Intel Buys an ARMy. Maybe?

[PorscheDan]

Is Intel in trouble? Since it is the #1 semiconductor company and, shipping 22nm in Q4 this year with 14nm in 2013, it is two process generations ahead of everyone else it is hard to see why it would be:


http://www.semiwiki.com/forum/content/509-intel-buys-army.html




DanielNenni, you have been told on numerous occasions that self-promotion in the interest of driving traffic off-site to your affiliate web-sites is not acceptable.

Making second-accounts (DanielNenni = PorscheDan) in an effort to get around your previous warnings, in the pursuit of further spamming the forum with your affiliate links, is all the more unacceptable.

We've given you ample rope in the past to come to terms with the anti self-promotion guidelines and you have taken actions that force us to treat your account(s) the same as we would any other spammer account. Good bye.

Moderator Staff at the AnandTech Forums

 

[Cogman]

I've never thought intel would be in trouble due to the ARM shift. They may take a hit, but I doubt it will be the end of the world for them.

They are able to buy ARM licenses just like everyone else.

Also, the article is pure speculation (it even admits to it at the end).

 

[Edrick]

I do not really understand all this hype about ARM processors. Sure they are great processors for phones and tablets, but are not even in the same class with x86.

And people are getting all excited about running Win8 on ARM processors. Anyone who has run Win7 on an Atom CPU knows that it is just not the same as a full power CPU.

So what is the point? Netbooks? They were a fad that are being replaced by tablets. I see tablets as also being a fad which will last only a few years until the next best thing comes out. I doubt anyone would use a ARM processor on a HTPC when IB will run cool and fast.

Perhaps I am missing something, but I just do not see ARM as a viable option for Windows or even any type of desktop system. When tablets can do almost anything a desktop can do, then I may think otherwise. But we are a long way from that.

 

[ElFenix]

I do not really understand all this hype about ARM processors. Sure they are great processors for phones and tablets, but are not even in the same class with x86.

And people are getting all excited about running Win8 on ARM processors. Anyone who has run Win7 on an Atom CPU knows that it is just not the same as a full power CPU.

So what is the point? Netbooks? They were a fad that are being replaced by tablets. I see tablets as also being a fad which will last only a few years until the next best thing comes out. I doubt anyone would use a ARM processor on a HTPC when IB will run cool and fast.

Perhaps I am missing something, but I just do not see ARM as a viable option for Windows or even any type of desktop system. When tablets can do almost anything a desktop can do, then I may think otherwise. But we are a long way from that.

it's going to be a long time before you're recoding video on your ARM whatnot, but for the 95% of crap that most people do (reading and watching video), ARM is just fine.

it's like a cell phone camera. no one is using it to shoot their wedding photos but for the vast majority of people for the vast majority of time it is good enough. that's really where the sweet spot is: good enough.

so, you'll still probably have some x86 hardware in your house. but most gaming is already done on something other than x86, most internet use will be done on something other than x86, etc.

 

[Cogman]

I do not really understand all this hype about ARM processors. Sure they are great processors for phones and tablets, but are not even in the same class with x86.

And people are getting all excited about running Win8 on ARM processors. Anyone who has run Win7 on an Atom CPU knows that it is just not the same as a full power CPU.

So what is the point? Netbooks? They were a fad that are being replaced by tablets. I see tablets as also being a fad which will last only a few years until the next best thing comes out. I doubt anyone would use a ARM processor on a HTPC when IB will run cool and fast.

Perhaps I am missing something, but I just do not see ARM as a viable option for Windows or even any type of desktop system. When tablets can do almost anything a desktop can do, then I may think otherwise. But we are a long way from that.

The thing about ARM, and where it gets its hype, is the fact that x86 doesn't even come close to it when it comes to power consumption. Atom, one of the leanest in the x86 domain, weighs in at 1 Watt at the lightest. ARM, on the other hand, can go clear down to about 0.136 Watts for a similarly clocked processor that can have a similar throughput to the Atom. That is impressive.

When you have such large power savings for the very small, it means that the very large can add more and more features that were previously unavailable due to power constraints. In other words, the ARM instruction set has a lot of room to grow.

The point is being able to do a heck of a lot more for less power.

 

[IntelEnthusiast]

Is Intel in trouble?

http://www.engadget.com/2011/04/19/intel-shocks-everyone-including-itself-with-record-q1-earnings/

I would have to say, No.

 

[ElFenix]

http://www.engadget.com/2011/04/19/intel-shocks-everyone-including-itself-with-record-q1-earnings/

I would have to say, No.

that, right now, is the biggest problem for ARM makers. heck, even AMD's net income dwarfed ARM's for the last reporting quarter.

 

[Edrick]

The thing about ARM, and where it gets its hype, is the fact that x86 doesn't even come close to it when it comes to power consumption. Atom, one of the leanest in the x86 domain, weighs in at 1 Watt at the lightest. ARM, on the other hand, can go clear down to about 0.136 Watts for a similarly clocked processor that can have a similar throughput to the Atom. That is impressive.

But as soon as you add in all the instruction sets that x86 already has, along with the other 'features' not currently found in ARM, then the power will increase.

But lets assume that ARM could build a viable competitor to the x86 chips in the next 5-10 years. And lets also assume they could do it with less power, say 20% less. Now companies would have to spend millions to re-optimize their code for the ARM uArch as opposed to x86. Unless ARM includes all the instruction sets that AMD/Intel does. Then I doubt we would be looking at 20% power savings.

So like I said before, for tablets and phones, sure it makes sence. Other than that, I do not see it.

 

[Cogman]

But as soon as you add in all the instruction sets that x86 already has, along with the other 'features' not currently found in ARM, then the power will increase.

ARM already has licenses for SIMD instructions, the most important addition to x86. The "extra" instructions of the x86 architecture aren't necessarily a feature, rather, a lot of added bloat. The x86 architecture has tons of instructions like "enter" that are no longer used in practice because they are too slow. These extra functions add a fair amount of overhead and power consumption increases.

Couple that with the fact that the x86 architecture has to support 4, 8, 16, 32, and now 64 bit general registers. It can't trim the fat because it still has to support MSDOS (This is, in fact, one of the tests that Intel does to make sure a processor is good, booting up MSDos).

Moving to a new architecture means you don't have to carry with you the rest of the bloat.

But lets assume that ARM could build a viable competitor to the x86 chips in the next 5-10 years. And lets also assume they could do it with less power, say 20% less. Now companies would have to spend millions to re-optimize their code for the ARM uArch as opposed to x86. Unless ARM includes all the instruction sets that AMD/Intel does. Then I doubt we would be looking at 20% power savings.

Don't get me wrong, x86 is going to be around for a LONG time. Legacy makes technology maintain unnatural lifespans. That being said, the only code that needs to be "optimized" for ARM is assembly. There is very little of that going on in consumer products. Compilers such as the GCC have been long optimized to produce some pretty good code for ARM (its been around for long enough).

ARM doesn't need to include every single instruction that AMD and Intel has to be competitive. (WTF would they need with MMX or 3dNow support?).

The power savings, however, is besides the point. My main point is that ARM has a much larger thermal range to play with compared to x86. Because it is lower power now means that it has much more room to increase speed later. It can do things that x86 can't do because it can spare the power.

So like I said before, for tablets and phones, sure it makes sence. Other than that, I do not see it.

Tablets and phones may very well be the future of computing. Even then, never say never. If a company like DEC can go under, anything can happen in the semi-conductor business.

I certainly can see ARM taking hold and generally replacing x86. However, that doesn't mean that x86 is going to die all together, it just means that we will probably start to see ARM netbooks, notebooks, and desktops, then those might start outselling their x86 counter parts, and later we might see it expand into server markets. Or it may just stay in the low power sector. Who knows? It certainly isn't impossible.

 

[Cogman]

that, right now, is the biggest problem for ARM makers. heck, even AMD's net income dwarfed ARM's for the last reporting quarter.

And I'll say this, there is nothing that is preventing Intel and AMD from being ARM makers.

 

[Dahak]

would be kinda funny if they did, they had an arm license back when they had the xscale chips before selling it

 

[Cerb]

I do not really understand all this hype about ARM processors. Sure they are great processors for phones and tablets, but are not even in the same class with x86.

And people are getting all excited about running Win8 on ARM processors. Anyone who has run Win7 on an Atom CPU knows that it is just not the same as a full power CPU.

No, but I can't imagine layin' back on the couch with a full powered notebook computer, and for light use, it's plenty fast enough.

So what is the point?

A decent MS OS, for those people in vertically integrated environments. Windows is only big and heavy because the hardware and users allow for that. With modern smartphone HW being what it is, there's no reason a phone could not effectively run a variant of the real Windows OS, instead of Windows CE.

 

[OCGuy]

Something tells me that nobody at Intel is boxing up their desk any time soon.

 

[ElFenix]

And I'll say this, there is nothing that is preventing Intel and AMD from being ARM makers.

well, if ARM decided to not license to them...

intel back in the day experimented with a lot of things other than x86. problem is the market inertia of x86 kept intel's risc designs from exceeding in anything other than the embedded market. and when choosing to spend capital between their wildly profitable desktop and server processor lines, and their embedded line, they chose the profitable side rather than the embedded side. ARM's desktop side died a couple decades ago, so all they had was the embedded side.

x86's main problem for power scaling, afaik, is that it has to have a decoder to turn complex x86 instructions into several internal micro-ops. when you're talking about 10 to 100 watts (or more) that decoder takes up a minuscule amount of the entire microprocessor's power consumption. when you're talking half a watt or less, that decoder is a significant power drain. the decoder has the same effect on die sizes.

 

[HopJokey]

Intel has a huge advantage with a big moat and that is in manufacturing. Right now they have a huge lead in process tech and it seems to only get bigger as time moves on.

 

[pcunite]

If Visual Studio will automatically turn standard C++ into insert_fav_cpu_cmd_here then anything is possible. It is all about how to sheild application developers from CPU changes.

 

[Edrick]

ARM already has licenses for SIMD instructions, the most important addition to x86. The "extra" instructions of the x86 architecture aren't necessarily a feature, rather, a lot of added bloat. The x86 architecture has tons of instructions like "enter" that are no longer used in practice because they are too slow. These extra functions add a fair amount of overhead and power consumption increases.

Couple that with the fact that the x86 architecture has to support 4, 8, 16, 32, and now 64 bit general registers. It can't trim the fat because it still has to support MSDOS (This is, in fact, one of the tests that Intel does to make sure a processor is good, booting up MSDos).

Moving to a new architecture means you don't have to carry with you the rest of the bloat.

Ok, so using that train of thought (which I do understand and agree with), why couldn't Intel or AMD create a new family of x86 CPUs which do not maintain 'legacy' functions. They would be much leaner and efficient. It would not be much different than moving to a new architecture. And all the current x86 developers would still have jobs.

 

[ElFenix]

Ok, so using that train of thought (which I do understand and agree with), why couldn't Intel or AMD create a new family of x86 CPUs which do not maintain 'legacy' functions. They would be much leaner and efficient. It would not be much different than moving to a new architecture. And all the current x86 developers would still have jobs.

it would be moving to a new arch. apple has moved mac's arch twice so far. it could be done but with intel's lead at manufacturing and total engineering resources, in the 10 to 100 watt space there's really no need to move to a new arch. risc designs with similar overall performance to intel's designs have usually consumed about as much power so on the desktop side there's really no reason for intel to change arch.

 

[Cogman]

Ok, so using that train of thought (which I do understand and agree with), why couldn't Intel or AMD create a new family of x86 CPUs which do not maintain 'legacy' functions. They would be much leaner and efficient. It would not be much different than moving to a new architecture. And all the current x86 developers would still have jobs.

They could totally do that, however, ARM is an already available and tested architecture. That is pretty much the only reason why they would choose to go with ARM vs cooking their own new architecture.

The x86 developers/engineers wouldn't really be out of jobs because the tricks that apply to optimizing x86 will almost directly translate to optimizing ARM. It wouldn't take too much retraining to get them going.

 

[hans007]

If Visual Studio will automatically turn standard C++ into insert_fav_cpu_cmd_here then anything is possible. It is all about how to sheild application developers from CPU changes.

microsoft will surely pu tthat in.

studio 6 had it for MIPs i believe, because windows NT 4 had alpha and mips and itanium support.

 

[Voo]

Now companies would have to spend millions to re-optimize their code for the ARM uArch as opposed to x86. Unless ARM includes all the instruction sets that AMD/Intel does.

Modern optimizing compilers use only an extremely small subset of the complete x86 ISA anyways and the important part: They mostly optimize not on the machine code representation but some intermediary representation (SSA and so on).
Other than that, the vast majority applications aren't hand optimized for x86, so we're talking about recompiling the code and playing with some compiler flags - that won't hardly cost millions.

And for those apps that DO include handwritten assembly code, the majority of those uses them quite localized in the few innermost loops where that really brings an advantage, which shouldn't be especially hard.

The rest are mostly things like drivers (well and the CPU dependent OS kernel parts) and that's surely a valid concern. Recompiling your code for ARM if there's a growing market is one thing, but who's going to offer drivers not just for x86 but also ARM? We all remember the x64 growing pains in that regard.

 

[Dadofamunky]

Based on the last reported results, I'd... uh... have to say no.

 

[Vesku]

Considering they sold off their XScale ARM processor product segment several years ago, I'd say it's pretty unlikely. Intel doesn't like the profit margins in the ARM cpu manufacturing sector, simple as that.

 

[Cogman]

Considering they sold off their XScale ARM processor product segment several years ago, I'd say it's pretty unlikely. Intel doesn't like the profit margins in the ARM cpu manufacturing sector, simple as that.

I know one of the engineers that worked on the XScale processor. According to him, they just made way too many mistakes. They launched late, they had several design flaws (which resulted in multiple prototype fab runs), and they just did things wrong. According to him, nobody was really pleased with the XScale situation. It was costing them too much money and headaches.