LPX Episode 9: Funeral for an Atom (processor)

Intel’s Atom processors have been powering low-cost computers since 2008. The first models were used in netbooks, but these days you can find Atom chips in Chromebooks, Windows and Android tablets, tiny desktop computers, and even a handful of smartphones.

But Intel never really made much headway in the smartphone space, where companies like Apple and Samsung typically use ARM-based processors for their devices.

So maybe it shouldn’t have been a huge surprise when Intel announced in April that it was scrapping plans to launch next-gen Atom chips code-named “Broxton” and “SoFIA” for phones, tablets, and other low-power devices.

That means 2015’s Intel Atom “Cherry Trail” processors are likely to be the end of the line… at least for a while.

The company isn’t exiting the low-power space altogether. Intel will continue to ship Cherry Trail processors for a little while.

Intel will also launch its first Celeron and Pentium chips based on the “Apollo Lake” platform soon. These are low-cost, low-power chips that are cousins to Broxton processors that have been canceled. But they’ll use a little more power and cost a little more money than the discontinued Atom/Broxton family.

Some folks are happy to see Atom go. Others are disappointed that this could mean the end of Intel’s investment in the 5-10 inch tablet space.

I started writing about affordable, portable computers at Liliputing.com shortly before the first Atom processor launched. A lot has changed since then: the rise and fall of netbooks, the rise and stagnation of tablets, and the rise and rise of smartphones.

But I wasn’t the only person documenting the changing mobile space. In order to properly mourn the demise of Intel’s Atom processor lineup, I decided to record a conversation with my friends and colleagues Sascha Pallenberg, Nicole Scott, and Steve “Chippy,” Paine.

Sascha and Nicole are co-founders of mobile tech blogs MobileGeeks.com and MobileGeeks.de, and Chippy runs a number of mobile-focused websites including UMPC Portal and CarryPad.

You can help support the LPX podcast by contributing to our Patreon campaign.

Podcast: Play in new window | Download (Duration: 48:08 — 38.1MB)

Shogmaster says

July 4, 2016 at 12:09 pm

There is lot of hyperbole and way out there predictions of near future of these processor in this podcast. Sascha especially had some way out there vision of the near future that seems not much to do with the realities. If cloud and IOT devices that leverage the cloud is the future, and these portable devices will become simple terminals to connect to the power of the cloud, then Intel and IBM are in the best position, not Apple nor other ARM chip manufacturers.

Atom’s demise in the hands of ARM chips had more to do with Intel’s terrible marketing strategy than any inherent lack of capabilities of Atom or X86. Yes, ARM has advantage over X86 in silicon size and amount of transistors in the early going due to RISC being more efficient and simple for hardware requirement than SISC/X86. But in the last 5 years, ARM processors have gotten just as big as some low end X86 processor (like Atom), and the price advantage when compared transistor to transistor had reached a near parity. At the end of the day, size and the yield of the chip determines the price, and Atom was neck and neck with mid range/high end ARM chips with Cherrytrail. The real reason for Intel killing Atom now is that simply, time had run out on Atom, where the years of subsidizing to gain marketshare strategy caught up with the finances by the time Intel ALLOWED Atom to be as good as it could be. For years, Intel had artificially limited capabilities of Atom for their marketing reasons. They could have added features like Out of Order execution much earlier than they did, but they held them back just to differentiate it and justify the cost difference between Atom and Core architectures.

Atom’s space in the low cost market cannot be replaced outright. Intel has, with this move, thrown in the towel for low margin chip business. What I imagine they will try to do is to concentrate on their high margin business to make their stock holders happy again, and perhaps tweak the Core M to fulfill some of the market demand they have vacated with Atom. Again, cost is determined by size and yield of the chip. And if you take Core M and reduce the die size by perhaps marrying the deduced size GPU from Atom with it, they can sell Core M cheaper than they do now (GPU takes up at least half of the Core M’s die). They can slot such version below Core M3, and reduce the price by 25~40% theoretically. That could potentially serve as a good fanless low(ish) cost chip for high end 10″ Windows tablets to compete with iPads and high end Android tablets. Low end tablet market in the 7~8″ range is done for Windows, with Atom being killed.

Also, I scratched a big bald spot on my head with your comments about Core M. I don’t think you guys understand what Core M is. It’s not a less capable brother of Core U. It is simply a thermally restricted Core U, and with same cooling overhead, will perform just as well as Core U. Before being call Core M, it was called Core Y, which was a binned (cherry picked) low yield, expensive lower voltage version of Core U. Core M, be it Broadwell or Skylake, will perform as respective versions of Core U when given similar cooling parameters. Only difference is that Core U can never operate low as 4W while Core M can. Core M is simply a more thermally flexible Core U, capable of operating from 4W all the way up to 15W. Core U is limited to operating from 11W to 15W. At 15W, both Core M and Core U will perform similarly. So stop painting Core M devices as lower end devices than Core U mobile devices. With Core M, you don’t have to stick in an expensive liquid cooling rig into a tablet to make it fanless. And if such Core U device can’t thermally sustain 15W operation for any decent duration, it’s all a waste of time anyways and would have been better to stick in a Core M with cheaper heatsinks in the first place.

And as for nVidia and their embedded processor for cars. all it is, is an ARM processor with a bigger GPU die married to it. It’s not some crazy special chip. X1’s 256 Maxwell streaming GPU processors is nice and all, but it’s not much of a success in the ARM space due to size and cost. It’s too big and thus expensive. In fact, nVidia is quite desperate to put that chip into more than just their Shield tablets and TV set top boxes. The chip is a bomb in ARM space, and thus they are desperate to stick into cars to recoup the development costs. At least AMD has big contracts with MS and Sony for their console CPU with the Jaguar core. NVidia has not much.going at all. Like of Rockchip and Mediatech has much more OEM wins.

And BTW, how the hell is Intel gonna go into embedded markets such as cars with Atom being killed off? ARM and Atom is the only viable candidate in that market. Fanless and low cost is the only way to compete in that market. Without Atom, it’s not happening.

Comments

D. Hugh Redelmeier says
May 21, 2016 at 1:50 pm
There are Celerons based on Core M, I infer. For example, the Celeron 2961Y in my Zotac CI321 Nano.
D. Hugh Redelmeier says
May 21, 2016 at 1:57 pm
Atom was intentionally crippled in so many ways to protect their other markets. This caused a lot of problems.
Add to that Microsoft’s licensing discontinuity at the 2G RAM / 32G disk point.
Compare memory restrictions of 2G for many Atom variants vs Avoton’s 64G.
Think of Baytrail for Windows limited to 32-bit UEFI (an Intel choice as I understand).
I’m really annoyed that Intel has left Atom poorly supported on Linux. Sound on HDMI and cstate code hanging are two important examples.
Bobby Salvin says
May 22, 2016 at 9:01 pm
For several years I used an 11.6″ Asus Vivobook with a Clover trail processor as my mobile computer, and I thought it was great. It got like 15 hours actual battery life. I replaced it with a core m device, which was the only device that could get close to the Atom battery life. The core m is a lot faster too.
Bobby Salvin says
May 22, 2016 at 9:04 pm
I don’t see how Microsoft can allow Windows to not be available at the low end.
Ajemo Haltom says
May 23, 2016 at 11:16 am
I appreciate Chippy’s forecast insight (if nothing else to ruffle our IT feathers and get us thinking). But I think its important to keep in mind that enterprise devices have not and will not for years adopt Apple or Android ecosystems. I agree it is happening, but Microsoft has a huge lead in this field. As long as the thrust is to make the homes function like a business, enterprise will continue to be the standard bearer for compatibility and therefore x86 compatibility is a necessity.
For example: It used to be you had to buy a computer monitor for VGA display, now any Walmart TV can be used as a computer monitor. Enterprise instructed that migration.
Jim Shortz says
June 2, 2016 at 11:07 am
My favorite tech blog meets my favorite YouTube channel. Loved the show
Shogmaster says
July 4, 2016 at 12:09 pm
There is lot of hyperbole and way out there predictions of near future of these processor in this podcast. Sascha especially had some way out there vision of the near future that seems not much to do with the realities. If cloud and IOT devices that leverage the cloud is the future, and these portable devices will become simple terminals to connect to the power of the cloud, then Intel and IBM are in the best position, not Apple nor other ARM chip manufacturers.
Atom’s demise in the hands of ARM chips had more to do with Intel’s terrible marketing strategy than any inherent lack of capabilities of Atom or X86. Yes, ARM has advantage over X86 in silicon size and amount of transistors in the early going due to RISC being more efficient and simple for hardware requirement than SISC/X86. But in the last 5 years, ARM processors have gotten just as big as some low end X86 processor (like Atom), and the price advantage when compared transistor to transistor had reached a near parity. At the end of the day, size and the yield of the chip determines the price, and Atom was neck and neck with mid range/high end ARM chips with Cherrytrail. The real reason for Intel killing Atom now is that simply, time had run out on Atom, where the years of subsidizing to gain marketshare strategy caught up with the finances by the time Intel ALLOWED Atom to be as good as it could be. For years, Intel had artificially limited capabilities of Atom for their marketing reasons. They could have added features like Out of Order execution much earlier than they did, but they held them back just to differentiate it and justify the cost difference between Atom and Core architectures.
Atom’s space in the low cost market cannot be replaced outright. Intel has, with this move, thrown in the towel for low margin chip business. What I imagine they will try to do is to concentrate on their high margin business to make their stock holders happy again, and perhaps tweak the Core M to fulfill some of the market demand they have vacated with Atom. Again, cost is determined by size and yield of the chip. And if you take Core M and reduce the die size by perhaps marrying the deduced size GPU from Atom with it, they can sell Core M cheaper than they do now (GPU takes up at least half of the Core M’s die). They can slot such version below Core M3, and reduce the price by 25~40% theoretically. That could potentially serve as a good fanless low(ish) cost chip for high end 10″ Windows tablets to compete with iPads and high end Android tablets. Low end tablet market in the 7~8″ range is done for Windows, with Atom being killed.
Also, I scratched a big bald spot on my head with your comments about Core M. I don’t think you guys understand what Core M is. It’s not a less capable brother of Core U. It is simply a thermally restricted Core U, and with same cooling overhead, will perform just as well as Core U. Before being call Core M, it was called Core Y, which was a binned (cherry picked) low yield, expensive lower voltage version of Core U. Core M, be it Broadwell or Skylake, will perform as respective versions of Core U when given similar cooling parameters. Only difference is that Core U can never operate low as 4W while Core M can. Core M is simply a more thermally flexible Core U, capable of operating from 4W all the way up to 15W. Core U is limited to operating from 11W to 15W. At 15W, both Core M and Core U will perform similarly. So stop painting Core M devices as lower end devices than Core U mobile devices. With Core M, you don’t have to stick in an expensive liquid cooling rig into a tablet to make it fanless. And if such Core U device can’t thermally sustain 15W operation for any decent duration, it’s all a waste of time anyways and would have been better to stick in a Core M with cheaper heatsinks in the first place.
And as for nVidia and their embedded processor for cars. all it is, is an ARM processor with a bigger GPU die married to it. It’s not some crazy special chip. X1’s 256 Maxwell streaming GPU processors is nice and all, but it’s not much of a success in the ARM space due to size and cost. It’s too big and thus expensive. In fact, nVidia is quite desperate to put that chip into more than just their Shield tablets and TV set top boxes. The chip is a bomb in ARM space, and thus they are desperate to stick into cars to recoup the development costs. At least AMD has big contracts with MS and Sony for their console CPU with the Jaguar core. NVidia has not much.going at all. Like of Rockchip and Mediatech has much more OEM wins.
And BTW, how the hell is Intel gonna go into embedded markets such as cars with Atom being killed off? ARM and Atom is the only viable candidate in that market. Fanless and low cost is the only way to compete in that market. Without Atom, it’s not happening.

Share this:

Comments