Intel outlines plans for next, next, next-generation Atom chips

1. Brad,

   Thanks for the info.  I have several quibbles though.  Process tech only increases processing performance because they can cram more transistors into a smaller space, and each then uses less current to operate, allowing for more advanced designs, and possibly more transitors to be crammed into the same space.

   I doubt that’s the direction Atom will head though.  Intel already has a performance line of chips.  They will need to be more agressive with Atoms performance in the future, because ARM and AMD have architectures which will actually be performing against it.  That said they will almost certainly never uncork the hardware and let it run at it’s most optimized best, because then it would start eating into the rest of their chip market, canabalizing sales on presumably more expensive chips.

   Also this announcement is somewhat confusing.  As far as I knew one of the points of Atom was to have a chip that was cheap to make, and could use some of the production capacity on older process nodes that Intel had and was under utilizing.  14nm in 2013 is going to be competing with their 2013 i-core processors (or whatever they’ll be called then) for production capacity…  Or ahead of the mainstream processors…  Since 2012 is the move to 22nm on Ivybridge, so the move to 14nm on main stream processors is…  2014 (or there abouts).

   So they’re blowing up the tick/tock schedule they JUST announced for Atom as point A.  Point B, it doesn’t make much sense given their own internal tick/tock cadence with their other market segments.  Point C, this would eat into production for mainstream processors right when they needed it.

   The only way this makes sense is if they’re saying that Atom chips will be the first onto 14nm with Core’s coming later, or they’re throwing the tick/tock process which has really worked out well for them, out the window.  Making the Atom chips first isn’t how businesses usually like to do these things since it would:  Steal the hype for the later mainstream processors, and would do little to recoup the massive expense of moving to the process tech, other than let them work out the production kinks…  However they’d be doing that on a very low margin high volume chip, which again doesn’t make any sense at all. 

   I understand that processors released in 2012 are manufactured in the end portions of 2011, so in a technical sense come 2013 they could theoretically be making Atoms on the processing node for 2014, but it just doesn’t seem likely, especially for products using those processors to come to market before 2014, but maybe this is technically true.

   I don’t know.  There’s a lot wrong here, on a lot of levels.  So with all that being said, I have to call BS on this.  Something isn’t right.

   1. Yup, they’re killing tick/tock and moving Atom to the same scheduled as the Core chips. 

      While Atom started as a way to make cheap, low power chips using last year’s tech, demand for low power chips for phones, tablets, set top boxes, netbooks, and other devices has blown up in the last few years and if Intel doesn’t want to lose more market share to ARM-based chip makers, the company needs to make some changes. 

      I guess we’ll know in a few years whether these are the right changes.

      1. Crazy.  Well thanks for the update.

2. This is all marketing ploy which you and me, the consumer fall for. If they want, they can even come out with the 14nm with optimum cost and productivity loevels tomorrow itself.

   We are all fools, falling prey to the marketing gimmics of chip manufacturers.

   1. Actually they can’t.  They are still working to perfect the 3d 22nm system for production runs, although from all accounts they should be volume production within months.

      14nm is a different beast, and is very much pushing the limits on what we can do with standard lithography.  They’re still working to perfect SRAM samples, which are always the first thing generated on a new process since they’re very easy to test.  Beyond that they’re going to need another esoteric process leap to get smaller, the same way they needed the advances of immersion lithography and other gains to get as small as we are now.  Something like quantum lasers will be required, although it’s not yet sure what that will be since those aren’t working out as scientists first hoped according to my sources (namely journals and Ars Technica).

      That said, you are on point that this is functionally marketing FUD.  That said I, I don’t think it’s directed at consumers.  I think this is directed towards Intel’s customers and partners, as a sort of ‘keep carrying the torch, we’ll have a good product soon’ kind of deal.

      Keep in mind that Atom has not in any way hit the performance metrics Intel set out for it.  It was designed specifically for MID’s (a space redefined by the iPad into Tablets not to be confused with the Tablets MS was pushing a decade and more ago), and in 3 years they don’t even have a Tablet win (not really, some very expensive boutique Win7 builds aside), and ARM has several.  We were supposed to start seeing Atom smartphones with Moorsetown, didn’t happen, it was too power hungry, big, and hot.  Worse the market it did find it’s way into, Netbooks, has helped gut the low end and highly profitable ultra portable markets, and possibly even massively distributable server computing (one of the most profitable markets in existence previously) driving margins down even further for Intel’s OEM partners, and hence putting even more downward pressure on Intel’s own chip prices above and beyond that provided by AMDs increasingly competitive offerings.

      All of this has meant that the Atom chip has been as highly disruptive of a force as Intel initially intended.  However it’s just not competing against ARM like they wanted, but is instead disrupting their existing markets.

      So again, yes to this being FUD.  FUD for their shareholders to feel good, and for their partners not to jump ship to ARM now that Win8 will run on it, and looks to be coming along at least on schedule if not ahead.  

      I think this is much less FUD for consumers though, making your comments about people being idiots for chip manufacturing marketing kind of ring false.  First off most people (not geeks but, you know people who are happy describing themselves as Joe Sixpack, gah), don’t know anything about chip process tech to begin with.  Low priced electronics are usually on a 2-3 year upgrade path anyway so this doesn’t really impact that except to give people a reason NOT to buy next years Cedartrail, assuming they did know that stuff.  And honestly having a low powered, low performance chip that’s REALLY small isn’t THAT exciting. None of this amounts to a reason that would keep me from buying something like an iPad that I could use today and instead keep me waiting for 14nm Atom…  Goodness?

      1. You still can’t do everything that you can do with a x86 system on a ARM system like the iPad.  So there are other reasons why people would still use x86 systems and despite lower production costs ARM devices are still priced higher than netbooks.

         Now while ARM devices are developing rapidly, keep in mind though that starting with 22nm that Intel chips will start taking advantage of new features like their 3D Transistor technology that promises a bigger boost in performance and power reduction than the size reduction would otherwise provide.

         Intel is also planning a complete architectural rework of the ATOM line code named Silvermont, adding for example out of order processing, when they go 22nm.  So the difference would be much more significant than the upcoming Cedar Trail that doesn’t change much besides adding a beefier GPU and higher clock speeds for lower cost and better power efficiency.

         For perspective though, you have to remember ARM offerings are under a pretty fast schedule themselves.  Companies like Nvidia for example will release a new more powerful version of their Tegra every single year.

         So while the iPad2 is significantly better than the original iPad, it mostly just catches up to what tablets can be offering this year from other companies like Qualcomm and their Snapdragon offerings.  Since we’re already seeing Smart Phones with dual core 1-1.2GHz and will soon see 1.5GHz.

         So even people who would go for products like the iPad have to consider what new better tablet is just right around the corner and is not just a competition with x86 solutions.

         Then there’s the quad core Tegra 3 that will be out later this year and it’s those next gen ARM offerings that will finally reach the CPU performance needed to properly run more powerful desktop OS and start to rival Intel ATOM CPU performance, which is the game changer that Intel is probably really worried about…

         1. Most people have only been aware of ARM products for a few years.  In reality, ARM started as an alternative to Intel in the spirit of offering ODMs a stronger lock on vertical integration, and that pedigree survives as its primary opportunity and advantage.  This is neither a good thing or a bad thing as much as it is a true thing, however, once you end up picking a side as either “consumer” or “product provider”, the balance of goodness and badness shifts considerably.  ARM has always done a good job of supporting its hardware partners to facilitate their time-to-market.  I’ve been involved in a number embedded projects in which the physical technology is just a small piece of the overall project.  It’s a little like throwing an event in which the food is just a small part of the overall needs, and in those situations in makes a lot of sense to call the caterer.  That’s what ARM products are like: catered meals.  It’s convenient for the host, and hopefully it’s delightful for the guests.  In contrast, Intel is like a Green Grocer.  They offer you the highest possible ingredients, but you have to bring the expertise if you’re going to make a meal.  To make use of this as an event host, you need to be supported by your own chefs and your own kitchen equipment.  MeeGo is one way that Intel is trying to accelerate time-to-market  for its hardware partners, and it’s sort of the equivalent of offering free recipes and cooking classes.  Although, you still need your own chefs and kitchens.  To that end, Linux itself stands as a repository of open source recipes and community supported cooking classes, and the reason that legitimate distributions of Linux and ARM don’t get on as well as Linux and Intel is that there’s only so much flexibility available in cooking when you’re ingredients are primarily pre-prepared food.  I think ARM is great, but I’m sticking with x86 for as long as possible simply because I know how to cook my own food and understand that it’s the only way to make sure that I’m eating a healthy, nutritional diverse and complete diet.  I’m certainly not going to eat ever meal of my life out at a fancy catered party, but when I throw my own party I’m calling the experts.

Comments are closed.