Intel's smartphone platforms

[tangey]

I guess it's Z2580, not Merrifield.

Indeed, given the roadmap from a while back.
http://www.cnx-software.com/2012/05...field-mobile-processor-microservers-and-more/

The latter is supposed to come with Intel IGP, while thhe former is coming with SGX 544MP2.

Any citation for the above? Baytrail (the nextgen tablet platform) has been widely cited as going with IGP, but I have not seen a single suggestion that the smartphone platform is moving from Imagination.

 

[Laurent06]

You are right, that was Bay Trail I had in mind, sorry.

 

[liolio]

A new piece from Anandtech

Interesting read. I'm surprised by the a15 power consumption, it spike really high, in extended (say a demanding game) use I would expect the tablet to turn pretty hot.
For me it looks like as far as phone as concerned the best products are Qualcom Krait, Intel Atom and Apple swift. Out of those three I would put swift on top but mostly because of the GPU (PowerVR perfs and power consumption are impressive). Wrt to the CPU it is a wash between Apple and Intel.

I was thinking that may be Intel should embrace ARM (again), but I strongly doubt it now.
ARM has nothing better than A15 anytime soon, I think that this fall is going to hurt if Intel delivers with its new Atom.
New processor, new process, I expect them to actually let the competition behind 8O

 

[Mintmaster]

Some good work by Anand on the power consumption measurements there. He's making some assumptions that those voltage rails really do represent the whole story, but they're probably correct.

There were some roadmap leaks recently that suggest the 22nm Silvermont core is delayed until 2014, but even so, Intel's current lineup is clearly quite competitive. At this point, it doesn't really matter for ARM or Qualcomm, as they're not getting any significant sales from Win RT, and an x86 version of Android isn't going take any marketshare away from them either.

It does make Win RT basically obsolete, though. The only advantage is cost savings from MS pricing RT lower and Intel pricing Clovertrail high (if they are choosing to do so). Other than that, there are zero compromises to be made in wanting full x86 compatibility in a tablet/hybrid, which is great news. My guess is the current price delta between ARM and x86 tablets (all else being equal) will be minimal in six months.

 

[Jubei]

A new piece from Anandtech

I was thinking that may be Intel should embrace ARM (again), but I strongly doubt it now.
ARM has nothing better than A15 anytime soon, I think that this fall is going to hurt if Intel delivers with its new Atom.
New processor, new process, I expect them to actually let the competition behind 8O

Leave the competition behind in what market?

 

[ltcommander.data]

It does make Win RT basically obsolete, though. The only advantage is cost savings from MS pricing RT lower and Intel pricing Clovertrail high (if they are choosing to do so). Other than that, there are zero compromises to be made in wanting full x86 compatibility in a tablet/hybrid, which is great news. My guess is the current price delta between ARM and x86 tablets (all else being equal) will be minimal in six months.

That begs the question of why Microsoft went to the effort of developing Windows RT? As an OS vendor they no doubt have visibility and access to everyone's SoC, ARM and x86, even more so now they are developing first-party devices. They would know early on that ARM and x86 performance and power consumption were converging and how soon that would be occurring relative to the Windows 8/RT release timeframe. If they wanted a low-cost licensing, touch/Metro focused OS option, they could have offered a version of x86 Windows 8 with those restrictions and cost structure like previous netbook/Starter edition programs instead of developing Windows RT.

 

[rpg.314]

That begs the question of why Microsoft went to the effort of developing Windows RT? As an OS vendor they no doubt have visibility and access to everyone's SoC, ARM and x86, even more so now they are developing first-party devices. They would know early on that ARM and x86 performance and power consumption were converging and how soon that would be occurring relative to the Windows 8/RT release timeframe. If they wanted a low-cost licensing, touch/Metro focused OS option, they could have offered a version of x86 Windows 8 with those restrictions and cost structure like previous netbook/Starter edition programs instead of developing Windows RT.

MS didn't think Intel could compete. Or at least be competitive in early 2013.

 

[rpg.314]

Some good work by Anand on the power consumption measurements there. He's making some assumptions that those voltage rails really do represent the whole story, but they're probably correct.

There were some roadmap leaks recently that suggest the 22nm Silvermont core is delayed until 2014, but even so, Intel's current lineup is clearly quite competitive. At this point, it doesn't really matter for ARM or Qualcomm, as they're not getting any significant sales from Win RT, and an x86 version of Android isn't going take any marketshare away from them either.

It does make Win RT basically obsolete, though. The only advantage is cost savings from MS pricing RT lower and Intel pricing Clovertrail high (if they are choosing to do so). Other than that, there are zero compromises to be made in wanting full x86 compatibility in a tablet/hybrid, which is great news. My guess is the current price delta between ARM and x86 tablets (all else being equal) will be minimal in six months.

Now all we need are good touch+keyboard+stylus devices with Haswell/Broadwell.

 

[liolio]

Leave the competition behind in what market?

According to you? Were we discussion HPC or servers chips here?

 

[Jubei]

According to you? Were we discussion HPC or servers chips here?

No we are discussing smartphones here. But im assuming thats not the market you were talking about since Intel doesnt stand much of a chance of becoming dominant in phones

 

[Silent_Buddha]

No we are discussing smartphones here. But im assuming thats not the market you were talking about since Intel doesnt stand much of a chance of becoming dominant in phones

Right now? Absolutely not. In the next 1-2 years, unlikely. 5-10 years from now? Possibly.

It's hard to say what the landscape will look like in 5 years. It all depends not only on how well Intel executes, but how well ARM and its partners execute.

1 year ago if you someone had said Intel would be able to release a chip and platform that would allow for x86 tablets the same size and weight as an iPad with competitive battery life and similar CPU performance, most would have laughed at the idea.

Yet here we are with x86 tablets the size and weight of an iPad with competitive CPU performance and battery life. Graphics performance is still far behind, but how long before Intel uses a competitive PowerVR graphics core for its mobile devices.

Of course, none of this means ARM is standing still. So as mentioned, it all depends on how the various parties execute going forward. At least ARM, unlike AMD when they had the performance lead over Intel, is well funded.

Regards,
SB

 

[liolio]

No we are discussing smartphones here. But im assuming thats not the market you were talking about since Intel doesnt stand much of a chance of becoming dominant in phones

Well that is your take. As performances are getting higher and higher so is the bar to keep up with the best.
It seems Texas Instrument has been mostly pushed out. I'm not sure about what is happening STM Erickson but I would say they are next.
Thanks to anandtech review we see that A15 are not really a fit for smart phone. They have great perfs, but overall for embedded devices Krait, Swift and Atom are better fit.
Samsung is stuck with A15 with no massive improvement in power efficiency in sight, neither they have a custom design.
I expect Nvidia to be no different that Samsumg wrt CPUs as they custom cores are not set to be release any time soon. They will have to use either A15 or A7, or possibly stick to high clock speed A9.

Be it Apple with its Swift or Qualcom and Krait, I expect not neat improvements till they jump to another lithography (as far as CPU performances are concerned).

On the other end fall 2013 Intel will have another architecture, design on their 22nm process, they are already pretty competitive. I expect them to take the performance lead this fall (2013) with their new Atom, that is what I mean by letting the competition behind. Another not leveraged advantage is support for 64bit memory space, Msft could have interest for example in having tablets with 4 GB of RAM (or more).

I don't see how they could not take the performance crown this fall (GPU perfs are another matter, and only apple seems to care that much ), and let the competition behind (as their lithography advantage is still not set to disappear).
I don't imply that ARM CPU are to disappear any time if ever, that is not the point.

 

[Jubei]

Right now? Absolutely not. In the next 1-2 years, unlikely. 5-10 years from now? Possibly.

It's hard to say what the landscape will look like in 5 years. It all depends not only on how well Intel executes, but how well ARM and its partners execute.

1 year ago if you someone had said Intel would be able to release a chip and platform that would allow for x86 tablets the same size and weight as an iPad with competitive battery life and similar CPU performance, most would have laughed at the idea.

Yet here we are with x86 tablets the size and weight of an iPad with competitive CPU performance and battery life. Graphics performance is still far behind, but how long before Intel uses a competitive PowerVR graphics core for its mobile devices.

Of course, none of this means ARM is standing still. So as mentioned, it all depends on how the various parties execute going forward. At least ARM, unlike AMD when they had the performance lead over Intel, is well funded.

Regards,
SB

If you want me to agree that anything is possible sure. But looking at the situation it doesnt seem very realistic. The two biggest smartphone vendors also happen to design their own chips wich cuts out a middleman like Intel, in Samsungs case they also have their own foundry so they really have no need for Intel

Then you have all these upcoming chinese companies using cheap ARM SOCs taking over the low range segment

Its not about performance but margins and flexibility.

 

[Jubei]

Well that is your take. As performances are getting higher and higher so is the bar to keep up with the best.
It seems Texas Instrument has been mostly pushed out. I'm not sure about what is happening STM Erickson but I would say they are next.
Thanks to anandtech review we see that A15 are not really a fit for smart phone. They have great perfs, but overall for embedded devices Krait, Swift and Atom are better fit.
Samsung is stuck with A15 with no massive improvement in power efficiency in sight, neither they have a custom design.
I expect Nvidia to be no different that Samsumg wrt CPUs as they custom cores are not set to be release any time soon. They will have to use either A15 or A7, or possibly stick to high clock speed A9.

Be it Apple with its Swift or Qualcom and Krait, I expect not neat improvements till they jump to another lithography (as far as CPU performances are concerned).

The solution for smartphones have always been ARMs big.Little technology. Two or four Cortex A7s for low power tasks and Cortex A15 only when requiring more CPU power. Samsung has an 8 core version in the pipeline already at 28nm. So i dont see how you can claim they are "stuck" in power efficiency when they have both a half node jump and low power processor available

Also there is no official word when next Atom is coming, the recent leaked roadmap says feb 2014 wich is pretty far away right now

 

[Mintmaster]

That begs the question of why Microsoft went to the effort of developing Windows RT?

I think it's primarily a low cost backup plan. By all reports Clovertrail supply is still tight, and there must have been some doubts about whether Intel could deliver, whether Intel even wanted Clovertrail to succeed since it can replace the sales of higher margin products, or whether A15 was going to be low power nirvana that would leave x86 tablets at a disadvantage. The ARM kernel is already there from WP8, and the UI/API is identical to W8, so I doubt it really cost them much.

The other thing is that RT is basically a "we can do it too" message from MS. It's proof they can do everything iOS/Android can and more, including a similarly secure/controlled marketplace, low power (true multitasking probably makes ultra-low power idle states a bit harder to get into), and usable productivity. The technical crowd probably never doubted it, but for the rest it answers those questions.

In the end it's just a way to make sure MS doesn't have all its eggs in one basket.

 

[Mintmaster]

The two biggest smartphone vendors also happen to design their own chips wich cuts out a middleman like Intel, in Samsungs case they also have their own foundry so they really have no need for Intel

What's interesting about that is Samsung is dominating the "productive smartphone" niche with its Galaxy Note series.

Looking at tablets, I think it's all but given that the Note tablets are going to be outsold by Win8 tablets with a stylus, because a stylus basically lets you use any legacy application without carrying around a mouse and using a table. Samsung knows this, so their Win8 tablets include stylus ability, despite competing with their own Note 10.1.

I think high end smartphones will eventually go the same way. I have a Galaxy Note 2, and when using Jump Desktop, the stylus works brilliantly in controlling remote PCs. Hover moves the mouse, tapping is left click, and pushing the button is right click. The next step, of course, is to have x86 applications run natively on the phone along with PadFone-like docking ability. What else are we going to do with the mobile CPU power that will be at our disposal 2+ years from now?

Under such a scenario, Samsung will be more than willing to use x86 chips in high margin superphones.

 

[iMacmatician]

Silvermont/next gen atom will reportedly bring much improved ipc, thanks to either full OoO execution or part OoO...and likely allow even greater clock speeds...

What's the difference between "part" OoO and "full" OoO? I've never heard of the former.

 

[liolio]

The solution for smartphones have always been ARMs big.Little technology. Two or four Cortex A7s for low power tasks and Cortex A15 only when requiring more CPU power. Samsung has an 8 core version in the pipeline already at 28nm. So i dont see how you can claim they are "stuck" in power efficiency when they have both a half node jump and low power processor available

Well I'm iffy about this big/little thinkg, ARM has a market for both A15 and A7, and it is convenient that they can run exactly the same code.
A quad core A7 should be cheap ti=o produce is really low power but for the costumer pov it doesn't bring anything on the table A9 could not deliver. Having such cores doing the house keeping could be interesting but I wonder how much power house keeping burns anyway.
Looking at competing design it seems that neither Qualcom nor Apple were comfortable going with a chip as wide/big as A15 on the processes readily available.
Looking at Anandtech tests, one may wonder if really low power cores but also sucky and slow are better at anything including house keeping. It seems that modern CPU can wake up fast get rid of whatever house keeping there is do in a blink and go to sleep again.
Thing is there is a market for A7 it allows an increase of perfs in the lower segment of the market.
But I'm not sure about the extend it can help A15 to to stay within a sane TDP for say a phone.

Samsung is to go for 28nm? I did not know, still it won't change much. The interesting thing would if they could somehow pass on 22nm and try to be on the same schedule as Intel for 14nm+finfet.
I wonder if Samsung is to develop its own ARM cpu, that would be interesting for sure.

Also there is no official word when next Atom is coming, the recent leaked roadmap says feb 2014 wich is pretty far away right now

As I was a bit surprised by your statement and made a search to find that indeed some leaks (from a few days ago) implies that the new atom are significantly delay. ~6months.
That makes a differences the latter they are the higher the odds for competing parts using 22nm processes. That could indeed makes a difference, but forgive me the news was indeed pretty new.
With this fact taken in account I would be more wary about Intel taking the performance crown for a significant period.
My pov was without that (new) information, if they have released their new Atom by Q3 early Q4 I still fail to see how they would not have take the "crown" for a pretty significant period of time.

 

[Jubei]

I would have to disagree about Cortex A7. Why would a vendor go for a much bigger powerhungry core like A9 when A7 offers the same performance with 40-60% less power draw at a fraction of the size. If anything i think Cortex A15 has gotten too much attention, the A7 is a fantastic core

I agree it remains to be seen just how well a big.Little implementation will work out in reality but the strategy is there

 

[Jubei]

What's interesting about that is Samsung is dominating the "productive smartphone" niche with its Galaxy Note series.

Looking at tablets, I think it's all but given that the Note tablets are going to be outsold by Win8 tablets with a stylus, because a stylus basically lets you use any legacy application without carrying around a mouse and using a table. Samsung knows this, so their Win8 tablets include stylus ability, despite competing with their own Note 10.1.

I think high end smartphones will eventually go the same way. I have a Galaxy Note 2, and when using Jump Desktop, the stylus works brilliantly in controlling remote PCs. Hover moves the mouse, tapping is left click, and pushing the button is right click. The next step, of course, is to have x86 applications run natively on the phone along with PadFone-like docking ability. What else are we going to do with the mobile CPU power that will be at our disposal 2+ years from now?

Under such a scenario, Samsung will be more than willing to use x86 chips in high margin superphones.

Sure but there is a major advantage in going with X86 for Windows 8, it gives you legacy applications. What advantage would Samsung have in using X86 processor for Android?