to 128 cores at 3GHz, X-Gene(TM) Server-on-a-Chip addresses cloud computing needs at less than half the power and cost of existing solutions

Source

Interesting.

 

[Fourstaff]

http://caxapa.ru/thumbs/229665/armcortexa8vsintelatomarchitecturalandbe.pdf

Arm's Whetstone and Linpack still fails against x86's

 

[qubit]

http://caxapa.ru/thumbs/229665/armcortexa8vsintelatomarchitecturalandbe.pdf

Arm's Whetstone and Linpack still fails against x86's

That's a 2009 article. A fairer comparison would be the latest models from both companies.

 

[Fourstaff]

That's a 2009 article. A fairer comparison would be the latest models from both companies.

Yeah, but its difference is still an order of magnitude, even if ARM manages to improve much more than x86 its at least quite a few years away, and another few years to scale them up. I couldn't find newer data, if people can then I think we need to re-extrapolate

 

[Kreij]

I think we need to re-extrapolate

That sounds dirty.

I'm not sure where these devices will successfully compete, but competition is always a good thing. It's still pretty early to make any kind of valid determination (although we do love our speculation here on TPU. ), but we should no more in the near future as new things hit the market using ARM's architecture.

 

[qubit]

Yeah, but its difference is still an order of magnitude, even if ARM manages to improve much more than x86 its at least quite a few years away, and another few years to scale them up. I couldn't find newer data, if people can then I think we need to re-extrapolate

Ok, if the difference is an order of magnitude, then it helps to make your point, to a degree. Still though, from what I can remember reading, the latest Cortex-A15 CPU is way faster than what was compared, the Cortex-A8, which will make quite a difference to the comparison. I'd really like to see a more up to date comparison, before coming to a definitive conclusion on this.

Regardless, if the kind of mojo that goes into an Intel x86 processor went into an ARM processor, then its performance would jump several orders of magnitude, no question. The way that the original 8088 from 1978 has been transformed into today's powerhouses is nothing short of phenomenal.

And Kreij is right, that does sound dirty.

 

[MilkyWay]

ARM dont manufacture anything they merely license the tech out to other companies, whats this about ARM using AMDs tech?

ARM back in the day was clock for clock faster than the Motorola CPUs of the time.

 

[darkreize]

Just a bit of a question here. Hypothetically speaking, if AMD is to sell itself/merge with another company, which company would have the money to acquire it, and what would be the benefits of that acquisition.

Just getting other people's opinion and doing some information gathering. Thanks. (^___^)

 

[qubit]

Just a bit of a question here. Hypothetically speaking, if AMD is to sell itself/merge with another company, which company would have the money to acquire it, and what would be the benefits of that acquisition.

Just getting other people's opinion and doing some information gathering. Thanks. (^___^)

Basically, that other company would acquire the rights to all of AMD's products except one: x86 and its derivatives. One of Intel's cleverest moves ever, was to put that exclusion clause in. AMD should never have agreed to that bit, if at all possible. If they had, we could have seen stronger competition for Intel now.

Welcome to TPU.

 

[suraswami]

If ARM really comes up with an awesome CPU, then most probably we will have 2 CPU companies and 2 Graphic card companies.