AMD: Moore's Law's end is near

[Idontcare]

See my comment about regulated monopolies. Not market monopolies that are unregulated, with unchecked prices.

Unregulated monopolies are bad for consumers, in every case.

To be fair though that is universally true because the common denominator there is "humans", not "monopolies".

Unregulated people (decision makers in the monopoly), as a rule, are bad for other people (consumers).

That was true even before the invention of business entities or regulation.

Competition is just a way of regulating people, a socially acceptable way that is also popular with the vast majority of people.

 

[tuxberg]

To be fair though that is universally true because the common denominator there is "humans", not "monopolies".

Unregulated people (decision makers in the monopoly), as a rule, are bad for other people (consumers).

That was true even before the invention of business entities or regulation.

Competition is just a way of regulating people, a socially acceptable way that is also popular with the vast majority of people.

Actually, market competition in the literal sense (not the heavily regulated thing we call a marketplace under the current regimes) is not a way of regulating people.

To have a free market is to allow people to self-regulate, to the extent that they are entirely responsible for their actions. The dimension of social propriety is really just an extension of economic utility. Even the most anti-capitalist maoist and the most impoverished monk are performing actions in the hopes that the products of those labors are going to benefit them (or someone they care for as much) later on.

The idea that human beings need regulation to be anything like socially cohesive or productive is absolute nonsense. Were that the case, north korea would be the most productive nation on earth. Secondly, to tip back to the point of monopolies: there is no such thing as a natural or unregulated monopoly. Monopolies are created by the presence of government in the market, be that in the form of an expressed license to monopolize the business or in the form of restrictions that make entry into the market extremely difficult (such as incorporation, intellectual property which are entirely state-generated concepts).

To get something-like on-topic: There is absolutely no way the reduction of competition could improve a market (for consumers). Anyone suggesting that AMD isn't reducing the cost of CPUs across the board hasn't got the most basic grasp on economics and is living in a Keynesian fantasy-world where Krugman's aliens are breaking windows and saving the world economy.

 

[Revolution 11]

It meant something physical up to the 0.35um node.

In the far future, decades from now, historians will look back on this era and label the 0.35um node as being the timeframe at which true traditional scaling broke down.

This begs the question of how big are "22 nm" transistors in a physical measurement. Are transistors still close to 0.35 um in size or smaller? (it does not matter but I am curious)

 

[Idontcare]

The idea that human beings need regulation to be anything like socially cohesive or productive is absolute nonsense.

I'm not sure why you would write that. What do you think laws are if not the regulation of humans in a society in the pursuit of productive initiatives. How long have cultures had laws? What was the written language created for?

Were that the case, north korea would be the most productive nation on earth.

You were probably going somewhere with this but you lost me.

North Korea does a lot of things, doesn't mean doing any one of them is to be automatically deemed "bad". They make babies, is making babies now to be considered bad because North Koreans do it?

 

[Idontcare]

This begs the question of how big are "22 nm" transistors in a physical measurement. Are transistors still close to 0.35 um in size or smaller? (it does not matter but I am curious)

They are 3-dimensional so labeling them by a specific dimension becomes all more meaningless. Like using the wheelbase measurement of your car to tell me about is gas mileage.

I suppose one could go by the minimum dimension of the cross-section of a fin, but that doesn't really cut the mustard either.

Nowadays we usually go by what we call the "contacted gate pitch" or the "M1 half-pitch". Neither are 22nm (they are more like 90nm), its just something we can measure with a SEM or TEM and document as a physical measurement.

 

[Maxima1]

Actually, market competition in the literal sense (not the heavily regulated thing we call a marketplace under the current regimes) is not a way of regulating people.

Adam Smith listed a wide range of functions that government can legitimately perform including regulation, enforcement of private obligation, and activity in markets to change outcomes.

The idea that human beings need regulation to be anything like socially cohesive or productive is absolute nonsense. Were that the case, north korea would be the most productive nation on earth.

If that's the case, then your hands off experiment has been Somalia.

 

[Homeles]

To get something-like on-topic: There is absolutely no way the reduction of competition could improve a market (for consumers). Anyone suggesting that AMD isn't reducing the cost of CPUs across the board hasn't got the most basic grasp on economics and is living in a Keynesian fantasy-world where Krugman's aliens are breaking windows and saving the world economy.

If there's anybody I'm okay with having my money, it's Intel. They're easily one of the best companies I can give money to and expect progress to be made for the betterment of the human race.

This begs the question of how big are "22 nm" transistors in a physical measurement. Are transistors still close to 0.35 um in size or smaller? (it does not matter but I am curious)

"Begging the question" refers to a rhetorical fallacy. The phrase you're looking for is "raises the question."

 

[tuxberg]

I'm not sure why you would write that. What do you think laws are if not the regulation of humans in a society in the pursuit of productive initiatives. How long have cultures had laws? What was the written language created for?

I am not aware of how long the ancient analogue of a "law," has been around, but that is immaterial to the argument. Slavery was around longer than the written word, but I don't think anyone would argue that was so prescient an idea, do you?

The writing system was demonstrably created for economic reasons. Keeping up with surplus goods and facilitating exchange that would otherwise be an educated guess. None of that; however, requires a legal system. A legal system has nothing at all to do with this exchange. Legal codes come from rulers, not traders.

You were probably going somewhere with this but you lost me.

North Korea does a lot of things, doesn't mean doing any one of them is to be automatically deemed "bad". They make babies, is making babies now to be considered bad because North Koreans do it?

If I lost you with that simple example, I doubt there's much else to be said. You advocated for human regulation. North Korea is notoriously high on human regulation. Therefor, using the logic you presented, they would be the most efficient nation on earth due to their incredibly high rate of human regulation.

If there's anybody I'm okay with having my money, it's Intel. They're easily one of the best companies I can give money to and expect progress to be made for the betterment of the human race.

And for every fanboy there is another objective consumer.

 

[tuxberg]

Adam Smith listed a wide range of functions that government can legitimately perform including regulation, enforcement of private obligation, and activity in markets to change outcomes.

I fail to see why your appeal to authority makes any sense in this context.

If that's the case, then your hands off experiment has been Somalia.

A nation torn apart by government-sponsored wars is the poster-child for statelessness. How convenient? I guess people need something to hold up when it's so obvious even the most liberal western democracies are comfortable murdering children with drones.

 

[Homeles]

And for every fanboy there is another objective consumer.

And they're all getting milked into the technological progress bucket. Sounds good to me.

 

[jaqie]

. . . . mrmt is not saying that a monopoly will always yield to lower prices, he is merely pointing out the mathematical possibility that it can . . . . That doesn't mean the product will cost the consumer less . . . . But it is not a certainty

thank you, that really makes a difference, I did totally misunderstand.

 

[iCyborg]

Generally on the matter of CPU monopoly, I'm not replying to anyone in particular:

http://www.columbia.edu/~brg2114/files/dynduo.pdf

Excerpt from conclusions, page 1188, 49 of the pdf doc (hopefully I'm not violating any copyright laws...)

"In this paper, we estimate a dynamic model of durable goods oligopoly
with endogenous innovation and use it to assess the effect of competition
on innovation in the PC microprocessor industry. Consumers are better
off under a duopoly because of lower margins: consumer surplus is
higher with AMD competing against Intel than without AMD. However,
in support of Schumpeter’s hypothesis, industry innovation is higher
with Intel as a monopolist."


Unless someone can convince me that Intel is a charity and/or a service with public interest before anything else, I will rather believe this study that may have flaws, but is still miles better than the mainly hand-waving arguments claiming that we would have better prices with just Intel. Though as they also claim, it would actually increase innovation, so I suppose it would benefit those consumers for whom "money is no object". For the rest of us, their conclusion is that we're better off without a monopoly.

 

[UaVaj]

software segment is so far behind in multi core processing. it will be a long long long while before 10nm chip even matters.

until software can utilize 100% of the hardware. the transistor count won't matter - at least yet.

 

[Idontcare]

Adam Smith listed a wide range of functions that government can legitimately perform including regulation, enforcement of private obligation, and activity in markets to change outcomes.

I fail to see why your appeal to authority makes any sense in this context.

If he lost you with that simple example, I doubt there's much else to be said.

If that's the case, then your hands off experiment has been Somalia.

A nation torn apart by government-sponsored wars is the poster-child for statelessness. How convenient? I guess people need something to hold up when it's so obvious even the most liberal western democracies are comfortable murdering children with drones.

Not too unlike an equally convenient example of using an extreme datum point of a police nation under the control of a despot like North Korea?

LOL at the irony.

 

[Homeles]

software segment is so far behind in multi core processing. it will be a long long long while before 10nm chip even matters.

until software can utilize 100% of the hardware. the transistor count won't matter - at least yet.

The main revolution, as it has always been, is with cost. With 10nm, you would quite literally be able to make a Sandy Bridge processor with a die nearly 1/10th the size.

I won't live to see this day, but someday high performing computers will be practically free. Really, if there weren't R&D costs to fund, we'd be able to quad core Sandy Bridge processors for something silly like $20.

 

[R0H1T]

I don't think they are pushovers at all. But based on the 40+ year history of Intel and the past 20 years of which I've been a keen observer that is my opinion. The manufacturers you mentioned are filling a void. A gas expands quickly in a void.

You've got to ask yourself why no one, no one on the freaking planet has been able to build a better x86 processor than Intel? And we know there is money to be made with x86 processors. You can claim they started first, or they had/have a monopoly with MS, or whatever, but in the end they still have far and away the best product. Only AMD is even close.

With all of their revenue Samsung has never even dipped a toe into Intel's water. In addition, the rate of increase of Intel's GPU's has been astounding over the last few generations. Another case of expanding into a vacuum. And by that I mean from where they came GPU-wise, which was crap.

There's nothing like a heavyweight fight and this is going to be a good one. I just don't see anyone having the design talent/experience, fabs, and marketing savvy to compete with Intel in any microprocessor market they decide they want. What should be even scarier for the likes of their competition you mentioned is the fact that they don't seem to be developing their own ARM parts. Either they are really stupid and are about to make a possibly fatal miscalculation or they are really smart.

To begin with you are asking the wrong question, just for starters, so lets say I am a phone manufacturer now ~ I'd love to get a 10% hike in performance at the same price(current levels) but I'd prefer it to be even cheaper than what it is now, talking about mobile segment here. Now have you ever considered the fact that Intel will want a 40% hike in price(per chip at current levels) for something like a 20% performance gain ? Can but more importantly will they settle for lower margins ?

Also that analogy of gas filling void is wrong, its rather something like CO2(ARM) filling more of the atmosphere at the expense of O2(x86) except here the proportion of the former is alot more than the latter in terms of volume. See this for instance ~

If the likes of Allwinner can provide the 99.9% of market that doesn't demand Crysis on a tablet for $10 at 28nm they are going to have a field day at 20nm.

It is from another thread but more than relevant here i.e. if Intel don't change their business model going forward, I don't see them even scratching the surface in mobile/tablet segment let alone dominate it ! The romanticism/fanaticism associated with Intel here is quite frankly hysterical especially when it concerns the lowest end of the market(smart devices) & there is absolutely no incentive for manufacturers to choose them over ARM, not without the legacy x86 baggage, unless they're selling really cheap parts!

 

[shady28]

Folks are missing some of the finer nuances of process shrink.

For one, once we went below 40nm, additional circuitry is required to take care of leakage within the chips. ARM's chief engineer alluded to this a couple of years ago, stating below 40nm we would see diminished returns on further die shrinks.

Process shrinks are used to make smaller and lower power chips - not faster ones. That's because there's a problem that they haven't solved yet - making a large, high density chip with a small process like 22nm (or even 28nm) where you can cram far more circuitry into a given area requires even more power.

So, they shrink the die, and put more chips on a wafer. This lowers power consumption and increases profit margins (more chips per wafer). What it does not do is significantly increase IPC or overall performance.

Take a look here :

http://preshing.com/20120208/a-look-back-at-single-threaded-cpu-performance

And that chart misses the last 2 years. What's happened in that time?

As for compute power - Sandy Bridge was the last decent bump, but Ivy was not so much, and Haswell appears to be very much a non-event in the performance arena.

If we look at just the last 3 years, we've now gone from ~21% per year from that chart to 5-10%/yr in the last 3.

That pales in comparison with the 51%/yr jumps in desktop compute power we had in the 1990s. It will now take about 4-5 years to equal one year of advancement in compute power in the 1990s.

I also agree about the software part, multi-core simply hasn't been used effectively in most applications. That also contributes, as there is no force driving chip makers to add more cores than what we've had for the past 5 years. Combined with negligible increases in IPC, we are stuck.

 

[MaxPayne63]

edit: My mistake; I assumed numbers were annualized and not literal.

 

[krumme]

Right now in the phone and tablet market Intel isn't there. I mean the 800lb gorilla isn't there. They just sent a monkey in to have a look around.
But the gorilla is coming...

Except this time they didnt end up in the jungle but the zoo

 

[krumme]

Still think it's ending at 14. That doesn't mean that less than 14 won't happen, it's just that the node progression is going to start slowing down.



AMD at this point has no non-windows strategy; and considering how much of a disaster Windows 8 is, Temash is a non-starter.

That should exactly be an opportunity for AMD and the bod must be standing on the shoulder of rr here. Windows got the money and needs exacly temash. If rr have not used thT strategcally he is not doing his job.

 

[ChippyUK]

I wonder if the future of the tick-tock cycle will change into a tick-tock-tock or longer to serve yearly releases of processors (to keep the income flowing).

I'm sure other methods will be implemented other than reducing die size will be used when it becomes more cost effective to employ that strategy. Surely further SOC development is likely in the future will yield some improvements.

 

[krumme]

Nobody is arguing for competition in water distribution, or sewage. In those industries a second player would mean just waste. Also if you have two competitors but one is very inefficient, and loses a lot of money and the other is too far away in the price curve from their most efficient price, then yes, consumer will be better with a given monopoly operating at maximum efficiency.

The point is that some times monopoly is the most efficient organization you can have, and some times it's even desirable. That's not to say that the situation is desirable in Intel's case, but it's not just ignorance to think that monopoly can yield lower prices to consumers.

Can we please stop this argument now. Delivering cpu is not a sewer system and all historical facts shows the lack of competition here raises prices. Thats not to say there is not a natural tendency towards monopoly here. There is. But who in their fucking sane minds thinks Intel will lower their prices? Come on. Do we have to get the old mails from Otellini again. They are running a business, they work each day to get monopoly like advantages. And you are asking them to lower prices?
I dont think this discussion will bring us somewhere. What the f is the purpose?

 

[beginner99]

And dont talk about socialism when you dont have a clue what it is. And seems you skipped capitalism as well in school.

A lot of Americans just don't get what a social democracy is and hence just call it socialism and think their social security thingy is the greatest thing on earth.

To give that perspective, were I live if you are unemployed you get 80% percent of your last salary for 2 years. There is an upper limit of course, but it's quite high, about $120'000 yearly salary. (eg you get a maximum of $240'000) And there are conditions, like you must have worked here for at certain time frame and you must send out a certain amount of resumes and attend courses and so on.
And when you do work, a certain percentage of your salary is automatically removed and put into that "pot". That's why it is called "unemployment insurance".

And about monopoly:

Train system in UK has become complete crap. It's less secure, less convenient and more expensive since the monopoly of the gov was removed. It is by far worse than were I live and at the same time more expensive and it is already very expensive here...

Basic infrastructure should IMHO never ever be privatized. It will just end up worse. It's much better to pay more taxes instead because government will not hire 100 additional high paid manager that just BS around all day...if you privatize, your money will go into pockets of such managers and not better infrastructure.

 

[ShintaiDK]

Can we please stop this argument now. Delivering cpu is not a sewer system and all historical facts shows the lack of competition here raises prices. Thats not to say there is not a natural tendency towards monopoly here. There is. But who in their fucking sane minds thinks Intel will lower their prices? Come on. Do we have to get the old mails from Otellini again. They are running a business, they work each day to get monopoly like advantages. And you are asking them to lower prices?
I dont think this discussion will bring us somewhere. What the f is the purpose?

Lack of competition in the CPU industry gave us the cheapest CPUs ever in history. Try think on why it happend.

You need to understand the difference between a dynamic and static demand segment. Intel is not selling food, water, electricity, heat or something else that you pretty much cant avoid. If Intel raises prices, volume shrinks, revenue shrinks and profit goes down. If Intel doesnt innovate, then revenue goes down, volume goes down, profit shrinks. Unlike water that we daily need, Intel needs to give us a reason to buy a CPU. Else your current CPU will work just fine the next 10 years+ while Intel is bankrupt and gone in the same timeframe.

Intel is gaining the maximum profit they can right now. I am certainly not blind for business. But you seem to think that the volume stays (close to) the same nomatter the price. And thats simply incorrect.

 

[krumme]

Folks are missing some of the finer nuances of process shrink.

For one, once we went below 40nm, additional circuitry is required to take care of leakage within the chips. ARM's chief engineer alluded to this a couple of years ago, stating below 40nm we would see diminished returns on further die shrinks.

Process shrinks are used to make smaller and lower power chips - not faster ones. That's because there's a problem that they haven't solved yet - making a large, high density chip with a small process like 22nm (or even 28nm) where you can cram far more circuitry into a given area requires even more power.

So, they shrink the die, and put more chips on a wafer. This lowers power consumption and increases profit margins (more chips per wafer). What it does not do is significantly increase IPC or overall performance.

Take a look here :

http://preshing.com/20120208/a-look-back-at-single-threaded-cpu-performance

And that chart misses the last 2 years. What's happened in that time?

As for compute power - Sandy Bridge was the last decent bump, but Ivy was not so much, and Haswell appears to be very much a non-event in the performance arena.

If we look at just the last 3 years, we've now gone from ~21% per year from that chart to 5-10%/yr in the last 3.

That pales in comparison with the 51%/yr jumps in desktop compute power we had in the 1990s. It will now take about 4-5 years to equal one year of advancement in compute power in the 1990s.

I also agree about the software part, multi-core simply hasn't been used effectively in most applications. That also contributes, as there is no force driving chip makers to add more cores than what we've had for the past 5 years. Combined with negligible increases in IPC, we are stuck.

Back in 1993 i was making a report. About 200 pages in word, with graphs on each page linked in from Excel.

It took minutes to open. Calculation had to be done manually in Excel. Printing took several hours.

Just to turn pages in word was a pain as there was no 2d acceleration yet !

All was done on 8 Mbyte of memory on win 3.11 on an 386.

The benefit for officework getting new computer then was enormous. Each time it was such a blast to get a new machine i got a rush.

The effect already started to fade in the start of the 00. And for 2005 forward it have just been a meh, with only the first ssd, to give some of the feeling back - and here to the start of the 00.

I think we forget how an extremely huge difference going from an 386 to an 486 with accelerated 2d was for the efficiency and experience in general.

The basic problem for the industri today, for both desktop and laptops, is that they dont deliver the same user benefit. Even for phones we are going to see the benefit stop,- and thats why Apple is betting on more gfx here.

All business is doing and going into commodity some day. The desktop is just heading that way and have for years, we just have to accept it.