So here it goes: possible beginning of the end for x86

[ElFenix]

Personally, I do not understand what is wrong with x86/64 in many environments. It does not compete on the very low power end (yet) but it seems to do fine for desktops and low power (like notebooks). What am I missing?

ARM designs could easily be good enough for desktop use (there is really nothing stopping them from designing processors on the performance level of the latest core whatever) and now you're getting an installed base of end user software (rather than ATMs and ovens) and experienced developers.


the other thing is that with lightweight operating systems that don't have all the baggage windows carries around you don't need as much processor power to have the same end-user experience. windows on ARM sounds like a disaster to me. but who knows? iOS is a stripped down version of OSX so maybe windows can do the same.

 

[Zap]

transition away from x86

IMO it isn't different architectures that may or may not drive the transition away from x86. It is the ability to do more and more tasks through the network/internet.

 

[Joseph F]

It would be very weird for me to see a ~125W desktop ARM CPU. It just doesn't sound right when all you ever see them in is portable devices. (As the main CPU) But, if a superior processor than today's is developed that is based upon ARM arch I'd welcome it with open arms.

 

[OS]

X86 won't die. They said the same thing about PowerPC killing x86 in the mid 90's.

These "End of x86 Prophets" have no more credit than those 2012 doom sayers.

seriously you can tell who has really been a long time follower of the computer industry from this topic.

Back then the articles claimed that Mac/PowerPC and DEC alpha architecture had superior performance to anything x86, that the x86 architecture and instruction set were inherently handicapped and the x86 days were numbered.

Well we know now intel blew by PowerPC, Apple abandoned PPC and switched to Intel, despite their massive ad campaign to convince the world that Intel was inferior to IBM/PPC and x86 was a dead end.

AMD later acquired Alpha technologies and incorporated them into their x86 product making it faster.

I guess it's been almost 20 years since those ongoing prognostications about x86 being a dead end architecture but now x86 performance relative to other architectures is even stronger now than back then.

you know what they say about predicting the future, but x86 has been counted out many times and somehow intel/amd manage to continue squeezing out more performance and flexibility.

 

[Idontcare]

all previous architectures failed because Windows did not work on them. Now with ARM, that is all different.

It could happen.

Windows NT

A main design goal of NT was hardware and software portability. Various versions of NT family operating systems have been released for a variety of processor architectures, initially Intel IA-32, MIPS R3000/R4000, and Alpha, with PowerPC, Itanium, AMD64 and ARM supported in later releases. The idea was to have a common code base with a custom Hardware Abstraction Layer (HAL) for each platform.

 

[Kingkazma]

i doubt it's going to die, you know why? because of intel

intel's going to 22nm right now while everyone's trying to get to 28nm. intel's always ahead of everyone when it comes to the cmos process. and as long as it has that advantage, i have a hard time seeing arm or anything else destroying x86

i only see arm destroying x86 if arm can get MORE performance with LESS transistors, if it just matches, intel can just increase transistor count to deal with the problem.

 

[podspi]

i doubt it's going to die, you know why? because of intel

intel's going to 22nm right now while everyone's trying to get to 28nm. intel's always ahead of everyone when it comes to the cmos process. and as long as it has that advantage, i have a hard time seeing arm or anything else destroying x86

i only see arm destroying x86 if arm can get MORE performance with LESS transistors, if it just matches, intel can just increase transistor count to deal with the problem.

Yea, Intel's advantage in manufacturing and obsession with x86 guarantees (imho) its existence for a very long time.

 

[soccerballtux]

Windows NT

that was back when nobody used windows NT. That's what the N stood for. Nobody's Technology.

 

[darkswordsman17]

I think AMD using ARM makes a lot of sense, as I don't see them being able to compete with Intel on shrinking x86 to smaller sizes (at least for the next few years), and AMD helping ARM with graphics will be a benefit to them. Nvidia seems to have taken a lone wolf approach and doing their own thing, so AMD would be a good match.

Also remember that just because Microsoft is putting Windows on ARM doesn't mean it'll be worth using that way. I think Microsoft is changing, but I don't see Windows and ARM being a great match for a few years (look at netbooks). Also, Intel can get a lot more aggressive with x86. I could see them showing off a 22nm updated Atom SoC in 2012, and there's a good chance it would be more powerful and possibly competitive power wise with high end ARM SoCs. Plus Intel is working to not be dependent on Windows, which will help their power usage in the mobile space.

There's a lot of big changes that we'll see in the next 5 years, it'll be interesting to see how things play out.

 

[Edrick]

Yea, Intel's advantage in manufacturing and obsession with x86 guarantees (imho) its existence for a very long time.

Intel has on at least 2 different occasions tried to build a cpu better (different) than x86. The first one was good in theory, but failed in the real world. The second (Itanium) appears to have its days numbered.

With that in mind, I do not see Intel moving away from x86 anytime soon.

 

[Edrick]

that was back when nobody used windows NT. That's what the N stood for. Nobody's Technology.

Maybe you didn't. The rest of us may have.

 

[catilley1092]

At some point, and sooner than many thinks, 128 bit computing will surface. It's not a question of if, but when. When, will be most likely shortly after XP is pulled from life support on 04/08/2014.

When Windows 7 was released, all that we heard on the forums was "should I stick with 32 bit, or go with 64?". Or the other way around.

There's no way that MS will support 3 different platforms here, which is why 128 bit won't be released with Windows 8. It will happen though with Windows 9 (or whatever MS decides to name it). Then it'll be a x64 (main version) & 128 bit (like 64 bit was for years, some strange, weird technology).

It's coming, believe me, it really is. 128 bit will be saturating the market, just as 64 bit does now. Where will that leave x86?

Cat

 

[Edrick]

128-bit processors could become prevalent as a method of addressing over 2^64 bits of information. Up to 2^128 could be directly addressed with 128 bits. That amount greatly exceeds the total data stored on Earth today (2010), which has been estimated to be around 1.2 zettabytes (over 2^70 bytes).

Yea, its coming........

I woundn't hold your breath however.

 

[Dribble]

Yea, its coming........

I woundn't hold your breath however.

Well I want all the information on the planet on my phone so if that's 2^70 bytes I'm obviously going to need a 128 bit phone!

 

[Vic Vega]

Some of you as old as me may have read Popular Mechanics as a kid in the 1980s, some of you probably read it in the 60s or 70s. Remember those old issues where some amazing new technology comes along and replaces what we have currently? In most cases, 30-40 years later, it hasn't happened.

This thread is like one of those old Popular Mechanics issues.

Yes, mobile platforms are increasing in use, power and popularity. Is x86 going anywhere? Don't count on it. Are traditional desktops and laptops going anywhere? Don't count on it. It still very hard to do actual work on these mobile devices and the people who have them know.

Not to mention the majority of the IT industry, you know, the back end, the mega networks, what keeps all of this running, is mostly x86. Consumer forget about this part or never knew about it to begin with. They think their devices run on magic. No, they run on servers. Most of the x86 variety. Why? They're cheap to build and offer good performance. Sure, lots of IT back end is on other platforms as well. Yes, mainframes and mid-range systems (non-x86) are still used in great numbers but 80% or so of back end IT is x86 from my experience.

Specifically, I work as a SAN administrator. This is sort of a unique position in IT where I get to interact with basically every type of platform and OS because everyone needs storage. Some of these monster Oracle servers I support, monster MS-SQL servers, massive Exchange clusters, etc, etc. They aren't going to run on a mobile platform and provide the level of performance needed for the service. x86 is here to stay.

 

[VirtualLarry]

It's coming, believe me, it really is. 128 bit will be saturating the market

How do you define a 128-bit processor? Today's CPUs process information in 128, and now 256 (Sandy Bridge and Bulldozer) bit chunks. Neither one has 128-bit address registers, which are pretty-much overkill.

 

[Idontcare]

Some of you as old as me may have read Popular Mechanics as a kid in the 1980s, some of you probably read it in the 60s or 70s. Remember those old issues where some amazing new technology comes along and replaces what we have currently? In most cases, 30-40 years later, it hasn't happened.

This thread is like one of those old Popular Mechanics issues.

Yes, mobile platforms are increasing in use, power and popularity. Is x86 going anywhere? Don't count on it. Are traditional desktops and laptops going anywhere? Don't count on it. It still very hard to do actual work on these mobile devices and the people who have them know.

Not to mention the majority of the IT industry, you know, the back end, the mega networks, what keeps all of this running, is mostly x86. Consumer forget about this part or never knew about it to begin with. They think their devices run on magic. No, they run on servers. Most of the x86 variety. Why? They're cheap to build and offer good performance. Sure, lots of IT back end is on other platforms as well. Yes, mainframes and mid-range systems (non-x86) are still used in great numbers but 80% or so of back end IT is x86 from my experience.

Specifically, I work as a SAN administrator. This is sort of a unique position in IT where I get to interact with basically every type of platform and OS because everyone needs storage. Some of these monster Oracle servers I support, monster MS-SQL servers, massive Exchange clusters, etc, etc. They aren't going to run on a mobile platform and provide the level of performance needed for the service. x86 is here to stay.

I hear that in the future electricity will be so cheap they won't even bother metering it!

 

[BBMW]

Other than power consumption, what advantage does ARM have over x86/64? I haven't heard of any. Most of the server world is going x86 from RISC.

I think they have their own markets, and will likely stay that way. X86 will continue to dominate the notebook/desktop/server market, and ARM will dominate the "mobile device" market.

 

[Tuna-Fish]

Other than power consumption, what advantage does ARM have over x86/64? I haven't heard of any.

Basically none. But the power difference is significant -- basically, in x86 you can make all the problems of the instruction set go away by paying a few tens of millions transistors and a few hundred milliwatts of power in the decode hardware. For a 95W desktop processor or a 30W laptop processor, this just isn't enough to care about. For a 500mW smartphone processor, it's quite significant.

If you wanted to be crazy, you could make a lot wider processor for ARM much cheaper -- arm decoders are orders of magnitude simpler than the x86 equivalents, and instructions that are always naturally aligned mean that there is no extra cost for having wider decode. But is there really a market for a processor that's wider than 4-issue?

A lot of people have been talking about "hybrid" ARM+x86. That's not so far-fetched, because modern x86 processors already decode instructions into risc micro-ops -- adding a separate set of arm decoders in front of the execution engine of, say, bobcat, would probably cost only fractions of a percent of die area. The question absolutely isn't of feasibility, but of market. Who would want such a processor?

 

[tweakboy]

Think of freeway lanes. 64 bit has 4 lanes and can pass RAM quicker, not noticiably tho.
x86 has 2 lanes to operate on.

 

[Idontcare]

That was worth a ROFL or two

 

[Dribble]

Other than power consumption, what advantage does ARM have over x86/64? I haven't heard of any. Most of the server world is going x86 from RISC.

From a business standpoint flexibility of design. With x86 you have to buy Intel/AMD cpu's in Intel/AMD boards configured how Intel/AMD have decided they should be. ARM don't force that one size fits all approach. Any company can develop whatever design they want with very few restrictions. Licensing ARM is also really cheap so the end result is something that is closer to what customer X really wants for a lower price.

For example google servers - instead of using general purpose high power x86 cores they'd be much more efficient with a massively parallel cpu with instructions just to do server-type-stuff. ARM is a good basis for this - it gives you some low power general purpose cores that you can incorporate into your own design to which you add your server specific cores (running something like open cl). This would probably be a good order of magnitude more efficient then x86 meaning the end of it.

Obviously you could do something like this with x86, only x86 is not as simple or efficient a starting point as ARM, and Intel/AMD don't let anyone else near their designs or let anyone else make something using an x86, and they'd want to charge a fortune if they did (unlike ARM who charge peanuts).

 

[Martimus]

Think of freeway lanes. 64 bit has 4 lanes and can pass RAM quicker, not noticiably tho.
x86 has 2 lanes to operate on.

Why do you continuously make definitive statements about things you don't know? At the very least throw in a caveat that you aren't sure about what you are saying, but that you believe what you are saying to be true. As it is, you often make completely untrue statements, and may end up confusing anyone who reads your posts and doesn't know any better.

 

[acx]

How do you define a 128-bit processor? Today's CPUs process information in 128, and now 256 (Sandy Bridge and Bulldozer) bit chunks. Neither one has 128-bit address registers, which are pretty-much overkill.

Same way we define a 64 bit processer and 32 bit processor. The processor is capable of performing 128 bit wide math operations and have 128 bit wide registers.

 

[Edrick]

How do you define a 128-bit processor? Today's CPUs process information in 128, and now 256 (Sandy Bridge and Bulldozer) bit chunks. Neither one has 128-bit address registers, which are pretty-much overkill.

128-bit address registers would be my definition. And yes, overkill for the next 10+ years at least.