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

[Charles Kozierok]

Doing something first isnt a risk in itself. Doing something the others didnt consider or didnt want to=risk.

You keep using that word. I do not think it means what you think it means.

 

[Phynaz]

Doing something first isnt a risk in itself.

Sure it is. "Hasn't been done before" is heavily weighed in business.

 

[ninaholic37]

Intel used 10billion$ on IA64 to frogleap performance for the future. AMD used what, 50mio$ on a slap on x64 to keep us locked for another 30 years. Who won? Do you call it innovation? I call it stagnation.

Didn't Intel finish Itanium? Can't we buy the processor? I imagine if people wanted to use IA64 and get the software ball rolling with it, they would have. I wouldn't mind seeing it go places if it eventually made things better, but it seems clear that more people wanted x64 for convenience / backwards compatibility anyway. Since the IA64 technology is already out there though... I suppose Windows/software/Intel could pick it back up anytime if they thought it was viable/beneficial to do so. Right?

In my view I dont see it that way. For me a risk needs to be a game changing event. Something to set a true difference. HKMG, trigates etc is just a natural step on the process manufactoring development. If Intel had gambled on graphene, photonics etc. Then yes.

I hope they do. If anything, just for sheer knowledge/experience of knowing what graphene/photonics/etc. can really bring to the table. From what I understand, with a lot of effort, graphene chips could finally enable stable clocks of 10Ghz+ into the mainstream without overheating easier. Is this true? Which company is willing to take the risk to find out (I read that Samsung is already attempting graphene in the mobile space)?

 

[Idontcare]

Discrete, IDC, discrete. Discreet is how you handle your mistress.

I fail at grammar/spelling

 

[Cerb]

Intel used 10billion$ on IA64 to frogleap performance for the future. AMD used what, 50mio$ on a slap on x64 to keep us locked for another 30 years. Who won? Do you call it innovation? I call it stagnation.

That's because it was never going to happen, because Itanium has generally sucked, compared to our commodity CPUs. Support for it hasn't waned because of x86. If that were the case, PPC support would also be a problem. The difference is that low-end PCC are still good, and Power still keeps kicking ass. IA64's greatest success was killing off PA-RISC and Alpha, and marginalizing SPARC, at a critical time, when many now-old series of computers were more or less dying off. IBM and Intel ended up the only strong survivors, though Sun/Oracle and Fujitsu still keep popping their heads up.

The goal of competition and capitalism is monopolies.

The goal of competition within capitalism is monopoly, and Intel is very good at working it. No big company truly values the need to innovate (provide more output with less input) that customers having options necessitates.

A duopoly is more damaging than a monopoly. If you wanted competition we would need much more companies competiting. But the ROI aint possible for that.

More competition would also be a scary idea to any representatives being lobbied to, when it comes to silicon-industry special treatment, as they thrive on a similar duopoly system.

 

[TuxDave]

I don't see why not.

It was a completely new technology and could have cost them billions of dollars and been a complete failure.

That's what test chips are for.

 

[IntelUser2000]

If Intel had gambled on graphene, photonics etc. Then yes.

The definition of "radical" all depends on perspective. I view "radical" technologies as merely enablers to continue progress. Early day cars were not viable against horse carriages, but that allowed futurists back then to imagine a different era.

Tri Gate is not as "radical" but still a big one. But it's still merely enabling the industry to continue progress on semiconductors, aka Moore's Law. I expect the same will be true for Graphene, Carbon Nanotubes, Photonics.

Initial results of Tri Gate were also worse off then cutting edge planar technology. They had to improve it to get to the point they have now. Same is true with others.

 

[pelov]

Initial results of Tri Gate were also worse off then cutting edge planar technology. They had to improve it to get to the point they have now. Same is true with others.

Was that the reason for the IB delays? I know they had yield issues (or at least that was the rumor) but I don't think Intel admitted anything publicly other than that it'll be late.

FinFETs/Tri-gate are a natural progression and not necessarily a "risk." It was something that had to be done sooner or later and it just so happened that Intel got there first so they were the guinea pigs (maybe not first technically as FinFETs have been around a long time, several years actually, but Intel's the first chip designer/fab to have produced 3d-gate tech at large quantities for consumers). Their fab advantage is huge and can't be overstated, particularly because their only competitor has far less money for R&D and now relies on an independent fab(s).

I think the most interesting thing would be to ponder just what truly radical advancements have been put on the shelf. Graphene maybe?

 

[CHADBOGA]

Sure it is. "Hasn't been done before" is heavily weighed in business.

Love the sig.

 

[IntelUser2000]

Was that the reason for the IB delays? I know they had yield issues (or at least that was the rumor) but I don't think Intel admitted anything publicly other than that it'll be late.

I believe the reason for Ivy Bridge delay was it was too much of a change. They would have been safe if both the CPU and Graphics had minimum of a change, or it was in an MCM like Arrandale/Clarkdale.

But they really had no choice and had to significantly change the iGPU, so it became a Tick on the CPU and Tock on the GPU. It's impossible to know if the situation would have been different if it was a true Tick and graphics were minimal advancement. There's no way to isolate whether its explicitely process related or whatever I mentioned, because Ivy Bridge is the lead vehicle for the 22nm process.

FinFETs/Tri-gate are a natural progression and not necessarily a "risk.

Hypothetically if traditional scaling never disappeared, we wouldn't have seen CPUs implement Strained Silicon, HKMG, and Tri Gates. I bet by the time we get whatever we think is "revolutionary" in real products, there would have been so much advancement on "less revolutionary" products that it might not be that big of a change.

Everything has a damn trade off that's in proportion to its benefits.

 

[Obsoleet]

Wot?

ARM =! Apple

Windows is the de facto standard for business software, and even if Microsoft utterly fails at everything over the next decade, they will likely continue to dominate the business computing sphere.

On the consumer end, Microsoft will become less relevant imo, but for that matter so will Apple as the products get better and cheaper from the competition.

ARM is also utterly terrible for certain tasks while being very efficient and effective at others. It's just a side effect of RISC architechture.

Apple is a major force in ARM's drive. Apple won't become less relevant over time, but more relevant. The competition will not produce better products- they are actually still going for the cheaper/more effective route. No one's trying to compete on the quality/high end with them, no one is even stating they are. Even Intel is saying "we're going cheap" on ultrabooks. http://www.tomshardware.co.uk/intel-ultrabook-Jason-Chen-Taiwan-Plastic-chassis,news-38475.html

It's too much work for a company to do what Apple has done: put out quality product, stick with the level of quality, and build a strong brand around said quality. That's the best business model- and the longest term annoyance for investors to wait for to happen. You need a dictator like Jobs to push it down everyone's throat, but once you're there no one wants it to stop. Mark my words, Microsoft's weak Windows8 front will prove ineffective.
It's always a day late and dollar short for the Android and Windows teams, if they don't have quality product out yet- they won't soon and are too late. Kindle/Nook are probably the best known Android tablets and are a mess, totally fractured from the rest of the Android world. The lack of control Google exerts over Android is damaging it- and the lack of well supported quality products is exactly why Google bought Motorola Mobile. http://www.tomshardware.co.uk/Google-Motorola-Acquisition-Deal-Final-Closed,news-38625.html

I also wouldn't put too many beans in the GPU/APU basket, Intel is headed there pretty fast, but in the very areas you're talking about, it's ironically ARM-powered devices that are replacing the units (think of how many folks pick up iPads or Galaxy tablets and use them to run dumbshit games and such on them rather than get a real laptop with a real GPU or Fusion/Llano/etc).

The APU/GPU basket is a good one, and one unaffected by the rise of ARM. GPUs are being sold in high quantity to number crunching firms, science and industry. APUs are going to be increasingly relevant as they offload the FP to the GPU and more applications are APU accelerated.
GPUs and ARM are two different markets. Both chipping away at x86, which is increasingly relegated to the business world but I can attest that many businesses are moving towards the x86-dependent world- using processor agnostic apps like GoogleDocs (supported in the enterprise) and using Apple devices (essentially cloud machines themselves with just enough local firepower in the age of excessive CPU power).

I'd put your money against MS and Intel and towards Apple- maybe Intel if Apple blesses their beggings. http://www.latestmobilesinfo.com/in-apples-future-intel-wants-to-be-the-mobile-chip.html

 

[2is]

errr... Apple uses "cheap" Intel CPU's in their Ultrabooks (MBA) and that's not changing anytime soon. They even toyed around with using ARM and abandoned the idea because quite simply, ARM sucks when you actually need processing power.

There's a big difference between powering a tablet/phone vs a laptop, much less desktop. ARM isn't going to chip away at that segment for a long time, if ever...

Heck, I own and like apple products too (have a 3rd gen iPad, iPhone 4 and a MacBook Air) but you really need to ease up on the koolaid.

 

[Obsoleet]

errr... Apple uses "cheap" Intel CPU's in their Ultrabooks (MBA) and that's not changing anytime soon.

Read this-> http://online.wsj.com/article/BT-CO-20120517-712450.html
I put a link in my post as to what the cheap comment is referring to. It's not the CPUs. Ultrabooks isn't the growth market today either way.

 

[dagamer34]

You guys really need to read the Innovator's Dilemma by Clay Christensen.

For the most part when dealing with effective monopolies as with Intel, the only way they can be "taken down" is by actually marginalizing their business to make it irrelevant, basically a "disruptive innovation"

A few examples:
1) Digital photography killed film photography, along with Kodak.
2) LCDs replaced CRTs.
3) GPS units replaced traditional maps.
4) Smartphones destroyed numerous markets, including GPS units, point and shoot cameras,
5) Wikipedia killed traditional encyclopedias
6) Digital downloads/streaming are killing CDs/DVDs, which killed floppy drives.
7) SSDs are slowly replacing magnetic HDs.

Source: http://en.wikipedia.org/wiki/Disruptive_innovation#Examples_of_disruptive_innovations

Intel is never going to be taken down by AMD, but by ARM licensee holders. TI, Nvidia, Qualcomm, Apple, and Samsung all have a combined R&D budget that is greater than AMD could ever hope for, and the future is clearly mobile which Intel isn't a strong player in because they rested on their laurels in x86 desktop/laptop/server land for so long (which, if anything, will be their downfall).

True competition will come from the fact that more and more people, though still a small number, are using iPads instead of traditional x86 PCs, and THAT is what would keep Paul Otellini up at night. It is not pure coincidence that as soon as the iPad came out, Intel started really targeting low power chips, especially with the promise of Haswell.

 

[Arachnotronic]

Intel is never going to be taken down by AMD, but by ARM licensee holders. TI, Nvidia, Qualcomm, Apple, and Samsung all have a combined R&D budget that is greater than AMD could ever hope for, and the future is clearly mobile which Intel isn't a strong player in because they rested on their laurels in x86 desktop/laptop/server land for so long (which, if anything, will be their downfall).

True competition will come from the fact that more and more people, though still a small number, are using iPads instead of traditional x86 PCs, and THAT is what would keep Paul Otellini up at night. It is not pure coincidence that as soon as the iPad came out, Intel started really targeting low power chips, especially with the promise of Haswell.

Intel's been lazy on the mobile front, but they're getting in gear. Medfield is a good start -- a chip that uses an old-as-the-hills CPU core. When Intel rolls out "Silvermont", in which they have the chance to do a brand new architecture, and when Intel puts their mobile lineup on a "tick-tock" cadence, I think Intel will be well positioned to lead mobile. Why should phone vendors deal with guys like Qualcomm or NVIDIA, who are subject to third party, inferior manufacturing technology when Intel will supply all the chips you need in mass quantities? If Intel actually hits one out of the park with a new Atom core, it could very well crush ARM offerings in all dimensions: power, performance, and price.

 

[pelov]

Intel's been lazy on the mobile front, but they're getting in gear.

They are but it's at a higher cost than the competing architectures. Intel isn't exactly known for producing cheap chips. In fact they've built their entire business on maximizing profit margins. This works great when your biggest and only competitor on x86 is 2-3 years behind but it's Intel that's playing catch up in mobile and not the other way around. Competing on phones/tablets means working on razor thin margins and relying on quantity to make up for larger profit margins. It's also far more cutthroat than x86 because of just how many players there are. A single misstep against TI/Qualcomm means they leap ahead whereas Intel can afford 2-3 missteps on the desktop and still be ahead.

Speaking of Medfield, how much does that chip cost?

 

[Charles Kozierok]

A single misstep against TI/Qualcomm means they leap ahead whereas Intel can afford 2-3 missteps on the desktop and still be ahead.

Okay, but isn't that a good thing?

Many people have been concerned that Intel has no competition in the x86 market, and that they would become an overgrown behemoth. What you're (correctly) surmising is that there is still plenty for them to compete against.

 

[IntelUser2000]

They are but it's at a higher cost than the competing architectures.

The mid-range Z2460 phones are supposed to be in $400 range unsubsidized.
The low end Z2000 phones are rumored to be $200 unsubsidized.

So that looks like it'll be competitive, as almost all initial phones based on it are largely based on reference designs with everything packaged.

The question now turns then are the prices enough for Intel to maintain satisfactory margins? They don't pursue without financial merit. Maybe selling it as packaged via reference design changes the equation.

 

[pelov]

Yes, but it also requires Intel lowers their cost.

Holding Microsoft's hand isn't helping as that's another unnecessary tax added on to a device. Tizen is an Intel/Samsung solution to the Google-free Android and moves Intel away from MS -- which is a great thing if you've seen the condition of win8.

Stiff competition means low pricing. Unless Intel is willing to quit relying on brand recognition for higher margins they'll get nowhere in the mobile market. They've got to price their chips like the ARM manufacturers price theirs. It's a win/win situation for consumers but it'll require Intel approaching the market in the complete opposite way they do their x86 desktop/server/laptop business.

The mid-range Z2460 phones are supposed to be in $400 range unsubsidized.
The low end Z2000 phones are rumored to be $200 unsubsidized.

I saw this too but you can't judge by the pricing of the phone itself. Intel obviously has a say in what the phones will look like but if it means OEMs have to skimp on screen quality or build quality to price match then it's a problem.

 

[IntelUser2000]

They've got to price their chips like the ARM manufacturers price theirs.

The pricing isn't unfeasible for Intel. The Netbook Atom chips have an ASP of slightly over $25. If they can get to premium status, then premium prices can be charged.

For pricing the Lava Xolo X900 goes for ~$420. You can judge from there.

 

[Ferzerp]

The whole ARM thing is funny. It's not really an evidence backed stance.

Certain people who style themselves experts have decided that it is inevitable, and then try to find evidence that supports the view that they formulated without evidence. Most of the analysts making these claims (and people parroting them) have very little understanding of what computing is. They see a shiny toy (ipad) see that it uses an ARM processor, and think "this is the future".

These models always assume many, many things, the most amusing of which is that Intel is stagnant. Where does all that R&D money go if they're stagnant?

Intel can back in to the ARM space much easier than anyone using ARM architecture can hope to grow in to Intel's space. Intel has (and appears to be able to for the forseeable future) a huge advantage in their ability to manufacture smaller, denser chips. Smaller chips are cheaper per unit to manufacture and they require less power.


We've been hearing about the death of the PC from analysts for how many years now? Well, if you look, PC shipments are still growing (not by much, and not by as much at the toy tablet market). There is a flawed assumption that what I'm calling toy computers increase in sales must come from a decrease in traditional PCs. So far, that just isn't the case.

 

[alyarb]

You can combine TSMC, UMC, and Samsung productivity and you won't come near a million wafers per month. And that's just wafers, not talking about 100% usable SoCs which must be a small fraction of that, and it's probably going to stay that way. intel has an empire of fabs. Even if their process wasn't 5 years ahead of the rest of the world, they still have better facilities and staff, in greater numbers, doing the work. so even if all intel SoCs were inferior to ARM SoCs, intel at least has the capacity to deliver as promised. TSMC does not.

Silvermont doesn't even need to be great. They can leave performance the way it is and use all the TDP headroom from 22nm to reduce power. They could even sell silvermont at a loss just to put everybody else out of business, and then make all that money back on the next node.

intel definitely beats the other fabs on quality, volume, and punctuality, and for most manufacturers, that is way more important than the relative performance of whatever they're buying.

 

[Golgatha]

You guys really need to read the Innovator's Dilemma by Clay Christensen.

For the most part when dealing with effective monopolies as with Intel, the only way they can be "taken down" is by actually marginalizing their business to make it irrelevant, basically a "disruptive innovation"

A few examples:
1) Digital photography killed film photography, along with Kodak.
2) LCDs replaced CRTs.
3) GPS units replaced traditional maps.
4) Smartphones destroyed numerous markets, including GPS units, point and shoot cameras,
5) Wikipedia killed traditional encyclopedias
6) Digital downloads/streaming are killing CDs/DVDs, which killed floppy drives.
7) SSDs are slowly replacing magnetic HDs.

Source: http://en.wikipedia.org/wiki/Disruptive_innovation#Examples_of_disruptive_innovations

Intel is never going to be taken down by AMD, but by ARM licensee holders. TI, Nvidia, Qualcomm, Apple, and Samsung all have a combined R&D budget that is greater than AMD could ever hope for, and the future is clearly mobile which Intel isn't a strong player in because they rested on their laurels in x86 desktop/laptop/server land for so long (which, if anything, will be their downfall).

True competition will come from the fact that more and more people, though still a small number, are using iPads instead of traditional x86 PCs, and THAT is what would keep Paul Otellini up at night. It is not pure coincidence that as soon as the iPad came out, Intel started really targeting low power chips, especially with the promise of Haswell.

According to the same theories though, Intel will brush up against the "good enough" point in desktop CPUs, and this will give ARM/AMD/etc a chance to disrupt them. Customers will not pay high margins for premium products forever.

 

[pelov]

We've been hearing about the death of the PC from analysts for how many years now? Well, if you look, PC shipments are still growing (not by much, and not by as much at the toy tablet market). There is a flawed assumption that what I'm calling toy computers increase in sales must come from a decrease in traditional PCs. So far, that just isn't the case.

Desktop sales are growing, but an overwhelming majority of those are in Asia and tend to be very cheap desktops for the newly emerging economies. Profits on those are less than stellar. In the US it's the laptops that are growing and now represent an overwhelming majority of PC sales. So PC sales are improving, if you look more closely at the numbers you'll see it's the laptops that make up a whole lot (75%?) of sales.

The desktop is growing, but a very measly rate when compared to tabs/lappies/smart phones. So the trend is actually still heavily favoring mobile and that only seems to be accelerating.

 

[Khato]

Silvermont doesn't even need to be great. They can leave performance the way it is and use all the TDP headroom from 22nm to reduce power. They could even sell silvermont at a loss just to put everybody else out of business, and then make all that money back on the next node.

No no, can't be selling product at a loss else those pesky regulators will take notice and actually have a valid antitrust case to bring against Intel. What they can do is sell at minimal profits though, which will have roughly the same effect as Intel has both a lower cost structure (other companies likely would have to sell at a loss in order to match pricing) and many other revenue sources.

Regardless, all the other SoC players should be quite concerned regarding Silvermont. I'm still expecting that it'll hit approximately core 2 duo IPC levels, but probably will only hit around 2GHz turbo in order to keep power low. If it does beat A15 (again, my expectation is that it easily will) then it'll be at least a year before there's any possibility of competition as Qualcomm's S4 successor should show up in 2014... but by that time it'd be going against the 14nm airmont. And it likely won't be until 2015 that we see the successor to ARM's A15... The possibility of having both a performance and power deficit against a company that has the capacity to fill the entire market with their design - I know I'd sure be nervous.