AMD: Moore's Law's end is near

[R0H1T]

Theoretical physicist Michio Kaku believes Moore's Law has about 10 years of life left before ever-shrinking transistor sizes smack up against limitations imposed by the laws of thermodynamics and quantum physics. That day of reckoning for the computing industry may still be a few years away, but signs of the coming Moorepocalypse are already here. Just ask chip maker AMD.

The company's Chief Product Architect John Gustafson believes AMD’s difficulties in transitioning from 28-nanometer chips to 20-nanometer silicon shows we’ve reached the beginning of the end.

"You can see how Moore's Law is slowing down," Gustafson recently told the Inquirer. "We've been waiting for that transition from 28nm to 20nm to happen and it's taking longer than Moore's Law would have predicted...I'm saying you are seeing the beginning of the end of Moore's law." A processor’s nanometer measurement tells you the size of the smallest transistors on a given chip.

Doomsday predictions about the end of Moore's Law are nearly as old as the famous observation posited by Intel cofounder Gordon Moore in 1965. In his 2011 book Physics of the Future, for example, Kaku predicted the end of Moore’s Law could turn Silicon Valley into a “rust belt” if a replacement technology for silicon isn’t found.

Not just about tech

Moore's Law, Gustafson argues, wasn't just about the technological ability to put more transistors on a chip, but also the economic feasibility of doing so. "The original statement of Moore's Law is the number of transistors that is more economical to produce will double every two years," Gustafson said. "It has become warped into all these other forms but that is what he originally said."

Gustafson's comments echo similar statements another AMD representative recently told PCWorld. "Moore’s Law was always about the cost of the transistors as much as performance increasing as you could afford more and more of them," Gary Silcott, the senior PR manager for AMD's APU and CPU products, said.

AMD's argument may also reveal a bit of corporate bias related to the company's recent struggles. While AMD's chips are currently stuck at 28nm, Intel is pushing ahead with smaller and smaller designs. Intel currently produces 22nm chips for its latest generation of Core processors, Ivy Bridge. The next generation, Haswell, will also feature a 22nm process. Intel, in 2014, expects to produce 14nm Haswell chips, and the company is aiming to produce 10nm chips by 2016.

But AMD is not alone in seeing the beginning of the end for Moore's famous observation. Computer processors in general are starting to lag behind the intent, if not the letter, of Moore’s Law, as PCWorld Senior Writer Brad Chacos recently reported. Chips may be getting smaller, but huge gains in processing power aren’t making the same jumps over that time as we saw in previous decades. Instead, smaller chips are more about improving graphics and energy efficiency rather than raw performance.

So if we are seeing the end of Moore’s Law, what’s next for computers? Kaku suggested a few interesting possibilities such as molecular transistors or, much further down the road, quantum computers. Until then, Intel, AMD, and other chip makers will continue to squeeze every last ounce of speed and power they can from silicon designs.

Source

 

[R0H1T]

Just to add a few things here ~ I don't see Intel going forward with their next full node shrink(10nm) unless they make serious inroads into the mobile/tablet market or better still they produce large volumes of chips for other fabless chipmakers like AMD

So if neither of these conditions are fulfilled it'll be doubly hard for them to pursue their tick-tock strategy any further than ~2016 ! Not saying that they won't go to 10nm rather their next full node shrink will be delayed by atleast a year or two :hmm:

 

[Atreidin]

...chip makers will continue to squeeze every last ounce of speed and power they can from silicon designs.

Ah yes, I was just talking to a colleague about how we aren't satisfied with the half ounce of speed that Haswell has over Ivy Bridge, even if we save a couple gallons of power every year on our desktops. I remember the days when I immediately noticed the extra couple of quarts of horsepower I felt when I upgraded my processor after only a year. It used to be that the difference between processors 5 years apart was at least a hogshead! Now, however, I expect that most people when using everyday applications only notice a couple of teaspoons of difference between a processor 5 years ago and now, if that.

 

[mrmt]

Moore's Law, Gustafson argues, wasn't just about the technological ability to put more transistors on a chip, but also the economic feasibility of doing so. "The original statement of Moore's Law is the number of transistors that is more economical to produce will double every two years", Gustafson said. "It has become warped into all these other forms but that is what he originally said."

This.

t's worth exploring what scenarios the Moore's Law wouldn't be applicable. When investing the kind of money Intel does, bringing new things to the process like finfet, then it might be worth it. Intel 22nm finfet brought nice efficiency gains and economic gains from reduced die size, I think they did a nice die shrink. OTOH AMD is tied to a subpar foundry that has practically canned their 20nm process because the implementation seemed too botched that could not attract a single customer.

With MPUs reaching maturity, it doesn't make sense to have volumes like AMD, or small foundry divisions inside a company. The benefits are still there, but the economic bar (CAPEX/Volumes) is too high for most people to afford, and this includes AMD.

AMD is too small compared to the big ARM manufacturers, or even with Nvidia, and they gave up their fabs, so they aren't an interesting customer for top tier foundries (TSMC) and can't afford to finance themselves. So the Moore's law indeed will slow down for AMD, not because of physics, but because of economy of scale. The botched GLF implementations will just make things worse.

 

[R0H1T]

Can/will Intel ever produce chips for AMD especially if its a good deal for them ? Given the overall state of the industry, it isn't such a long shot anymore :awe:

 

[mrmt]

Can/will Intel ever produce chips for AMD especially if its a good deal for them :awe:

Never. With the economic bar raising up, meaning higher volumes and higher CAPEX for the investment to pay off, Intel will try to push AMD out of the market, not help them to stay in the market.

 

[jihe]

Moore's Law: AMD's end is near.

 

[R0H1T]

Never. With the economic bar raising up, meaning higher volumes and higher CAPEX for the investment to pay off, Intel will try to push AMD out of the market, not help them to stay in the market.

But IIRC most tech enthusiasts agreed, even here on AT, on the common notion that Intel wouldn't let AMD go belly up fearing antitrust issues ! Has something changed in the meantime ?

 

[Fx1]

Once we get to 10nm they will just add more cores and push the paralel computing power increases. Its a joke we are still only at 4 cores on a £200 CPU. Intel is milking the market because it knows once they are in to single nm numbers its going to be to hard to keep shrinking.

 

[BallaTheFeared]

Meanwhile, Intel says:

Mark Bohr who overseas much of Intel’s processor research, was quoted as saying, “The end of Moore’s Law is always 10 years away. And, yes, it’s still 10 years away.”

()

 

[ShintaiDK]

AMD sounds bitter...being stuck on 28nm when Intel ships 14nm.

 

[mrmt]

But IIRC most tech enthusiasts agreed, even here on AT, on the common notion that Intel wouldn't let AMD go belly up fearing antitrust issues ! Has something changed in the meantime ?

If AMD went belly up because of Intel practices, yes, that would be a problem, but going because of incapacity to compete? A regulator propping up a company that failed in the open market? I doubt it.

Plus reality is a lot different of that of 10 years ago. ARM is around the corner, with resources that matches or even dwarfs Intel's. So in the grand scheme of the things, I don't think AMD is that relevant anymore.

 

[R0H1T]

If AMD went belly up because of Intel practices, yes, that would be a problem, but going because of incapacity to compete?

That's understandable I guess.

A regulator propping up a company that failed in the open market? I doubt it.

You mean something like a GM/Ford bailout ? You never know how the govt thinks !

Plus reality is a lot different of that of 10 years ago. ARM is around the corner, with resources that matches or even dwarfs Intel's. So in the grand scheme of the things, I don't think AMD is that relevant anymore.

The point I'm making here is relevant to the x86 market, as a subset of the overall industry, so whether the regulators/govt see it fit to intervene is totally up to them but I'm neither holding my breath nor anticipating them to turn a blind eye especially considering the number of OEM's that'll be greatly affected if AMD indeed goes bankrupt !

 

[mrmt]

The point I'm making here is relevant to the x86 market, as a subset of the overall industry, so whether the regulators/govt see it fit to intervene is totally up to them but I'm neither holding my breath nor anticipating them to turn a blind eye especially considering the number of OEM's that'll be greatly affected if AMD indeed goes bankrupt !

I explained in an older thread why AMD bail out isn't a straightforward issue, but the basic idea is that there is no liability hindering AMD like the autos had, but structural problems.

The other problem is that AMD is largely irrelevant for pricing. They are already in their bottom limit, they cannot go any further. So you have to question what utility they have for oems when it is them asking for higher prices, not Intel. Last, but not least, how can they compete with that subpar foundry partner of them?

 

[frozentundra123456]

AMD sounds bitter...being stuck on 28nm when Intel ships 14nm.

Sounds like sour grapes from a company stuck with a sub par foundary. The bigger problem for Intel is that node shrinks seem to be bringing less benefits and introducing heat dissipation issues.

 

[frozentundra123456]

That's understandable I guess.
You mean something like a GM/Ford bailout ? You never know how the govt thinks !
The point I'm making here is relevant to the x86 market, as a subset of the overall industry, so whether the regulators/govt see it fit to intervene is totally up to them but I'm neither holding my breath nor anticipating them to turn a blind eye especially considering the number of OEM's that'll be greatly affected if AMD indeed goes bankrupt !

Personally I don't think there would be a bailout because amd does not have the political and financial clout the unions and automakers had.

 

[Idontcare]

Yet another "we are the new VIA" article from AMD. I think they are finally getting over their delusions and beginning to see reality as every other x86 business before them.

But they still can't look in the mirror and accept blame for anything here, can they? Now they have to blame a non-existent "law" for their follies.

Yeah it was Moore's Law that made them hire Hector only to piss $5.4B into ATI and then throw off their fabs while tying themselves to the mast. Yeah, AMD's woes are because of Moore's Law! :|

Never mind the historical fact that AMD can't correctly predict when Christmas will get here, let alone their own products or the performance characteristics of them...these guys have no business telling the rest of the adult world how business is going to look 10yrs from now.

 

[R0H1T]

Personally I don't think there would be a bailout because amd does not have the political and financial clout the unions and automakers had.

So you don't think the lost sales would impact mobo, display, psu, dram/nand makers enough(not to mention loads of Online/brick & mortar stores et al) to warrant a bailout from the govt ? Don't get me wrong here but I can see why the lobbying system in so fucked up in the US !

 

[mrmt]

So you don't think the lost sales would impact mobo, display, psu, dram/nand makers enough(not to mention loads of Online/brick & mortar stores et al) to warrant a bailout from the govt ? Don't get me wrong here but I can see why the lobbying system in so fucked up in the US !

What impact? Intel, also an american company, has both the capacity and the will to supply the market if AMD fails. Except for the fanboys, nobody would notice if AMD went under.

 

[Idontcare]

So you don't think the lost sales would impact mobo, display, psu, dram/nand makers enough(not to mention loads of Online/brick & mortar stores et al) to warrant a bailout from the govt ? Don't get me wrong here but I can see why the lobbying system in so fucked up in the US !

What "lost sales"? The sales volume is fixed, all you are doing is shuffling around who those sales go to.

In the absence of AMD those sales would go to Intel or Asus or MS or any number of other existing players in the hardware market.

It wasn't the end of the world Qimonda went bankrupt, or Spansion, or Elpida, or Hynix...etc. If AMD goes bankrupt it will be devastating in a personal way to those who's personal income is tied to AMD, but the rest of the industry and world would barely notice outside of reading about it in a headline or two.

edit: mrmt beat me to it

 

[lagokc]

But IIRC most tech enthusiasts agreed, even here on AT, on the common notion that Intel wouldn't let AMD go belly up fearing antitrust issues ! Has something changed in the meantime ?

Yeah, a bunch of ARM chips will be pushing into the desktop and server space in the next few years.

 

[MisterMac]

So you don't think the lost sales would impact mobo, display, psu, dram/nand makers enough(not to mention loads of Online/brick & mortar stores et al) to warrant a bailout from the govt ? Don't get me wrong here but I can see why the lobbying system in so fucked up in the US !

dGPU possibly.
But it will be bought out and continued by someone.
That's hawt intellectual property right there.

CPU?

Naw - AMD has no clout or grounds or is the same size as Ford\GM.

We're talking what 7000-8000 jobs of HIGH PAYED (mostly) qualified adult educated invididuals versus ....100's of thousands of general average public jobs.

If AMD goes bellyup OEM jobs would be same - they'd just buy intel.
Can't be said for the car selling dealerships and Ford\GM itself.

Monsterous difference - while yes not good for the consumer in a free market theory type of thinking the question is if in this subset of the market it would matter much at all.

 

[Charles Kozierok]

This has all been talked about for years. AMD is just late to the party.. as usual.

Whether it is in ten years or twenty or thirty, at some point the end for silicon will indeed be reached; the writing is already on the wall. The real question is who is going to take us beyond that... and it's not going to be AMD.

Even before Moore's Law truly reaches its end, it has effectively reached its end for those who really care about increasing performance. They are still packing more transistors in there, but they're not using them to make chips run faster, just adding cores, cache and GPU.

 

[R0H1T]

What impact? Intel, also an american company, has both the capacity and the will to supply the market if AMD fails. Except for the fanboys, nobody would notice if AMD went under.

The overall x86 market would shrink dramatically & I can't see how that helps anyone bar Intel & Via to a lesser extent ! Then there is the GPU market that'll be hit hard as well, I really don't see how you can accept monopolistic firms that'll virtually own 60~70 billion $ market with their only real competitor going bankrupt ? This certainly can't be good for the consumers !

 

[BallaTheFeared]

Natural Selection, only the strong survive!