How long can intel maintain such a large process lead over rivals?

[Irenicus]

I keep reading about performance shortfalls of chips like bulldozer and piledriver and wonder to myself when will it ever end?

Even if AMD or some other company designed a chip that was better than haswell or beyond from intel... would it even matter if the other foundries are 2-4 years behind in process technology? Wouldn't intel be able to just brute force a larger die size to compensate?

I saw the die size comparison with the latest trinity review and trinity was over 200 square mm compared to 160 for intel..

Now it's probably the case that production of 22nm trigate chips is more expensive than what the competition is doing, such that trinity chip and ivy bridge chip of equal die size would be more expensive for intel to produce, but still, how are the other foundries supposed to compete?


Is the only relief going to be an intervention by the limits of physics with ever increasing difficulty of going smaller?

I am probably off base on much of this, I don't know about this stuff in any depth, so maybe someone here can set me straight.

 

[ShintaiDK]

Intel is expanding its lead. And the competition is dropping of. New processes got a ROI to forfill. For example for 14nm only Intel and TSMC can actually make a profit out of it. Samsung is in a 50/50 area. And beyond 14nm for logic. Samsung is out for sure.

The others are already 3½-4½ years behind if you look beyond the node scale itself.

Intel could bruteforce a bad design and still win. But there are limits to how bad the design could be. Essentially Intel could produce a quadcore IB today with 3-4x the IGP power of the current HD4000. And still make it cheaper than what AMD can make its dualcore Trinity with CMT. Not to mention AMD would have to share profits with the foundry they use.

Another point is that Intel, unlike any other can design its manufactoring and chip design together for maximum performance, yield and so on.

Its simple economics tho. Intel spends several times more on chipdesign than AMD as well. AMD would need a miracle to beat Intel. While Intel could still beat AMD on a bad day. Its the same that happend to the mass of previous x86 makers. The scale of economics killed them off one by one. Thats what competition is.

So to your question. They cant compete.

 

[Don Karnage]

Anyone have an idea how many chips intel sells a year? 100 million? 200 million?

 

[blastingcap]

Anyone have an idea how many chips intel sells a year? 100 million? 200 million?

Um, define "chips." Are you talking about desktop, server CPUs? Chipsets? Flash drives? etc. You get the point....

Anyway I don't think INTC cares how many chips they sell (within reason). They care about profit. They leave the low-margin stuff to AMD. INTC's gross margins are crazy, like 67% last time I checked.

 

[ViRGE]

In short: massive volume of chips + deep pockets + the ability to hire and retain some of the brightest people in the field. They're so far ahead that they make money hand over fist, which they invest in making sure they stay so far ahead.

They are quite frankly a fab company that happens to design chips on the side so that they have something to run through their fabs.

 

[Don Karnage]

Um, define "chips." Are you talking about desktop, server CPUs? Chipsets? Flash drives? etc. You get the point....

Total. Desktop, Server and Mobile etc etc etc

 

[Arachnotronic]

Intel rocks.

 

[blastingcap]

Total. Desktop, Server and Mobile etc etc etc

A lot.

And they like to focus on higher-margin stuff, like high end CPUs and SSDs, leaving the lower-margin stuff to rivals like AMD.

Plus they are making a serious push into graphics processing now, with Haswell rumored to be up to 3x the speed of Ivy Bridge's HD4000. If that rumor holds up, we're talking speeds of ~HD4850 which is impressive for an embedded GPU.

Plus they are actually making progress in mobile, with their first real competitor to ARM in smartphones. Anandtech reviewed it recently. It will only get better.

They still have a lot of work to do in terms of fighting AAPL in ultrabooks/MacBook Air clones though.

They are also the only company out there with proven non-2D transistors in the mass market that I know of. (I hesitate to call tri-gate 3D. 2.5D maybe?)

 

[ShintaiDK]

Anyone have an idea how many chips intel sells a year? 100 million? 200 million?

Around 350mio x86 CPUs. Some 60% is mobile CPUs. And about 5mio CPUs are server CPUs if I recall right.

World PC shipments account for around 375million. But I dont think that includes servers, settop boxes, appliances, tablets, smartphones etc where you also see atom/bobcat and regular CPUs.

 

[IntelUser2000]

such that trinity chip and ivy bridge chip of equal die size would be more expensive for intel to produce, but still, how are the other foundries supposed to compete?

That may be true if everything else is equal but its not. Higher volumes mean lower costs.

Around 350mio x86 CPUs. Some 60% is mobile CPUs. And about 5mio CPUs are server CPUs if I recall right.

Intel shiped 4 million server CPUs per quarter last year, and was close to 16 million for the year.

servers, settop boxes, appliances, tablets, smartphones etc where you also see atom/bobcat and regular CPUs.

All except servers are irrelevant in volume(the highest is Atom in settop boxes at few million/year with Appliances/Tablets/Smartphones at less than 1/10th of that, probably all 3 combined). For 2011 the total sales for both companies were little over 400 million. Without servers that's close to 390 million, and Intel has approximately 80% of that making it little over 300 million.

 

[2is]

AMD had approximately 2 years which to at least be competitive, and that was back in the Athlon 64 and early Athlon 64 X2 days. They had a better product, they had no problems selling it for MORE than competing P4's and Pentium D's and for the first time in a long time, they were actually making a profit. They took advantage of that opportunity by remaining largly stagnent in their development for several years, until BD of course which was a move in the wrong direction. And they aren't coming back from it. Intel would have to royally mess up two generations worth of processors and AMD would have to actually improve on what they have, then improve on it again in the next generation for them to catch up, which isn't going to happen.

Sad part is, given AMD's track record as of late, I think it's for more likely Intel will screw up an upcoming CPU then AMD actually coming out with any sort of breakthrough.

 

[jpiniero]

I think Intel is hitting a wall because of the heat density.

 

[amenx]

AMD is already producing better chips than Intel. Intel sell more chips due to slick marketing. Not everyone is a gamer, and those who are should just focus on the GPU not CPU. Reviews that show Intel ahead are from pro-Intel sites and the benches they use are cherry picked to favor Intel. These are 'bentmarks'. Those who buy Intel are just sheep who were fed misleading information or plain out fanbois. So goes the arguments at AMD Zone, the last bastion of feel-good-about-my-AMD-CPU-and-spit-on-Intel remaining on the internet.

 

[pm]

I heard an interesting lecture from Andy Bryant (chairman of the board for Intel) about the costs of fabs. I did a google search for it and couldn't find it, but I did find this article from the UK's Financial Times which seems to make a similar arguement.

http://www.ft.com/cms/s/0/d2706a60-758d-11de-9ed5-00144feabdc0.html#axzz1vNKBOy6c

Basically the thinking is that in order to be able to build a new 22nm fab for ~$4 billion US dollars, you then have to actually make enough chips at enough of a profit to make this investment pay off. And there are only a few companies left in the world who, first can afford to throw US$4bn at a factory, and then can sell enough chips at enough profit to actually make a profit from it. According to the arcticle it came down to Intel, Samsung, Toshiba, Texas Instruments and STMicroelectronics, and possibly TSMC.

There's a few people who say that the end of Moore's Law won't come from a physical limit imposed by the size of atoms, or the laws of physics, but the fact that the manufacturing equipment to make the chips will be so prohibitively expensive that it doesn't make financial sense any more.

 

[ShintaiDK]

AMD is already producing better chips than Intel. Intel sell more chips due to slick marketing. Not everyone is a gamer, and those who are should just focus on the GPU not CPU. Reviews that show Intel ahead are from pro-Intel sites and the benches they use are cherry picked to favor Intel. These are 'bentmarks'. Those who buy Intel are just sheep who were fed misleading information or plain out fanbois. So goes the arguments at AMD Zone, the last bastion of feel-good-about-my-AMD-CPU-and-spit-on-Intel remaining on the internet.

That was a funny read. You sure got humour

 

[ShintaiDK]

There's a few people who say that the end of Moore's Law won't come from a physical limit imposed by the size of atoms, or the laws of physics, but the fact that the manufacturing equipment to make the chips will be so prohibitively expensive that it doesn't make financial sense any more.

This I agree on. Its all about the ROI. Higher volume helps to offset that. But at a certain point it just wont be worth it. And by then Intel is alone in the game. In that list of yours. Some of those companies already gave up long ago. TI for once.

Today there is only TSMC, Samsung and Intel left. GloFo is already losing money on the processnode and they desperately try to increase volume to compensate. And Samsung is close to jump ship too.

Fab42 cost Intel 5 billion$. And they use 9 billion$ alone this year on fabs. (New, upgrades etc.) To compare they used 5 billion$ in 2010 on the same.

 

[Charles Kozierok]

Conventional CMOS technology is hitting the wall. We've already seen that Intel has had to go to FinFET technology and that's really just the latest in a long line of "games" that have been used to deal with process challenges -- gating, immersion lithography, multipatterning, etc.

The more of these tricks that are required, the more difficult and expensive it gets to set up a fab, and the harder it is for it to be cost-effective. The best hope for companies like AMD is for a major breakthrough in transistor technology, such as the 3D chips being currently investigated.

Or, to shift their market focus, which is what they are doing.

 

[pelov]

I was reading this recently and it really put the costs of such endeavors into perspective. Needless to say, it's a metric butt-ton of money required

Makes going fabless sound like a pretty good deal until you realize your arch rival has the best fabs and the most money

 

[Imouto]

Since the future in this field will be graphene based I think that Samsung will have a huge lead. They have way more money to burn in development as they're doing atm researching graphene.

Intel and IBM are researching aswell, but Samsung is really interested in graphene since they can make touchscreens, NAND, batteries and a lot of stuff out of it.

Miracle material they say.

 

[Hatisherrif]

As long as capitalism lives, companies like Intel will be on top. As long as engineers seek only the jobs with the highest pay, Intel will win. As long as we are robots, Intel will win.

 

[Charles Kozierok]

As long as capitalism lives, companies like Intel will be on top. As long as engineers seek only the jobs with the highest pay, Intel will win. As long as we are robots, Intel will win.

Yawn.

Not everything is about a capitalist conspiracy to control the world. Sometimes one company wins because they deserve to win.

When Intel got too full of themselves, we ended up with the Pentium 4 "space heaters" and AMD had its chance to really dominate. But Intel has done the right things over the last few years, while AMD has made one mistake after another.

That's why Intel is leading.

And let's not forget that there are markets aside from x86 -- Intel only dominates its own specific CPU segment.

 

[Imouto]

Intel doesn't deserve to win. Is miles away from deserving it.

Is amazing how fast ppl forget about all the antitrust shit Intel has been doing for decades.

 

[Charles Kozierok]

Intel doesn't deserve to win. Is miles away from deserving it.

Why?

Is amazing how fast ppl forget about all the antitrust shit Intel has been doing for decades.

Long over and settled. And not even remotely related to why Intel is doing well process-wise.

 

[Fjodor2001]

It would be interesting to break down the costs of building a FAB into sub-costs. How can it cost 7 billion dollars to build a 22 nm chip factory? I mean after all it's just a building, with clean room facilities, and production lines with machines for producing chips?

Intuitively, one would expect the research costs to be much higher than the cost of producing the FAB. Once everything works, and you know how to produce those production line machines you just have to mass produce them, put them in a factory building and you're done? How can such a FAB cost 7 billion dollars?

I know things are always more complicated than they original seem, but still...

 

[Imouto]

Long over and settled. And not even remotely related to why Intel is doing well process-wise.

If I drive you out the circuit in a car race rendering your car way slower than mine you won't keep the pace even if I get a stop and go penalty. If I do it every single lap you can guess the outcome.