Intel CEO predicts DOOM for fab industry and competitors The semiconductor industry is at a major inflection point Intel's CEO Paul Otellini predicted today at the company's analyst day, with the increasing cost of manufacturing causing a shake-out among the biggest chip players that, he said, would only leave two or three companies at the leading edge of chip design. Intel veteran …
Actually I have a feeling that Intel smartphones may do well in Canada and Russia. With the brutal winters in parts of those countries having a pocket warmer that can make calls might be useful. Then again having to carry 4 batteries in order to make it through a day may limit sales somewhat.
Intels Medfield platform benchmarking seems fairly competitive with e.g. a Galaxy S2 type phone for performance and power - where does your 4x statement comes from? No reason to think that a more up to date design plus the 22nm process shrink won't yield a very competitive product to ARM based SoC within a year or so.
The issue is, nobody gives a f about x86 compatibility on tablets and smart phones. Nobody includes Microsoft this time, that is the horrifying part for Intel.
Big Blue, if convinced there is a market can easily move to 22nm. The "they couldn't give us 3ghz cpu", remember that? That was "they wanted out of consumer cpu so they didn't care" since they moved to 4.7 (!) ghz months later.
So if I am reading this right he thinks that atom chips will become the biggest thing in mobile devices, thats phones and tablets.......
What OS is it they will be running ?
If android then I expect Arm to stay number one for many years to come as Intel are still trying to catch up with Arm moving ahead all the time
If he is banking on Windows then he is dreaming, no matter how good it is the ipad will be the number one tablet for several more years even if outsold overall by android, and as to phones windows is too restrictive on hardware spec.
Ultrabooks are just overpriced laptops that lets be honest unless they are at least £200 cheaper will not outsale the Air purely on the "I want one" factor.
Yes as a company they are ok for a long time but as to some of what he says is pie in the sky thinking.
why does the os matter, most of the code will be written in a high level language with the board support package have to be written for most chip change anyway. If apple see an advantage in using atom through a combination of cost, power, speed, size I'm sure they would use it to maintain market dominance.
The OS and its architecture can have a big impact on power dissipation depending on how it handles task switching and idle operation.
Also one argument for x86 is the huge (but often ugly and unsupported) "wintel" legacy of code that businesses rely on.
So far MS has failed to succeed on non-x86 (they supported then unceremoniously dropped Alpha, MIPS, etc with NT4) so the new WOA (WinRT) on tables is therefore a big gamble. However, Apple has made a successful transition from PowerPC to x86 for the Mac, and conceivably could do it again if ARM became attractive enough in cost/power/performance (given the success of iOS on ARM that is not unreasonable). Similarly Linux runs on practically everything.
So the OS choice matters in that Intel's success was tied to MS' past glory, and the x86 is a horrible design that no one in their right mind would *want* to use, so outside of MS' old PC ecosystem there is little reason to favour x86.
Most of code on os x is cpu independent by nature. It is merely an "arm" option in xcode upgrade to ship same app for a future Mac air. Funny part is, user wouldn't even notice unless he/ she is technical since universal binary exists. It is way ahead of win, even Linux.
These things must be really making Intelddsa afraid and same time, keeping arm innovative as same terms exist for them too. As they talk about drivers, guess what? They are universal too and they too are written in a special dialect of c.
They're moving to 45cm wafers? And at 22nm? That's a lot of chips per wafer. Wonder what the yield figures are; whether this isn't simply a brute force approach to too many broken chips in production. Just too bad they're not doing too well with the other side of chip innovation. How's itanic, these days?
The thing is of course that it's becoming increasingly clear you don't really need all that processing muscle for day-to-day computing needs. An ARM core or two like in a recent smart phone would do pretty well on the desktop, too. Provided your software is halfway efficient and doesn't waste most of the cycles available on eye candy. We might even see people attach a keyboard and mouse to a tablet, stick it in a stand, and write emails, letters, do homework, that way. Leave the keyboard at home but take the tablet. Another upshot: No active cooling, so not as noisy as a laptop. Even atom still runs too hot to fit this purpose.
It might be that intel simply isn't making enough headway fast enough in the lower end of the computing spectrum to not end up in a slowly contracting but, of course, for the foreseeable future still rather large niche. Should, say, ti manage ARM cores in something comparable to 22nm tri-state, intel will have a serious problem, even if they're two nodes ahead by then. And though I actually know very little about the economics involved, I wouldn't be surprised if it turns out that 45cm wafers aren't necessary to stay competetive if the chip itself is small enough. The high end are monsters, but if say only a quarter of the same chip real estate is already enough for a nice product with a good margin....
Which is to say, intel has some very, very nice cards in its hands, but not all its cards are on an equal footing, not quite. So the competition needn't fold yet, which I'm sure they won't do.
Yep Intel's saving grace is their state of the art fabs. It allows them to charge the sick margins on the desktop and server that ARM only dreams of. Then again by going fabless ARM gets just about every other chip maker to do its dirty work against Intel (Samsung, AMD, TI, Freescale, IBM, etc).
Fabless is a great division both of labour and reward: licensees know that while ARM has a good margin it is delivering a good product at a fair price and gives them the chance to make chips for customers at a profit as opposed to Intel's winner takes all model, as PC makers know only too well. Should point out that AMD is no longer a chip maker - it outsourced that to Global Foundries and you missed TSMC from your list.
Nevertheless, at the same time as the commodification of chip design has significantly reduced the prices for such chips, the prices for the new processes and equipment sill continue to rise and there are fewer and fewer companies able to produce the kit (lithography, etc.) necessary to build the fantastically complex designs. It is a paradox that Intel's prowess in this area is driving up costs at the same time as chip costs are falling. Intel has made great progress even with the Atom. Reports suggest that the new phone chip really does come close to the TDP of comparable ARM designs and it has more oomph. But it costs a lot more to make and to put in a phone.
"...the new phone chip really does come close to the TDP of comparable ARM designs and it has more oomph"
How much of that is clever design (i.e specific to the x86 system) and how much of that is down to having a process technology a year or two ahead of the competition?
In other words, if you could get an ARM chip done with on the same Intel fab, would it then thrash it?
"...the new phone chip really does come close to the TDP of comparable ARM designs and it has more oomph"
>How much of that is clever design (i.e specific to the x86 system)
If true than it really would be kudos to Intel as the x86 instruction set is a dog all around especially for sub watt chips. Definitely a case where Intel succeeded in spite of the instruction set and architecture not because of it.
As for 45cm Intel would love to move to them for new fabs if it could but the issue is all its 3rd party partners. The tool makers took a bath on moving to 300mm for many years (had to do massive R&D, new tool software necessary, etc). The industry largely needs to get the majority of the tool makers, wafer makers, FOUP supplier, etc to jump all at once which is like herd cats many of which are going broke.
Paul Otellini technically just says what Intel shareholders want to hear. What he does address is the growing frustration of CIOs with proprietary platforms, like SPARC and POWER, which are, when equipped with enterprise grade memory, SAS and SSD hard drivres, HBAs, etc, 10-15 times more expensive, but not 10-15 times more performing. The advantage of proprietary systems in feature richness to ensure availability is diminishing by the day.
What he does not address is the solution: vendor independence, which also means independence from x86 architecture, and the fact that CPUs have become a commodity. Sure enough, there should be a standard, but the standard can be ARM.
I predict that in not to distant future, big corporations will send a wishlist to Intel, ordering a customized 256 core, 2048 thread ARM processor, or the like, maybe already soldered on a motherbard, next to the 4TB SSD chip. If Intel can produce 100K units of these, fine. If not, they will order from Samsung. Or TSMC. Or someone else.
Yes and no.
SPARC and POWER have two great advantages over Linux/Windows and x86/64: hardware compatibility. Linux, Windows and every other OS, application, and more that's built for that platform has to stand up against a dizzying array of hardware options.
With SPARC and POWER and ... you have one manufacturer of hardware and one developer of the OS (and generally a small number of developers for supplemental OSes). That means that your compatibility is higher. You also usually have *very* good service contracts with those people, so you can call out and have someone in your doorstep in 2 hours or less with replacement hardware, they're certified and know how to deal with it. Getting a response like that from a x86 vendor is hard, and costs a lot of money on top of what you buy.
One of the things I think that Linux needs to work on is hardware detection and interaction. Compared to Solaris and AIX, it sucks (and I say this as a Linux fan and user). Naturally some of this comes down to the earlier issue with fragmentation and large number of items to be compatible with et al.
that Intel and nVidia have a peace treaty, paying each other chump change for each others tech and expertise .. nVidia GPUs are showing up next to Intel CPU .. including on high-end intel main boards and in HPCs
nVidia's kicking butt in ARM processors and GPUs .. and Intel the CPU and systems to tie them together .. it's everyone else that has to play catch up on the 3 platforms, smartphone, tablet and laptop/PC
"We'll be top dog for years, Otellini tells investors" - of course, what CEO is going to say "we suck, we'll be bust in a year". CEOs...They're going to smile until the bitter end, it's their job. Well, that and blaming anything else.
"We have incumbency, legacy support, and the capabilities of Intel for scale and power - and yet you don't have the ability to deliver that to a device that'll run for a day off a battery the size of a two euro coin...
I notice he didnt say anything about AMD spinning off their fab into a separate company, nor anything about Oracle and SPARC, nor about IBM, much less about Toshiba. And the only real mention about ARM was some bullshit about an "uphill fight", while ARM has pretty much become the dominate player in low-voltage microprocessors for quite awhile now.
On the defense side of things, Harris also isnt going anywhere, nor is TI. Much less Qualcomm when it comes to mobile communications.
x86 also isnt going anywhere for now, but neither are the other players. It seems to me that every one of the semiconductor manufacturers with the exceptions of AMD, ARM, and Intel wound up specializing in markets where they're pretty much pseudo-monopolies.
This is one of the few times Id actually like to hear what Larry Ellison has to say about this subject.
Sounds like Intel just wants to spread some FUD to me.
For over 30 years they have been trying to convince consumers that they should over-pay and get less value than buying from AMD. It hasn't worked yet so I suspect Intel is in for a long uphill battle now that they can't freely bribe, intimidate, coerce and violate anti-trust laws to prevent AMD from selling better products at a lower price.
Ivy Bridge looks pretty bad compared to Trinity.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
