back to article Intel to put pedal to metal in 14nm Atom upgrade

When the Intel Developer Forum (IDF) kicks off in San Francisco on September 10, Chipzilla may announce that it will radically accelerate its development of 14-nanometer Atom processors. So says a report by Barron's, citing "a person close to Intel" who told the financial news service that the chipbaking giant will trim its …

COMMENTS

This topic is closed for new posts.
Silver badge

Fab 42

From Jimbo's emporium of knollij: On February 18, 2011, Intel announced that it will construct a new $5 billion fab in Arizona, designed to manufacture chips using 14 nm manufacturing processes and leading-edge 300 mm wafers. The new lab will be named Fab 42, and construction will start in the middle of 2011. Intel billed the new facility as "the most advanced, high-volume manufacturing facility in the world," and said it would come on line in 2013.

1
0
Anonymous Coward

Re: Fab 42

That's great. Now if they would just use it to make 14 nm ARM chips, they might sell a few...

3
2
Anonymous Coward

Re: Fab 42

Gotta laugh at Chizilla so obviously shitting bricks. Don't remember it happening for quite a while. Not since AMD pulled the 64bit rug out from under Itanium? Good to see a bit of healthy competition galvanising Intel's resolve again.

Methinks it might already be to late though: The phone market seems close to saturated and moving rapidly into a pricewar mode. Same for tablets. Hard to see a premium x86 part suddenly making huge inroads into that sort of market. Niche "premium" products like MS's next surface no doubt but their "premiumness" will almost certainly continue to guarantee their nicheness.

The server racket's another matter though. May well help Intel fend off the ARM advance into servers for a while longer. I wonder how the performance/power/price metrics will compare.

But what next? Won't 14nm leave only one more shrink in Intel's arsenal? Or is there an ace lurking up Chipzilla's sleeve?

2
2
Anonymous Coward

Re: Fab 42

"May well help Intel fend off the ARM advance into servers for a while longer. I wonder how the performance/power/price metrics will compare."

I wonder that too.

I wonder specifically because I'd love to know how much of a datacentre power budget is influenced by the CPU chip in use, rather than (say) storage, connectivity, etc. Any references around?

Whatever ARM (and partners) do, they're not going to significantly change the datacentre power budgets for storage, connectivity, etc. If they have much of a power/thermal impact on servers (and on the corresponding cooling) I'll be pleasantly surprised.

Desktops etc are a different ball game.

1
0
Anonymous Coward

With Intel not really having any competition on the desktop/server side of things, maybe they should introduce their mobile (Atom) chips first on the new process? Find any issues with a lower volume part and they still maintain the lead in process size over their competitors.

1
0

Competition is a wonderful thing

Sounds like Intel are all out of ideas and now left only with process improvements to compete with ARM - that may work initially but ARM and their partners won't be hanging around so how much time does this buy Intel, a year of competitiveness? Maybe 18 months.

And then there is still the issue of unit price - Intel being way more expensive than ARM though the move to 14nm should help here - and lack of SoC customisability. So still a few hurdles before Intel can really compete with ARM, but I'm sure Intel will achieve some 14nm tablet/smartphone design wins by hook or by crook... they know they have to.

2
1
Silver badge

Re: Competition is a wonderful thing

It's common to see the assumption thrown about here that Intel on mobile is uncompetitive, slow, expensive etc. But honest question - why are we seeing Android tablets now based on Intel, including flagships like the Samsung Galaxy Tab 3?

Unlike a Windows (or Linux or OS X) device, there's no reason to do so from a compatibility point of view, on the contrary, it's a slight downside as some Android applications won't run on Intel.

0
1
Anonymous Coward

Re: Competition is a wonderful thing

"why are we seeing Android tablets now based on Intel, including flagships like the Samsung Galaxy Tab 3?

Unlike a Windows (or Linux or OS X) device, there's no reason to do so from a compatibility point of view, on the contrary, it's a slight downside as some Android applications won't run on Intel."

Are you aware that Intel have a history of paying market-leading customers to use their chips rather than the competition's? Obviously they wouldn't be doing that again, new management etc.

Anyway, best-known might be the Dell-Intel deals as reported here and elsewhere e.g.

http://money.cnn.com/2010/07/27/news/companies/dell_settlement_intel.fortune/index.htm

"Though last week's settlement between the U.S. Securities and Exchange Commission and Dell was anticipated -- the computer maker admitted no wrongdoing and agreed to pay about $100 million in fines -- it was nevertheless a stunning event both in terms of the charges it leveled against Dell and those it reinforced against semiconductor giant Intel.

Amid accusations of both earnings manipulation and apparent antitrust violations, the SEC paints a picture of the computer industry in which the near-monopolist Intel (INTC, Fortune 500) is the master puppeteer and the computer makers are little more than its marionettes. When Intel rival AMD (AMD, Fortune 500) first propounded this dark view of the world in an antitrust suit against Intel in June 2005 it seemed like a wacky conspiracy theory; now it has been endorsed by regulators for 30 countries, including ours.

In its complaint, the SEC alleges that, from May 2001 through January 2006, Dell (DELL, Fortune 500) created the false impression that it had met or exceeded analysts' consensus earnings-per-share expectations in 20 straight quarters. In reality, says the SEC, Dell wouldn't have met its numbers once during that period without secret payments from Intel that were made in exchange for Dell's agreement not to use any AMD chips.

Thus, the SEC has now become the sixth regulatory body worldwide -- and the third in the United States -- to conclude that Intel made improper payments throughout much of the last decade to persuade computer makers to bar or sharply limit their use of AMD chips. Though these payments allegedly began in 2001, under Intel's now retired CEO Craig Barrett, the practice allegedly continued and expanded under Intel's current CEO, Paul Otellini, according to the SEC.

Intel has always denied making such payments" (article continues)

1
1
Silver badge

Re: Competition is a wonderful thing

"Are you aware that Intel have a history of paying market-leading customers to use their chips rather than the competition's?"

But the original claim included "And then there is still the issue of unit price - Intel being way more expensive than ARM", so which is it - is it more expensive than ARM, or less so?

Yes, it may be that it's only less expensive because Intel are currently subsidising it; it may be that Intel have higher prices for some customers and lower prices for "market-leading" ones; but that's fair game for Intel to do - the point is, we're looking at it from the point of view of those market-leading customers. I don't see how you can simultaneously claim Samsung are better off with ARM because it's cheaper, whilst saying they only use Intel because, er, it's cheaper. The alleged wrongdoing of Intel you reference is more about collusion and antitrust issues (which perhaps would apply far less in a market where they aren't dominant?) - in principle, if a company says "Our TVs cost £1000 but we'll pay you £500 to buy one" is no different to "Our TV costs £500"; the issue is more dodgy deals and collusion to limit competition.

I can accept that maybe Intel are only getting somewhere in mobile because of offering a lower cost to companies, but they can't then be criticised for being more expensive.

So now that we've established the price issue isn't true, how does performance compare? Again, an honest question - I've long wanted to see benchmarks comparing Intel to ARM on CPU and GPU, but I've so far only found charts that compare within each family. I realise benchmarks can be misleading, but is there any attempt to come up with a rule of thumb for comparing? (E.g., it's possible to have a rough idea of how Intel and AMD CPUs compare, or how Intel, AMD and NVIDIA GPUs compare.) That I'll probably get voted down for merely asking the question rather than given evidence doesn't gives me even less confidence on these claims...

0
0
Anonymous Coward

Re: Competition is a wonderful thing

"I can accept that maybe Intel are only getting somewhere in mobile because of offering a lower cost to companies, but they can't then be criticised for being more expensive."

Please consider the difference between cost for chip vendor to build, cost to system vendor to buy, etc.

It's perfectly possible (and probably even legal in the phone market) for Intel to sell chips to Samsung for less than it costs Intel to build them. But it may not be sustainable for very long. Intel could do it if they wanted to because they have plenty of cash in the bank. They might do it if they needed to encourage a lead customer or two to seed the "Intel in phones" market. Intel need a success or two outside their legacy IT marketplace. See where all that leads? Selling chips at very low (negative?) margins in order to seed a market?

ARM and partners don't need to play that kind of silly games with subsidised prices. The integration options already available from ARM partners are way ahead of what's available from Intel, and mean that the system cost is likely to be more attractive on ARM than Intel for a while yet, where subsidies aren't involved.

"is there any attempt to come up with a rule of thumb for comparing? "

Think about what you want to use the "system" for, and find a relevant benchmark, perhaps an indepently reproducible one. Coremark is good from that point of view for 'core' number-crunching functionality. It's not a phone benchmark though, nor is it a GPU benchmark. Those aren't really my areas, and the benchmarks in those fields that I'm aware of aren't exactly agreed as independent yet, afaict.

Benchmarks measure what they measure. In order to avoid being duped by benchmarketing, it helps to understand whether what a given benchmark measures is being measured fairly, and whether what they measure is relevant to the matter at hand, e.g. any design/purchase decision you may be thinking about.

A benchmark which is easily reproducible (such as CoreMark) can be good for that - you can get the sources, look at them, build it and run it yourself. BigData SAP benchmarks and that kind of thing are less easily reproducible because of the costs (including expensive skills) involved.

"I'll probably get voted down for merely asking the question"

Not by me you won't. It's a perfectly fair question, one which should be frequently asked.

That any good to you?

0
0

ARM is not Intel is not ARM

ARM is a design house. Intel is a foundry and design house. Comparing ARM and Intel doesn't really make sense. What makes sense is to compare Intel processor design with ARM processor design. As someone who has worked with both Intel processors and ARM cores, both are excellent for the tasks designed; both design houses are top-notch.

The question that I see will be answered is whether having design and foundry people as one can produce a better (someone else can define "better") product. ARM has numerous independent foundries using their IP which has worked out quite well to date for ARM. However, as Intel starts moving into ARM arena, will the advantage of having chip designers being part of the same team with the foundry outweigh ARMs business model; which model will win out when it comes to creating greater power efficiencies coupled with increasing computing power.

In the end, power and efficiency boils down to who makes the best transistor ergo my money is on Intel even if ARM comes up with a better design.

2
1
Anonymous Coward

Re: ARM is not Intel is not ARM

However, as Intel starts moving into ARM arena, will the advantage of having chip designers being part of the same team with the foundry outweigh ARMs business model;

I think you'll find that ARM and their designers work very closely with their foundry/silicon partners, you'd be naiive to think otherwise.

In the end, power and efficiency boils down to who makes the best transistor ergo my money is on Intel even if ARM comes up with a better design.

Intel has more invested in process technology than anyone else, and are using that rather than better designs to achieve similar power efficiency of existing ARM designs built on larger nodes. It's likely Intel will hit the process wall before ARM, making ARM the better bet over the longer term. Any performance advances Intel makes as a result of improved process nodes are likely to be temporary, achieving mostly parity with existing ARM designs until ARM silicon partners catch up and move ahead once again.

The other thing is whether Intel can survive on the wafer thin margins found in the ARM marketplace - Intel have enjoyed fat margins for most of their existence which has funded their expensive fab development. That's not likely to continue with Atom, particularly if Intel sell at cost (or less) to achieve market share. On price alone it's difficult to see how Intel can realistically (and honestly) compete with ARM products, but given Intels history I fully expect them to find "a way".

2
0
Alien

Re: ARM is not Intel is not ARM

I recognise your point when it comes to the traditional processor market. However when it comes to the mobile and high power efficiency markets then SOC becomes very important and I believe that ARM will continue to have a real advantage there unless Intel make a big move to open up their fabs.

An Intel closed design SOC, no matter how good it is, will define the capabilities of a device. All that the downstream manufactures will be adding is a case so margins will be very slim. With ARM a company can design their own SOC, built out of IP blocks from multiple sources, and have a chance to differentiate themselves from the competition and charge a higher margin.

If Intel SOCs turn out significantly cheaper than sticking with ARM manufactures will have to go Intel (I'm not sure that is going to happen, smaller processes are getting VERY expensive) otherwise I think they will be reluctant to move.

2
0
Anonymous Coward

Re: ARM is not Intel is not ARM

"power and efficiency boils down to who makes the best transistor ergo my money is on Intel even if ARM comes up with a better design."

So the system architecture, instruction set, etc, is largely irrelevant?

The fact that an SoC from an ARM partner potentially includes far more than just the CPU core(s), which potentially requires mixed technologies not traditionally associated with Intel's x86-centric processes, is largely irrelevant?

Well the market's got news for folk like you.

x86 depends on Windows. Where there's no need for Windows, it's x86 that's largely irrelevant. Look around you. x86 computing has been irrelevant in volume terms for a few years outside the desktop and datacentre, and I don't currently see any way that Intel can change that whilst remaining as profitable as their x86 range has historically been.

Interesting times.

2
0
Anonymous Coward

Re: ARM is not Intel is not ARM

> So the system architecture, instruction set, etc, is largely irrelevant?

Yes, at least according to "Power Struggles: Revisiting the RISC vs. CISC Debate

on Contemporary ARM and x86 Architectures”

Emily Blem, Jaikrishnan Menon, and Karthikeyan Sankaralingam

University of Wisconsin – Madison (HPCA 2013)

http://research.cs.wisc.edu/vertical/papers/2013/hpca13-isa-power-struggles.pdf‎

1
0
Anonymous Coward

Re: ARM is not Intel is not ARM

"So the system architecture, instruction set, etc, is largely irrelevant?

Yes, at least according to "Power Struggles: Revisiting the RISC vs. CISC Debate"

That looks like interesting reading, thank you.

It covers the instruction set(s), and understandably concludes that it many cases there's not much in it, except at low power and high power extremes. I think many of us suspected that already, but it's always good to see objective analysis with references.

It doesn't seem (at a quick glance) to say much what goes on the SoC itself and what goes elsewhere (they make no claim to address that?), which is part of what I meant by "system architecture". In many cases that's going to be far more important than the ISA, and is where Intel have so much to catch up on. If they even can.

0
0
Silver badge
Thumb Down

Naturally, it will come with a new socket

requiring all the 'i' fans to buy new mobo's

gotta keep the market going somehow...

for that reason it gets a --->>>

2
0
Silver badge

Re: Naturally, it will come with a new socket

Atom chips don't use sockets.

They are soldered directly to the mainboard.

Atoms are primarily used in portables and embedded devices - ease of upgrade isn't such an issue there..

2
0
This topic is closed for new posts.

Forums