Intel abandons future Atom line of processors


Intel has canceled SoFIA and Broxton, but not all Atoms. For example, the recently revealed Apollo Lake platform.

Broxton was supposed to be a convergence point servicing both phone and tablets, so it remains to be seen if Apollo Lake will have low enough TDP to be put in small tablets and handhelds like Cherry Trail, or if it'll only be suitable for large tablets, convertibles, and laptops like Braswell. I have a feeling that Apollo Lake will go lower because they outfitted it with LPDDR3/4 support.
 
It looks like choosing an ARM processor for the Pyra was a good choice.

http://www.pcworld.com/article/3063...-tablet-markets-after-cutting-atom-chips.html

So it's a good thing the end-of-line OMAP 5 28nm process SoC from TI, who dropped out of the mobile SoC business right after releasing it? There is no successor for the OMAP 5, released in 2012.

That vs Intel Bay Trail 22nm alternative that also launched in 2012 that was followed up by the Cherry Trail on a 14nm process that was last updated in Q1 of this year? Apparently, from this news, there won't be a successor process released next year.

So - other than the chosen one getting development and additional versions cut off 4 years before the other...?

Yes, the Pyra is designed with a daughter card SoC setup. I would love to see a working Intel X7-Z8750 version, though I'm pretty sure it will never happen unless the Pyra sells 100,000 units and 3rd party manufacturers make the board and BIOS - which isn't that far fetched given that there are gumstick Cherry Trail HDMI computers running Ubuntu now - which means they've cracked the BIOS barrier. http://liliputing.com/2016/04/mele-pcg02u-is-a-70-ubuntu-pc-stick.html

I'm thrilled that the Pyra is getting built, regardless of the SoC used. I'm just not so sure we should celebrate the demise of a considered alternative SoC's future develoments when the one chosen was orphaned long ago.
 
So it's a good thing the end-of-line OMAP 5 28nm process SoC from TI, who dropped out of the mobile SoC business right after releasing it? There is no successor for the OMAP 5, released in 2012.

That vs Intel Bay Trail 22nm alternative that also launched in 2012 that was followed up by the Cherry Trail on a 14nm process that was last updated in Q1 of this year? Apparently, from this news, there won't be a successor process released next year.

So - other than the chosen one getting development and additional versions cut off 4 years before the other...?

From the perspective of getting followup support from the same manufacturer, it doesn't get too much worse than TI - they'll be doing more ARM SoCs but they don't seem interested in mobile applications at all. In that sense it doesn't really make a difference.

But from the perspective of picking a stable ISA, it makes a difference. If Intel drops out Atom that are suitable for handhelds that could be the last chance of getting any updated x86 SoC that's suitable for a device like Pyra (not holding out much for AMD here). So if the plan was to stick with a single ISA for years down the road that would have had more severe repercussions than OMAP5 because there are still other ARMv7+ SoCs out and coming out.

That wouldn't strictly have to be the case because a multi-ISA device is possible. Multi-arch repos can me maintained, possibly even if fat binaries if necessary. But I hope people aren't under the impression that this a negligible or trivial amount of overhead. The increased workload on the system software writers, a very limited resources, will be significant. And at least some developers will feel some significant strain supporting multiple ISAs with testing on multiple devices.

Yes, the Pyra is designed with a daughter card SoC setup. I would love to see a working Intel X7-Z8750 version, though I'm pretty sure it will never happen unless the Pyra sells 100,000 units and 3rd party manufacturers make the board and BIOS - which isn't that far fetched given that there are gumstick Cherry Trail HDMI computers running Ubuntu now - which means they've cracked the BIOS barrier. http://liliputing.com/2016/04/mele-pcg02u-is-a-70-ubuntu-pc-stick.htm
http://liliputing.com/2016/04/mele-pcg02u-is-a-70-ubuntu-pc-stick.html

AFAIK it's only with Apollo Lake that Intel has finally enabled booting from primary storage instead of an external BIOS chip. What about this device do you think removes that? Bear in mind that a tiny capacity SPI flash chip can be pretty small.
 
From the perspective of getting followup support from the same manufacturer, it doesn't get too much worse than TI - they'll be doing more ARM SoCs but they don't seem interested in mobile applications at all. In that sense it doesn't really make a difference.
One positive thing I see from Ti is what I see from the Kernel mailing list, it's nice that Nikolaus is getting a bit of technical support from the Ti guys. Not sure too many other companies would give even that much unless we're shilling out big monies.
 
One positive thing I see from Ti is what I see from the Kernel mailing list, it's nice that Nikolaus is getting a bit of technical support from the Ti guys. Not sure too many other companies would give even that much unless we're shilling out big monies.

We're actually getting quite a lot of support, but that's not officially by TI. They're basically doing this in their spare time because they're interested :)
Some of them also were part of the official OMAP5 team, so you can't have any better help :D
 
AFAIK it's only with Apollo Lake that Intel has finally enabled booting from primary storage instead of an external BIOS chip. What about this device do you think removes that? Bear in mind that a tiny capacity SPI flash chip can be pretty small.

Not that they have removed the need for BIOS but that they have one working that can load Linux instead of being locked to Windows or Android on an Intel Mobile Atom SoC with UEFI.

That stick is an Atom Z3735F. I may have misspoke above and called it Cherry Trail - it's still Bay Trail.
http://ark.intel.com/products/80274/Intel-Atom-Processor-Z3735F-2M-Cache-up-to-1_83-GHz

But, the fact that they're shipping it with Ubuntu loaded - that means they've figured out a key to the desktop OS on Intel Mobile Atom platforms conundrum. I'm easily impressed, but this is the first product I've seen ship with an Intel mobile phone/tablet SoC running a desktop Linux distribution. That is the accomplishment that makes it relevant here - proof of concept.
[doublepost=1461967704,1461967306][/doublepost]
We're actually getting quite a lot of support, but that's not officially by TI. They're basically doing this in their spare time because they're interested :)
Some of them also were part of the official OMAP5 team, so you can't have any better help :D

That is a great thing too. It's always impressed me how there are several communities that are all helping for the Pyra.

I still hope you sell so many of them (100,000+?) that you or even 3rd party manufacturers start selling and supporting X86 boards for it - in addition to future ARM SoC options down the road. Heck, with process shrinkage a single daughter board with both X86 and ARM SoC sets and users select platform at boot would be cool too, but we're more likely to see pork on the wing.
 
I really thoughtt the Atom was a success, at least the new and promising Cherry Trail. It enabled alot of these new "Stick-PC" things and it would still be an great option for an x86 Pyra, so it MUST be a good thing. ;)
Well, maybe ED can get some of these abandoned Atom chips for cheap if they are no longer an important piece of the market and Intel wants to empty its warehouses.
 
Well, maybe ED can get some of these abandoned Atom chips for cheap if they are no longer an important piece of the market and Intel wants to empty its warehouses.
If you're going to plonk down the overhead of designing and testing a design, you want to be sure it'll last for a while. Using EOL parts in the design phase is just about the worst idea ever.
 
Release date and product life cycle are two completely different things. There's application processors released over 15 years ago still in production available over-the-counter. On the other hand, a product that you know is discontinued and no longer produced, you have no idea if you'll be able to source two years from now. You don't design in EOL products, so the whole "get a bunch of unwanted processors that aren't made anymore" concept is just stupid. Unless you're a chinese company that quickly design the thing, make a single batch and then forget about it ever happening.
 
I'm not sure what your point is, Grench. The probability that there will be future systems on chips with ARM processors is nearly 100%, if not from Texas Instruments, then from some other organization. The probability that there will be future systems on chips with Atom processors for handheld devices is now close to 0%.

I do not celebrate the demise of the Atom processors for handheld devices. I do not feel any emotion regarding the matter.
 
I wouldn't be surprised in the slightest if Nvidia announced they are getting out of the mobile SoC business also. They have threathened to do it in the past, now their market is even smaller , almost non existent for the X1 outside of the Shield TV and Pixel C . Every generation of their SoCs has seen the up take for them shrink
 
According to this link Intel will most likely keep the Core-M line of processors. I think that's fine enough, because in both cases (Atom and Core-M) the question remains how to get the heat out of the Pyra. So I don't see how killing the Atom line will influence the chance of getting a x86 cpu for the Pyra one day.
 
CoreM is expensive though. Like 10 times what a Cherry Trail SoC costs. A lot anyway what ever the figure is, can't remember if it was $140 or $280

I think that was pretty much the crux of the story, ie that Intel are moving away from producing the cheap low margin stuff. So yeah other stuff might be still made by them that may be low power consumption, but they will probably be expensive (prohibitively so for handhelds anyway)
 
Last edited:
Release date and product life cycle are two completely different things. There's application processors released over 15 years ago still in production available over-the-counter. On the other hand, a product that you know is discontinued and no longer produced, you have no idea if you'll be able to source two years from now. You don't design in EOL products, so the whole "get a bunch of unwanted processors that aren't made anymore" concept is just stupid. Unless you're a chinese company that quickly design the thing, make a single batch and then forget about it ever happening.
Well, is the OMAP5 still officialy supported by TI? Not sure about this.

CoreM is expensive though. Like 10 times what a Cherry Trail SoC costs. A lot anyway what ever the figure is, can't remember if it was $140 or $280

I think that was pretty much the crux of the story, ie that Intel are moving away from producing the cheap low margin stuff. So yeah other stuff might be still made by them that may be low power consumption, but they will probably be expensive (prohibitively so for handhelds anyway)
Yupp, CoreM are starting way over $200, not really an option for a small niche Handheld device. And their TDP is afaik easily 6W which you have to get out of such devices somehow.

skylake-core-m-100610678-orig.png
*
 
Well, the reason (I believe) for this disaster of the Atom line is the Architecture implementation in terms of the software environment. Whether you like it or not, Android is a pretty slick piece of software. That is demonstrated by the HUGE acceptance in the smartphone market, and market share. Intel got into the game too late; I'm not saying into the SFF device race as Intel had been into the Stealy A1xx ARM chip race since the early/mid 2000's. Then they had their U-XXXX series Core series CPU's, which where used in just about every UMPC until the catagory was absorbed by Netbooks and went away. The performance/Power consumption was very much in Intel's favor, but once the Software moved away from x86 to a mobile concentric OS like Android it spelled out the slow demise of Intel's Atom/ARM lines. For quite a while x86 never really had an acceptable mobile platform, as you seen once Windows 8/.1 came out they tried to bring back their user base from iOS/Android but that was a bigger disaster and made things far worse. In my eyes solidifying the notion that Intel will Never (long time) have a solid candidate for an Android/iOS killer.

Funny enough, It used to be price that made people get a cheap non-Windows tablet, now Android/iOS tablets/devices are even more expensive than that of the x86 counterpart and they STILL sell better.
 
Back
Top