Why on earth would you want that? It wouldn't be compatible with any legacy Windows Phone apps so you'd only have WinRT universal apps, which are thin on the ground at the moment.
If you're under the impression it would support Win32 apps, < 8" screens don't support desktop apps. Only WinRT.
No, Microsoft is literally stripping the OS of the features required to run desktop apps to reduce the install size on < 8" devices. The actual interface on these devices is going to be quite a bit different from larger Windows 10 devices. I know Microsoft likes to claim that they're all "Windows 10" but the small devices are functionally going to feature only a subset of features.
Yeah, I hate that Microsoft is calling both platforms Windows 10. Windows 10 for phones is clearly a different OS, as anyone who has used both can attest to. People seem to be expecting the phone SKU to be the PC version without the desktop, but that's not at all what is happening.
It really should have been "Windows 10" and "Windows 10 Mobile" or something of that sort, because as it is now, "Windows 10" is a different OS from "Windows 10" and it makes it far more convoluted to talk about the two.
And yet people even on this site, a highly technical one, are already thinking that there's a possibility of docking a larger screen to the phone/small tablet version in order to get a full desktop.
The general public is going to get confused. That's the whole reason for the naming in the first place, to trick people into thinking they're the same thing when they're clearly not.
Well there probably will be a way to do this but it will require some hacks ( unsupported and maybe even a bit over what the license says you can do ones ). As long as the core OS is there, kernel and drivers and stuff and it's the same Windows some libs and UI and whatever could probably hacked over that. The only problem so far was that the core OS was different and there weren't that many x86 chips.
I'll be interested to see if Microsoft expressly prohibits this functionality or if they resort to other market trickery, like making OEMs pay for a full license, thus adding $100 to the price of a phone/phablet. It's not like the hardware isn't fast enough. I've got a Bay Trail tablet that runs Windows 10 desktop just perfectly. What difference does screen size make?
I would do just about anything short of murder to have a Galaxy S6/Note-style device with the full W10 experience. Set it on the desk, have it wirelessly charge while it wirelessly connects to my display and peripherals... That's the future-world I want to live in.
Even so, there are likely to be many more apps for such a version of Windows as compared to the current version of Windows Phone - no need to recompile to tailor an app to a phone vs. tablet/PC.
Maybe at some point in the far future. Basically all current WinRT apps will need to be recompiled to be "Universal", they haven't even given us a finalized SDK for Windows 10 "Univeral Apps" either. The current one for Windows 8.1 actually requires two different interface projects with the option to share code in a third library-type project.
My hope is they're going for a mess now, for a cleaner tomorrow. But it's not like they haven't tried and failed to do that before. I don't recommend anyone buy anything based on possible future support, particularly with Microsoft's track record.
As a huge WP, I still need to warn you to clearly put "sarcasm" on comments like this :)
Intel can still catch the train. But WP has no chance of being a competitive 3rd place. With A LOT of luck and perfect execution, best case scenario is 10%.
they lost the battle for USA, they don't even participate in China. Europe alone can not keep them afloat. Especially since a lot of people in Europe are pissed at being utterly neglected (no Cortana, few flow keyboard languages supported, no xbox, no xbox music, no any MS service at all actually).
Really? Because what I see is a "my vaporware can beat up your vaporware" presentation. Intel has "addressed" its lack of a phone chip by admitting there's a problem. I guess that's a start, but by the time they actually ship something will we already have Cortex A-72 on 14nm in production?
I'd also love to read the fine print of that Rockchip agreement. Intel may think they can control the golem, but at some point I expect Rockchip's ever cheaper prices for ever more cores are going to conflict with Intel's price-maintenance agenda. There's pretty much ALWAYS a conflict brewing when you are simultaneously a vendor to some other business AND you sell a product that directly competes with that business. (ARM avoids this by not selling CPUs.). I imagine Intel think they have this carefully controlled, that Rockchip is condemned to forever selling crappy low-end CPU+modems that Intel doesn't want to sell anyway. But that low end stuff has a way of making itself very powerful when it's cheap enough, and the alternatives aren't...
Bunch of mistakes in this article: SofIA and Baytrail are not the same - Baytrail released in fall 2013 and aimed at Windows (tablet, netbook/nettop, ...), Android tablets, Chromebooks. No Baytrail skus had integrated cellular modems. Baytrail (and Merrifield and Moorefield) are manufactured in Intel's 22nm fab.
In contrast SofIA chips (the x3 line) are fabbed on 28nm (TSMC most likely), have integrated modems, and are coming to market now. They are also much cheaper and thus a better/viable competitor in the entry/value smartphone market - no contra revenue.
If it produced on 28nm TMSC it is surely an ARM core then, not a x86. I don't think than Intel is ever going to allow for a production of x86 chips outside its foundries and most of all, there was not enough time to convert a x86 core to a TMSC PP.
Of all of this, I would like to know what Intel IP are in there but the name.
For sure these are using Intel IP - silvermont cpu, intel modem, ... The only external IP that is announced so far is Mali GPU.
Yes, it seems off that Intel is dabbing soc at TSMC , but this was clearly explained in earlier interviews- porting the modem to intel 14nm was harder than porting intel cpu to 28nm. This is continuation of intel WPRD feature phones from previous Infineon which was manufactured at TSMC. Intel had been manufacturing discrete 3G and LTE modems has been at TSMC for as while.
Intel will be moving modem on to Intel 14nm this year for products next year.
Indeed. Bay Trail was all 22 nm within Intel's fabs. It appears SoFIA will be fabbed elsewhere as many other sites are reporting it to be 28 nm -- Intel does not run a 28 nm process, unless it is top secret and not mentioned anywhere. Most likely being made with TSMC.
Usually many hints make a proof... It is manifactured outside Intel's fabs It' s on a different nanometry and PP in general (which is surely not better than those provided lately by Intel in terms of quality) It has been created quite soon (so doubts may raise on proprer x86->28nm TMSC PP port) It uses an ARM based IP for GPU (not even a PowerVR as previous Atoms)
It seems to me that all these x3 are simply ARM SoCs made by RockChip with an Intel label on them.
We know Intel can't compete in term of costs with chinese (but probably all) ARM SoCs. And we know that on the same PP Atoms can't compete in terms of perf/watt.
If these x3 are really x86 based, we have a combination of the worst of the two cases above: worse PP and a x86 architetcure. How can these SoCs compete with very cheap and low power Cortex-A7 SoCs?
It's obviously an x86 CPU... Otherwise why would Intel be involved at all? To sell their awesome modems?!?
It's made with TSMC because of price and Intels own 14nm/22nm being busy with Broadwell. In this segment price is the most important factor, so the fact that Intels own fans could make a higher performing chip means little.
That the GPU came from ARM is as insignificant as the fact that one of their old Atom GPUs came from Power VR. They chose ARMs Mali GPU because their own 5000-series GPU is too big, too fast and too powerhungry for this market. ARM's Mali solution fit Intels performance and price metrics, it's as simple as that.
So I ask again, if it is x86, what's the contribution of Rockchip to it? Intel already knows how to make an Atom with PowerVR. Building it at 28nm may be cheaper than using its own 22nm, but surely not at 32nm. And seen all the fuss about the 32nm Intel quality against competing 28nm, Intel (od and now cheap) 32nm should still perform better than TMSC PP. Without speaking about the fact that producing a x86 core outside its foundries exposes its architectural secrets to the rest of the market. Then, he fact that the chip is restricted to Asian market (probably because in cheap Chinese product sponsorized by Rockchip itself) still raises more than eyebrow.
Basically, the low-end x5 only supports up to 2GB of RAM, but the higher end x5 and the x7 model both support up to 8GB. Hopefully we only see higher-end x5 and x7 models in OEM designs.
Intel uses own GPU for Baytrail and Cherrytrail (now x5, x7). Mali is being used for very low end die area, perf, power (and somewhat reduced feature set eg 3d/compute APIs)
given that intel are using the Mali GPU's then when can we expect the user space FLOSS driver code and documentation to be written and made available ?
also what SIMD will these low power Mali GPU equipped intel CPU's use ? AVX2 or the better NEON SIMD, with neither current SIMD then its not worth the trouble for the enc consumer to invest there now with octacore 64bit/128 bit NEON coming around now on mass
So they don't have an up to date LTE or eMMC implementation and the LTE is connected via USB 2? Intel needs a kick in the @$$ to get them to actually compete for market share in this segment.
All this fairly recent modem and peripheral integration aside, I could never really care much about the atom line.
It is not a product line where the company wanted to make the best possible chip at a low power budget. It always seemed like intel wanted to make some chips good enough to capture the low power device market, at the same time ensuring them to be crappy enough to show off how fast their main line of chips (now broadwell) are. No wonder they still haven't captured the market quite.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
37 Comments
Back to Article
nathanddrews - Monday, March 2, 2015 - link
Atom/W10 phone, please!Flunk - Monday, March 2, 2015 - link
Why on earth would you want that? It wouldn't be compatible with any legacy Windows Phone apps so you'd only have WinRT universal apps, which are thin on the ground at the moment.If you're under the impression it would support Win32 apps, < 8" screens don't support desktop apps. Only WinRT.
edzieba - Monday, March 2, 2015 - link
"If you're under the impression it would support Win32 apps, < 8" screens don't support desktop apps. Only WinRT."On the phone display, yes. but it would make for a rather nice dockable device (like the old Atrix 4G, but with Windows 10 rather than Linux).
Flunk - Monday, March 2, 2015 - link
No, Microsoft is literally stripping the OS of the features required to run desktop apps to reduce the install size on < 8" devices. The actual interface on these devices is going to be quite a bit different from larger Windows 10 devices. I know Microsoft likes to claim that they're all "Windows 10" but the small devices are functionally going to feature only a subset of features.jhoff80 - Monday, March 2, 2015 - link
Yeah, I hate that Microsoft is calling both platforms Windows 10. Windows 10 for phones is clearly a different OS, as anyone who has used both can attest to. People seem to be expecting the phone SKU to be the PC version without the desktop, but that's not at all what is happening.It really should have been "Windows 10" and "Windows 10 Mobile" or something of that sort, because as it is now, "Windows 10" is a different OS from "Windows 10" and it makes it far more convoluted to talk about the two.
markiz - Monday, March 2, 2015 - link
people will still refer to it as "Windows phone" so I don't see confusion there.There will also be no overlap between the two with line set at <8"<.
jhoff80 - Monday, March 2, 2015 - link
And yet people even on this site, a highly technical one, are already thinking that there's a possibility of docking a larger screen to the phone/small tablet version in order to get a full desktop.The general public is going to get confused. That's the whole reason for the naming in the first place, to trick people into thinking they're the same thing when they're clearly not.
nathanddrews - Monday, March 2, 2015 - link
That assumes that no vendors will build sub-8" devices with full Windows 10. Has Microsoft officially stated that they won't allow it?mcirsta - Monday, March 2, 2015 - link
Well there probably will be a way to do this but it will require some hacks ( unsupported and maybe even a bit over what the license says you can do ones ).As long as the core OS is there, kernel and drivers and stuff and it's the same Windows some libs and UI and whatever could probably hacked over that.
The only problem so far was that the core OS was different and there weren't that many x86 chips.
nathanddrews - Tuesday, March 3, 2015 - link
I'll be interested to see if Microsoft expressly prohibits this functionality or if they resort to other market trickery, like making OEMs pay for a full license, thus adding $100 to the price of a phone/phablet. It's not like the hardware isn't fast enough. I've got a Bay Trail tablet that runs Windows 10 desktop just perfectly. What difference does screen size make?http://youtu.be/o7irMplW88A
I would do just about anything short of murder to have a Galaxy S6/Note-style device with the full W10 experience. Set it on the desk, have it wirelessly charge while it wirelessly connects to my display and peripherals... That's the future-world I want to live in.
jospoortvliet - Friday, March 6, 2015 - link
Ubuntu has it.../me hides
mkozakewich - Monday, March 2, 2015 - link
I've got a < 8" device (7"), and it runs desktop apps fine. Unless that's a change in Windows 10?jhoff80 - Monday, March 2, 2015 - link
It is.Ktracho - Monday, March 2, 2015 - link
Even so, there are likely to be many more apps for such a version of Windows as compared to the current version of Windows Phone - no need to recompile to tailor an app to a phone vs. tablet/PC.Flunk - Monday, March 2, 2015 - link
Maybe at some point in the far future. Basically all current WinRT apps will need to be recompiled to be "Universal", they haven't even given us a finalized SDK for Windows 10 "Univeral Apps" either. The current one for Windows 8.1 actually requires two different interface projects with the option to share code in a third library-type project.My hope is they're going for a mess now, for a cleaner tomorrow. But it's not like they haven't tried and failed to do that before. I don't recommend anyone buy anything based on possible future support, particularly with Microsoft's track record.
bradleymikej - Monday, March 2, 2015 - link
The Wintel freight train is rounding the bend at full steam!!!markiz - Monday, March 2, 2015 - link
Sure..As a huge WP, I still need to warn you to clearly put "sarcasm" on comments like this :)
Intel can still catch the train. But WP has no chance of being a competitive 3rd place. With A LOT of luck and perfect execution, best case scenario is 10%.
they lost the battle for USA, they don't even participate in China. Europe alone can not keep them afloat. Especially since a lot of people in Europe are pissed at being utterly neglected (no Cortana, few flow keyboard languages supported, no xbox, no xbox music, no any MS service at all actually).
name99 - Monday, March 2, 2015 - link
Really? Because what I see is a "my vaporware can beat up your vaporware" presentation.Intel has "addressed" its lack of a phone chip by admitting there's a problem. I guess that's a start, but by the time they actually ship something will we already have Cortex A-72 on 14nm in production?
I'd also love to read the fine print of that Rockchip agreement. Intel may think they can control the golem, but at some point I expect Rockchip's ever cheaper prices for ever more cores are going to conflict with Intel's price-maintenance agenda. There's pretty much ALWAYS a conflict brewing when you are simultaneously a vendor to some other business AND you sell a product that directly competes with that business. (ARM avoids this by not selling CPUs.). I imagine Intel think they have this carefully controlled, that Rockchip is condemned to forever selling crappy low-end CPU+modems that Intel doesn't want to sell anyway. But that low end stuff has a way of making itself very powerful when it's cheap enough, and the alternatives aren't...
croandeimos - Monday, March 2, 2015 - link
I think you misspelled the name for upcoming high end SoC.. its 'Broxton' and not 'Braxton'.rootheday3 - Monday, March 2, 2015 - link
Bunch of mistakes in this article:SofIA and Baytrail are not the same - Baytrail released in fall 2013 and aimed at Windows (tablet, netbook/nettop, ...), Android tablets, Chromebooks. No Baytrail skus had integrated cellular modems. Baytrail (and Merrifield and Moorefield) are manufactured in Intel's 22nm fab.
In contrast SofIA chips (the x3 line) are fabbed on 28nm (TSMC most likely), have integrated modems, and are coming to market now. They are also much cheaper and thus a better/viable competitor in the entry/value smartphone market - no contra revenue.
And, lastly, its Broxton, not Braxton.
CiccioB - Monday, March 2, 2015 - link
If it produced on 28nm TMSC it is surely an ARM core then, not a x86. I don't think than Intel is ever going to allow for a production of x86 chips outside its foundries and most of all, there was not enough time to convert a x86 core to a TMSC PP.Of all of this, I would like to know what Intel IP are in there but the name.
rootheday3 - Monday, March 2, 2015 - link
For sure these are using Intel IP - silvermont cpu, intel modem, ... The only external IP that is announced so far is Mali GPU.Yes, it seems off that Intel is dabbing soc at TSMC , but this was clearly explained in earlier interviews- porting the modem to intel 14nm was harder than porting intel cpu to 28nm. This is continuation of intel WPRD feature phones from previous Infineon which was manufactured at TSMC. Intel had been manufacturing discrete 3G and LTE modems has been at TSMC for as while.
Intel will be moving modem on to Intel 14nm this year for products next year.
Exophase - Monday, March 2, 2015 - link
While there may not be other external IP announced, I doubt they teamed up with Rockchip if the GPU was the only foreign part.CiccioB - Tuesday, March 3, 2015 - link
Sorry, it is not clear if the core are x86 or ARM based for x3 series. If the latter there's nothing but the modem as Intel IP in those SoCs.JumpingJack - Monday, March 2, 2015 - link
Indeed. Bay Trail was all 22 nm within Intel's fabs. It appears SoFIA will be fabbed elsewhere as many other sites are reporting it to be 28 nm -- Intel does not run a 28 nm process, unless it is top secret and not mentioned anywhere. Most likely being made with TSMC.CiccioB - Tuesday, March 3, 2015 - link
Usually many hints make a proof...It is manifactured outside Intel's fabs
It' s on a different nanometry and PP in general (which is surely not better than those provided lately by Intel in terms of quality)
It has been created quite soon (so doubts may raise on proprer x86->28nm TMSC PP port)
It uses an ARM based IP for GPU (not even a PowerVR as previous Atoms)
It seems to me that all these x3 are simply ARM SoCs made by RockChip with an Intel label on them.
We know Intel can't compete in term of costs with chinese (but probably all) ARM SoCs.
And we know that on the same PP Atoms can't compete in terms of perf/watt.
If these x3 are really x86 based, we have a combination of the worst of the two cases above: worse PP and a x86 architetcure. How can these SoCs compete with very cheap and low power Cortex-A7 SoCs?
V900 - Wednesday, March 4, 2015 - link
Nonsense...It's obviously an x86 CPU... Otherwise why would Intel be involved at all? To sell their awesome modems?!?
It's made with TSMC because of price and Intels own 14nm/22nm being busy with Broadwell. In this segment price is the most important factor, so the fact that Intels own fans could make a higher performing chip means little.
That the GPU came from ARM is as insignificant as the fact that one of their old Atom GPUs came from Power VR. They chose ARMs Mali GPU because their own 5000-series GPU is too big, too fast and too powerhungry for this market. ARM's Mali solution fit Intels performance and price metrics, it's as simple as that.
CiccioB - Wednesday, March 4, 2015 - link
So I ask again, if it is x86, what's the contribution of Rockchip to it? Intel already knows how to make an Atom with PowerVR. Building it at 28nm may be cheaper than using its own 22nm, but surely not at 32nm. And seen all the fuss about the 32nm Intel quality against competing 28nm, Intel (od and now cheap) 32nm should still perform better than TMSC PP. Without speaking about the fact that producing a x86 core outside its foundries exposes its architectural secrets to the rest of the market.Then, he fact that the chip is restricted to Asian market (probably because in cheap Chinese product sponsorized by Rockchip itself) still raises more than eyebrow.
djvita - Monday, March 2, 2015 - link
will the x5/7 support more than 2Gb RAM? if not, i'm keeping my venue 8 pro for longer....better to save for a surface in that casekyuu - Monday, March 2, 2015 - link
For some reason this article doesn't show the slide that details the x5/x7 specifications. Link here: http://cdn.arstechnica.net/wp-content/uploads/2015...Basically, the low-end x5 only supports up to 2GB of RAM, but the higher end x5 and the x7 model both support up to 8GB. Hopefully we only see higher-end x5 and x7 models in OEM designs.
nevertell - Monday, March 2, 2015 - link
When will they start using their own GPU's in these low power SoCs ? What is it with these gpus Mali and their superbad linux drivers ?rootheday3 - Monday, March 2, 2015 - link
Intel uses own GPU for Baytrail and Cherrytrail (now x5, x7). Mali is being used for very low end die area, perf, power (and somewhat reduced feature set eg 3d/compute APIs)jasperjones - Monday, March 2, 2015 - link
When can we expect to see Intel NUCs based on the Atom x5 or x7?Other question: Aren't the x5 and x7 roughly in the same thermal/performance bracket as Core-M? Do any mini PCs (NUCs or other) exist with Core-M?
BMNify - Monday, March 2, 2015 - link
given that intel are using the Mali GPU's then when can we expect the user space FLOSS driver code and documentation to be written and made available ?also what SIMD will these low power Mali GPU equipped intel CPU's use ? AVX2 or the better NEON SIMD, with neither current SIMD then its not worth the trouble for the enc consumer to invest there now with octacore 64bit/128 bit NEON coming around now on mass
toyotabedzrock - Monday, March 2, 2015 - link
So they don't have an up to date LTE or eMMC implementation and the LTE is connected via USB 2?Intel needs a kick in the @$$ to get them to actually compete for market share in this segment.
aryonoco - Tuesday, March 3, 2015 - link
That feeling when you have to read everything Ian writes 3 times to understand sentences like this:"Firstly is the use of 28nm is the same node as previous Intel Atoms, and thus should be derived from a TSMC source."
chimeM - Thursday, March 5, 2015 - link
All this fairly recent modem and peripheral integration aside, I could never really care much about the atom line.It is not a product line where the company wanted to make the best possible chip at a low power budget. It always seemed like intel wanted to make some chips good enough to capture the low power device market, at the same time ensuring them to be crappy enough to show off how fast their main line of chips (now broadwell) are. No wonder they still haven't captured the market quite.