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

[Hard Ball]

A couple of decades from now, 2011 maybe remembered as one of the titanic shifts in the chip industry. Not only with the success of iPad, the meteoric rise of Android in mobile, the rise of a number of ARM manufacturers (which include nVidia), and MS porting their main client OS to ARM, but now comes another possible piece in the puzzle:

http://www.eetimes.com/electronics-news/4215518/ARM-working-on-AMD-to-drop-x86
http://www.eetimes.com/electronics-news/4215519/Analysis--Why-ARM-AMD-makes-sense?pageNumber=1

I don't think that AMD would abandon x86 wholesale for the next several years, for doing that leaves too many usefull assets on the table, but I think after the management team, they are finally showing the signs of waking up to the times. Heterogeneous chip with both x86 and ARM, plus some component from throughput design (GPU), along with the appropriate software stack, has been something that's been advocated by many for a while now, including myself. This would allow a smooth and less painful transition away from x86, and let each ISA stand on its own merits without the legacy of binary compatibility standing in the way.

While there isn't anything definitive yet at this point, and may not be for some months; if ARM and pull this off, it would be another significant move in the industry toward the eventual final confrontation between ARM and intel. At the risk of making it sound overly dramatic, these are exciting times. It will take some time to see everything where they may fall, but we should watch this closely.

 

[Dribble]

I can see AMD developing some soc's like nvidia has, but they are up against it as they started so late (well as we all know they really started very early then stupidly sold all that hard work off for peanuts). To develop an ARM soc the one bit you don't need is a cpu (ARM provide perfectly good cpu cores) so AMD's knowledge is a bit wasted there. However they could obviously provide a good gpu, but there's more to a soc then that, e.g. nvidia years ago was out buying companies such as good audio chip designers (the ones that developed the chip used in the iPod). Then there's the software support - AMD are used to piggy backing on Intel/MS there, but if you want your own soc with your own graphics you need a good software team. For example I bet a major reason google used tegra 2 for andriod 3 was nvidia was able to offer very good software support with all that TWIMTBP and cuda experience.

It'll be an uphill battle for a few years as they catch up but they've got to go for it imo - staying under Intel's thumb in the slowly shrinking market of x86 is not a good long term strategy.

 

[Genx87]

I'd think including ARM and x86 on the same processor would be wasteful and require more energy to pull off. If AMD is to get into this space. They need to drop x86 all together. The overhead with x86 will hurt them against ARM. They can attempt to push fusion into this space. But I think it just wont compete with ARM on power consumption.

 

[Hard Ball]

I'd think including ARM and x86 on the same processor would be wasteful and require more energy to pull off. If AMD is to get into this space. They need to drop x86 all together. The overhead with x86 will hurt them against ARM. They can attempt to push fusion into this space. But I think it just wont compete with ARM on power consumption.

All the system level code would run on the ARM cores, x86 would only be necessary for specific applications that are compiled in x86 binary. I don't think power consumption would be an issue, but yield and cost might be. So MCM or off package co-proc design would be another possibility, which would mitigate the transistor budget/yield side of the equation, cost would be measured in terms of how important binary application compatibility would be to the client.

 

[tweakboy]

x86 will never die. All big companies and what not use Windows XP and custom applications which are x86..... the day x86 dies is not anytime soon. Im talking 6 years or more... imo

 

[VirtualLarry]

All the system level code would run on the ARM cores, x86 would only be necessary for specific applications that are compiled in x86 binary. I don't think power consumption would be an issue, but yield and cost might be. So MCM or off package co-proc design would be another possibility, which would mitigate the transistor budget/yield side of the equation, cost would be measured in terms of how important binary application compatibility would be to the client.

What about an ARM quad-core, with an x86 hardware decoder stage, that translated into ARM opcodes, or whatnot? Kind of like Itanium has or had.

 

[busydude]

I can see AMD developing some soc's like nvidia has, but they are up against it as they started so late (well as we all know they really started very early then stupidly sold all that hard work off for peanuts).

I don't think they were stupid at all.. their x86 business was at risk of going bankrupt. They had to make AMD leaner.. and concentrate on their core x86 market. It was just part of their business plan of staying alive in the x86 space.

 

[Genx87]

I don't think they were stupid at all.. their x86 business was at risk of going bankrupt. They had to make AMD leaner.. and concentrate on their core x86 market. It was just part of their business plan of staying alive in the x86 space.

For the cash they recieved it was pretty stupid. 65 million is peanuts. And now they are on the outside looking in on a market that is exploding.

 

[tweakboy]

Even Windows 8 I learned has x86 code in it. But also robust x64 compat,, thx,

 

[gevorg]

IMHO, x86 has at least 10 years to live, possibly more. Beyond that, it will die sooner or later, since it carries too much baggage from early computing days.

 

[tweakboy]

IMHO, x86 has at least 10 years to live, possibly more. Beyond that, it will die sooner or later, since it carries too much baggage from early computing days.

I believe that!

 

[ElFenix]

What about an ARM quad-core, with an x86 hardware decoder stage, that translated into ARM opcodes, or whatnot? Kind of like Itanium has or had.

you've basically described every new x86 core since the pentium pro. the real problem is x86's 8 registers. iirc, x64 basically solved that, though.




intel still owns the rights to alpha, does it not?

 

[smakme7757]

I don't care if it's an ARM processor or an X86 processor or a super duper quantum Ghz ass kicker. As long as it gets the job done it just doesn't matter to me at all!

Of course new technology is exiting and anything new and exiting is something i look forward too. I also read a while back that IBM is onto something although the name escapes me. Apparently it's a silicon replacement which might keep x86 in the game a little longer, but who knows.

 

[aigomorla]

lolol...

i'll be waiting for my bluetooth HUD sunglasses which pair up with my phone b4 i wait for the end of x86.

 

[Tsavo]

I'm pretty confident that x86 will be around a LOT longer than I will be.

 

[LiuKangBakinPie]

Intel by the rate their going is shrinking themselves into ARM production. The competition must be pissin themselves

 

[nenforcer]

x86 may be around for still longer than 10 years but the difference is you will have alternatives.

Now you could argue that even though Microsoft had an DEC ALPHA version of Windows that x86 eventually killed it off but the problem was the sales of those workstations wasn't that great.

When you have millions of cheap, low powered ARM devices which will soon be running Windows you will have a definite alternative instead of obscure, high cost workstations.

 

[RobertPters77]

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.

 

[ViRGE]

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.

Agreed. We may see some kind of equilibrium where x86 and ARM co-exist in the same market, but nothing is killing x86. The ISA is nothing but a formality, what really matters is who's designing and building chips.

 

[gdansk]

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?

 

[Idontcare]

As a member of the "x86 generation", I too share the personal views stated by many others that "x86 will never die, evar!" but I am required to note that there were a great many number of architectures that preceded x86 and that when x86 came on the "scene" it was not exactly highly regarded or viewed as a threat to the existing big-iron architectures of the day.

With this in mind, I have to conclude it would be entirely foolhardy of me to assume/presume/state/expect/anticipate that x86's future is somehow any more or less secure than that of the architectures it displaced in the past 40 yrs.

Just like our lives, x86 has an expiration date, we just don't know what it is or why it will come about. Could be ARM, could be something entirely different. But to be of the position that x86 is somehow special, somehow unique, compared to the historical progression of computing architectures that have been spawned for over a century's time, that just strikes me as laughably short-sighted.

I personally can't imagine needing anything beyond x86, but then again if I were a child of the typewriter age, predating electronic calculators even, then I probably would have been prone to regard the newfangled invention of the personal computer as merely being a fad that would blow over in time.

 

[sxr7171]

It seems to me, with what little I know, that the end game is performance/watt. Battery technology is progressing at a rate that is an order of magnitude or 2 slower than semiconductor technology. People want machines that do everything they need and weigh a pound or less and last 12 hours. Can x86 reach that kind of performance or will it take ARM to do it? But there will always be a market for plugged in high performance machines and x86 will do just fine there, the issue is that that market will one day become very small. The vast majority of people don't need that kind of power. The main demands today are photo/video manipulation, transcoding and encryption. If ARM CPUs with the right specialized co-processors/GPUs can handle these things well, then there won't be a need for x86. But it may well be that x86 develops in a way that it becomes more specialized for what the market needs. But I somehow doubt that it will meet the performance/watt figures of ARM. But who knows?

 

[videoclone]

Its already found a replacement and its called X64 .. give it enough time and we wont have x86 it will be x64 that we all see in our super fast servers and workstations

 

[Fox5]

I can't see AMD going ARM, they don't have the design resources to do two-three x86 designs plus ARM.
Besides, right now they're #2 in the large x86 market, they'll be lucky to be a player in the ARM market. They don't have any wireless tech, they don't have any other integrated mobile tech that the market demands (sold that all to qualcomm), if they just license the A15 they're doing nothing special compared to everyone else, and they have several powerful graphics makers to compete against in the ARM market. Not the least of which is their own IP via Qualcomm.

 

[soccerballtux]

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

It could happen.