Intel first introduced their Atom based architecture CPUs 5 years ago. Back then, Intel wanted to make a low performance but low power consumption core for a developing niche of small size notebooks, namely Nettops. And thus began the Nettop revolution. Selling far more than Intel could have ever anticipated (nobody in the industry saw that coming), Intel sold Atom in millions and made loads of profit from the same (thanks to the chip having an extremely tiny foot print, yet having a relatively high cost). But the nettop revolution soon died out with Apple's iPad, and the tablet revolution followed. Since then Intel's Atom has been going more and more out of focus. And it had reached a stage where it was simply standing still, and stinking.
Intel didn't push Atom hard enough
Intel basically slackened up. They didn't not introduce architectural enhancements, process changes or anything new to the Atom line up, which was totally unlike what they were doing with their Core iSomethingmeaningless (thank you Charlie, semiaccurate) lineup. Atom kept getting revisions where more and more components got integrated into the chipset. Intel was moving to an SOC design, but it took them 4 years to move to 32 nm and they still haven't made any architectural advancements.
Enter 2012, Atom SOC for your phones
Intel's push to an SOC design is primarily because of the growing threat that ARM poses to the behemoth (and rightly so). And this year, we saw the first smartphone (the Lava Xolo X900) running on Intel's Atom SOC. Additionally, a Motorola Razer i was released running an updated chipset with faster Turbo. The chipset consumes more power than a traditional ARM SOC, but the single core Atom processor, even with it's limitations such as an in-order execution design and cache size trades and reduced clock speeds for TDP targets, was able to beat dual core ARM SOCs easily. Web browsing performance was especially (and notably) better than competition.
However, the Atom SOCs are hardly selling that well. That shows with the number of design wins Intel has had (takes two fingers to count, two). I believe that the SOC still consumes too much power (at least when compared to an ARM SOC). Add to that x86 architecture compatibility for Android (which Intel and Google are indeed working on together) and you know why this isn't really working out for Intel).
2013 brings 22 nm Tri-Gate transistors and an architectural refresh
Intel's plans for 2013 are big. 22 nm manufacturing process has finally been optimized for SOCs. You have to factor in that an SOC requires lots of variety, i.e. a mix of high voltage, low voltage, high and low speed transistors on the same piece of silicon. That is perhaps why the desktop and mobile release of 22 nm Ivy Bridge will be nearly a year old when Intel introduces Silvermont, the 22 nm architectural refresh of Atom with a process shrink from the 32 nm Medfield at the same time.
The process will shrink down to 14 nm as well, although we believe that would only happen by 2015 at the earliest (unless Intel is forced by competition to move even faster). Intel will make major architectural changes to Atom, including scaling up to 8 cores, enhanced Turbo, out-of-order execution of instructions and much more. Hyper threading is a feature that is still in question. This new generation of Atom processors will bring the much needed die shrink (drop in power consumption) as well as performance increase for Intel to capture back some of the tablet market (and enter the Android tablet market) as well as get more design wins for smartphones running on their new SOCs.
[wzslider autoplay="true"]
So what's next?
The next 3 years are all set to get heated up. ARM's Cortex A15 will appear in smartphones next year, just in time for the new Atom (which will take longer). Intel maintains a lead in the manufacturing process, the industry is still at 28 nm while Intel is already on 22 nm. However, the next shrink should more than even the playing field as the industry will move to 20 nm in 2014 (or 2H 2013), and at the same time SOCs based on ARM's first 64-bit ARMv8 processors will appear in products. These will pose a serious threat not only to Atom but low end Core processors as well (Celeron and Pentium, move aside). Add to that rumors that Apple might just abandon Intel and move to ARM based SOCs for their Macs (which is entirely feasible, even more deeper integration in their Mac and iOS operating systems, true cross-platform apps and games).
The next few years are more exciting than ever (if the last 2 years haven't already been, look where we have come since Samsung released their Galaxy S smartphone, to today where 1080p panels are becoming a common sight). Who would have thought mobile computing and those tiny devices that we carry in our pockets would become such an important element, in our lives as well as in the industry as a whole.
We can only wait and watch (and be a part of course!).
Intel didn't push Atom hard enough
Intel basically slackened up. They didn't not introduce architectural enhancements, process changes or anything new to the Atom line up, which was totally unlike what they were doing with their Core iSomethingmeaningless (thank you Charlie, semiaccurate) lineup. Atom kept getting revisions where more and more components got integrated into the chipset. Intel was moving to an SOC design, but it took them 4 years to move to 32 nm and they still haven't made any architectural advancements.
Enter 2012, Atom SOC for your phones
Intel's push to an SOC design is primarily because of the growing threat that ARM poses to the behemoth (and rightly so). And this year, we saw the first smartphone (the Lava Xolo X900) running on Intel's Atom SOC. Additionally, a Motorola Razer i was released running an updated chipset with faster Turbo. The chipset consumes more power than a traditional ARM SOC, but the single core Atom processor, even with it's limitations such as an in-order execution design and cache size trades and reduced clock speeds for TDP targets, was able to beat dual core ARM SOCs easily. Web browsing performance was especially (and notably) better than competition.
However, the Atom SOCs are hardly selling that well. That shows with the number of design wins Intel has had (takes two fingers to count, two). I believe that the SOC still consumes too much power (at least when compared to an ARM SOC). Add to that x86 architecture compatibility for Android (which Intel and Google are indeed working on together) and you know why this isn't really working out for Intel).
2013 brings 22 nm Tri-Gate transistors and an architectural refresh
Intel's plans for 2013 are big. 22 nm manufacturing process has finally been optimized for SOCs. You have to factor in that an SOC requires lots of variety, i.e. a mix of high voltage, low voltage, high and low speed transistors on the same piece of silicon. That is perhaps why the desktop and mobile release of 22 nm Ivy Bridge will be nearly a year old when Intel introduces Silvermont, the 22 nm architectural refresh of Atom with a process shrink from the 32 nm Medfield at the same time.
The process will shrink down to 14 nm as well, although we believe that would only happen by 2015 at the earliest (unless Intel is forced by competition to move even faster). Intel will make major architectural changes to Atom, including scaling up to 8 cores, enhanced Turbo, out-of-order execution of instructions and much more. Hyper threading is a feature that is still in question. This new generation of Atom processors will bring the much needed die shrink (drop in power consumption) as well as performance increase for Intel to capture back some of the tablet market (and enter the Android tablet market) as well as get more design wins for smartphones running on their new SOCs.
[wzslider autoplay="true"]
So what's next?
The next 3 years are all set to get heated up. ARM's Cortex A15 will appear in smartphones next year, just in time for the new Atom (which will take longer). Intel maintains a lead in the manufacturing process, the industry is still at 28 nm while Intel is already on 22 nm. However, the next shrink should more than even the playing field as the industry will move to 20 nm in 2014 (or 2H 2013), and at the same time SOCs based on ARM's first 64-bit ARMv8 processors will appear in products. These will pose a serious threat not only to Atom but low end Core processors as well (Celeron and Pentium, move aside). Add to that rumors that Apple might just abandon Intel and move to ARM based SOCs for their Macs (which is entirely feasible, even more deeper integration in their Mac and iOS operating systems, true cross-platform apps and games).
The next few years are more exciting than ever (if the last 2 years haven't already been, look where we have come since Samsung released their Galaxy S smartphone, to today where 1080p panels are becoming a common sight). Who would have thought mobile computing and those tiny devices that we carry in our pockets would become such an important element, in our lives as well as in the industry as a whole.
We can only wait and watch (and be a part of course!).