Intel SoFIA & Broxton Killed

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Apr 30, 2015
131
10
81
ARM have sold 4 licences for the Artemis processor; this is a 10nm product, following the A72, and maybe up to 50% more powerful, drawing up to 750 mW; they have also sold 4 POP IP licences for Artemis on 10nm. The announcements were low key in the 2015 Q2, Q3, Q4 results, and the 2015 Analyst and Investor conference.
I would think that the next battle will be over AIO and notebooks, with Artemis v. Core-M.
The licences may have been for Mediatek, Qualcomm, and possible Samsung and Huawei, at a guess.
There are successors to Artemis in the pipeline.
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,784
136
Huawei? Do they produce an ARM core? Huawei has signed a lot of deals with Intel as of late, even going so far as to take free Cherry Trail boards from Intel for every Core M laptop that they build.
 

NTMBK

Lifer
Nov 14, 2011
10,208
4,940
136
Huawei? Do they produce an ARM core? Huawei has signed a lot of deals with Intel as of late, even going so far as to take free Cherry Trail boards from Intel for every Core M laptop that they build.

Huawei owns HiSilicon.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
it's die area and feature set is probably too beefy. SoFIA is probably because they don't want to compete with the low-end, low-margin stuff. Not interesting.

They'd like to see their Core chips that way. It simply isn't good as the claims.

Core chips are using too much power as a platform and not integrated enough for its performance, and especially for its cost.

Atom chips are too expensive for its performance. Current Atom X5-8500 should have been what's aimed at SoFIA, now. Then you'd have Broxton at 1.5-2x performance. And if anyone's confused, that's for CPU.

You're right - it's always kind of been assumed that this was about x86 and Atom but Apple could have been interested in XScale.
Somebody was saying that they never had chance with XScale. The division was apparently plagued with setbacks and delays. Something is more deeply wrong than just technical issues. Some people say its the culture at Intel.

Look at it this way. It isn't "x86's fault" that they are delayed with majority of 14nm products and subsequently underperformed.

-What if rather than Haswell Refresh, we had Broadwell at Q2 2014?
-What if SoFIA was indeed late 2014, with SoFIA 14nm Q2 2015?
-What if Knights Landing was officially released Q3 last year?
-What if Broxton/Apollo Lake was Q2 last year?

And delays defeature products. Add to all the above, what if the performance turned out to be better too? Was it a coincidence that Broadwell-C and Skylake K had availability issues? Was it a coincidence that Broadwell Core M sucked as a product? What about SoFIA? What about Airmont?
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
There is no giving up on tablets: both iOS and Android will use the tablet as an entrance vector into the productivity space. Unless Intel has a clear strategy to deliver better perf/watt in passively cooled devices with lower platform cost than today's Core M, they'll be giving up on more than tablets.

Tablets are in free fall to say it mildly. They are on a fast track out of this world. They are not going to takeover anything.

http://www.idc.com/getdoc.jsp?containerId=prUS41218816
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Smartphone market isn't growing either. Apple sales are down as well. Just like with the PCs, market is over-saturated with devices.

Nope. People dont have money.

Let me give you a perspective.
US companies now got lower profits 3 quarters in a row. And they lost revenue 5 quarters in a row. Way beyond PCs, tablets and phones.
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
I wonder how much of this is just 14nm sucking so badly at Intel. Intel was overconfident with its 22nm success (which was a great node), with first wafers being developed in house in 2009 and displayed at IDF also in 2009 (products did not ship till 2012) so they started thinking about buying more RF properties announcing their intention to purchase Infineon in 2010 and completing the purchase in 2011.

Infineon kept its future products on non Intel Fabs with the 14nm line being the first intended Infineon rollout of modems and modem+soc.

But unlike 22nm which had amazing yields and introduced FinFets, 14nm was a dog and heavily delayed due to the yields being so shitty and that Intel had to spend an extra year to fix the yield problems. It was supposed to come out in 2013, with the Intel Ceo announcing at IDF 9/2013 showing a working ultrabook and announcing 14nm will be shipping in the next quarter, but instead of shipping 1 quarter later it came come out in quarter 4 2014 and volume only appeared to occur in 2015.

14nm was so delayed that the desktop i7s of broadwell and skylake came out within only a month with each other in the retail channel.

-----

Making modems is hard stuff for its not something that scales well with die shrinks, and having your first attempt on new companies process be intel 14nm would not be something I want to design on.
 

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
Nope. People dont have money.

Let me give you a perspective.
US companies now got lower profits 3 quarters in a row. And they lost revenue 5 quarters in a row. Way beyond PCs, tablets and phones.

On the surface this also puts a crimp in Intel CEOs designs on being a major IoT supplier. IoT devices need wireless communication as much as phones do.

Are they killing wireless modem and Atom off for a few years to reduce the embarrassment of spending 100s of millions to billions in incentives and not making much headway? Will there still be some skunkwork teams kept from this to work on low power wireless and sub 1W SoCs?

Edit: Might just be, Quark team is in Ireland. Wouldn't be surprised if they are using these layoffs to shift jobs from production focused R&D in US to future product focused R&D in Ireland (Quark team) and Israel (LAN and WiFi teams).
 
Last edited:

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
On the surface this also puts a crimp in Intel CEOs designs on being a major IoT supplier. IoT devices need wireless communication as much as phones do.

Are they killing wireless modem and Atom off for a few years to reduce the embarrassment of spending 100s of millions to billions in incentives and not making much headway?

From the anandtech article

Earlier this decade Intel used a ‘contra-revenue’ strategy, investing into OEMs that would buy their chips, causing operating losses for the mobile division of $3.1 billion in 2013 and $4.2 billion in 2014 with a much lower revenue stream

http://www.anandtech.com/show/10288/intel-broxton-sofia-smartphone-socs-cancelled

Now it is not so obvious that this 7 billion was wasted for the stock price would be hammered if they were not doing external things to chase markets of growth such as when tablets were in 2012, 13, and 14. Only in 15 and 16 have we seen that this market is shrinking and not the next big growth market.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
But unlike 22nm which had amazing yields and introduced FinFets,

None of their products showed us 22nm was amazing though. Bay Trail was eventually contra-revenued. Ivy Bridge with less overclocking and 100MHz higher stock clock.

The 37% transistor gain was only applicable to Bay Trail, and non-22nm ARM SoCs reached Atom speeds pretty easily. The process gain was only in theory because the products were all mediocre. Call me cynical but no one said "22nm is awesome!" If anything it made us all realize that they are sacrificing PC for mobile and still going nowhere in mobile.

It was supposed to come out in 2013, with the Intel Ceo announcing at IDF 9/2013 showing a working ultrabook and announcing 14nm will be shipping in the next quarter, but instead of shipping 1 quarter later it came come out in quarter 4 2014 and volume only appeared to occur in 2015.
2012 Q2 was 22nm. 14nm was NEVER Q4 2013. If it did not delay it would have been Q2 2014.

Of course, I'm not disagreeing in general, but you have to learn a realistic schedule for process introductions. 2 years is really the bare minimum, because manufacturers only accept absolute minimum of 1 year per "generation", whether Tick or Tock.

Up until recently, when Intel said production year, that means when the products were available for users to buy. Over time, as CPUs got more complex, that terminology was shifted. 2001 - 0.13u, 2004(Feb) - 90nm, 2005(second last day of 2005) - 65nm, 2007 - 45nm, 2010(actually early first month 2010), 32nm, 2012: 22nm, 2014: 14nm

They still follow the old terminology when they say "production". therefore 2007 - 45nm, 2009 - 32nm, 2011 - 22nm, 2013 - 14nm, 2015(now 2016) - 10nm
 
Last edited:

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
The chance to get into mobile slipped more than 10 years ago when they said no to apple.

When Apple and Samsung own soc hit market there was absolutely zerro chance for Intel to be in the market. The thought of your competitor buying your product is crazy.

They were never close. Even if Atom had been better and LTE/3G solution of good quality integrated it was meant to fail.

Secondly Intel process tech is tuned for another product type. They are competing with arm eco system in a red ocean. With an organization tuned for quality and performance not price. The tune of 4b a year.

And add to it, nobody could give a solid reason for Intel to be in mobile. It was weak reasons.

It was so damn obvious it would fail. Still all the talk is not about the above reasons but about atom and the design and execution of the product stack. How ironic even after Intel fall flat on mobile people dont seem to grasp the real reason.

I have taken so much agressive flak in these forums for saying this over the years its incredible.

Its all fuelled by reviewers like Anand telling the same crap from Intel marketing. I remember one of the most sad pieces in years. Do you remember the one about an Intel engineer that just happen to visit Anandtech, with a Voltmeter, learning us ohms law, and busting the x86 power myth?
 

lefty2

Senior member
May 15, 2013
240
9
81
On the bright side, this means contra revenue is finished too. We should see a surge in profits as $4 billion is added to their bottom line.
 

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
On the bright side, this means contra revenue is finished too. We should see a surge in profits as $4 billion is added to their bottom line.
Absolutely. As a sideeffect the 4b a year killed eg beema and denver. We will never know if any of those would have been viable product.

Stupid brute force killing innovation. Old style monopoly otellini thinking. Except this time they were facing apple and samsung. They brought a knife to a fight with heavy armor.

There is so much technological muscle in Intel its incredible. World class engineering all over all the time. Seeing it wasted on this "me too" strategy is just sad. Glad its over.

Now instead of thinking how to improve profit for their shareholders or some weak long term strategy they might try the idea of thinking how to improve consumers experience and the business for companies. Not how to get most money from them. Entertaining that idea might bring them to a new place.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Absolutely. As a sideeffect the 4b a year killed eg beema and denver. We will never know if any of those would have been viable product.

Denver was the only one of the two with even a remote distant option to be a viable product. In short, no.
 

JDG1980

Golden Member
Jul 18, 2013
1,663
570
136
Look at it this way. It isn't "x86's fault" that they are delayed with majority of 14nm products and subsequently underperformed.

-What if rather than Haswell Refresh, we had Broadwell at Q2 2014?
-What if SoFIA was indeed late 2014, with SoFIA 14nm Q2 2015?
-What if Knights Landing was officially released Q3 last year?
-What if Broxton/Apollo Lake was Q2 last year?

And delays defeature products. Add to all the above, what if the performance turned out to be better too? Was it a coincidence that Broadwell-C and Skylake K had availability issues? Was it a coincidence that Broadwell Core M sucked as a product? What about SoFIA? What about Airmont?

Sounds like the common problem was 14nm, not x86. Intel's 14nm node was late, with crappy yields, and probably less performance than Intel expected. It's slightly denser than Samsung/GloFo and TSMC FinFET - maybe a half node ahead - but Intel clearly ran into the same problems with node shrinks that the other foundries did, despite a lot of hype to the contrary.
 

JDG1980

Golden Member
Jul 18, 2013
1,663
570
136
Tablets are in free fall to say it mildly. They are on a fast track out of this world. They are not going to takeover anything.

http://www.idc.com/getdoc.jsp?containerId=prUS41218816

Looks like you just read the first paragraph of the article and ignored the rest. For Intel (and AMD), the relevant part is this:

"With the PC industry in decline, the detachable market stands to benefit as consumers and enterprises seek to replace their aging PCs with detachables."

By "detachables" they mean products like the Surface Pro. These still need compatibility with legacy Windows applications, so they need to be x86. And since they are full-fledged PCs, they can sustain higher profit margins on the CPU.

Intel wants 60% gross margins, and that was never going to happen on low-end commodity "slate" tablets or smartphones. And it won't happen on IoT devices either. Their mobile strategy was doomed from the start.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Beema has chances too since it can run Windows... You fanatism is incredible.... At least admit that Intel sucked hard on mobiles.

The Discovery tablet wasn't even sold. And it was not worth to write home about. And it wasn't due to Atom, but the poor characteristics by Beema.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Looks like you just read the first paragraph of the article and ignored the rest. For Intel (and AMD), the relevant part is this:

"With the PC industry in decline, the detachable market stands to benefit as consumers and enterprises seek to replace their aging PCs with detachables."

By "detachables" they mean products like the Surface Pro. These still need compatibility with legacy Windows applications, so they need to be x86. And since they are full-fledged PCs, they can sustain higher profit margins on the CPU.

Intel wants 60% gross margins, and that was never going to happen on low-end commodity "slate" tablets or smartphones. And it won't happen on IoT devices either. Their mobile strategy was doomed from the start.

How much did the tablet market decline, and how much did the PC market decline?
 

NTMBK

Lifer
Nov 14, 2011
10,208
4,940
136
The Discovery tablet wasn't even sold. And it was not worth to write home about. And it wasn't due to Atom, but the poor characteristics by Beema.

We know almost nothing about Beema's characteristics. :/ We do know that Intel made negative revenue (!) on its mobile division, I.e. it paid more in contra revenue payments than it made in sales. Against that sort of practice, AMD could have offered a fantastic product and still been unable to compete. Intel was burning billions of dollars, and AMD didn't have the finances to do that.
 

zentan

Member
Jan 23, 2015
177
5
36
Based on which criteria?

Design wins perhaps? :)
At least Tegra line was seeing mainstream design wins. Tegra 3 in Asus Nexus 7'' , Xolo play tab, etc. Tegra K1 in HTC Nexus and Xiaomi Mi Pad. Nvidia moved away because they couldn't see enough margin or otherwise losses,in a way it could not compete.

One only has to look at the Chinese market for reference.
Lots of Core-M based tablets and 2-in-1s starting as low as ~340$. Some links given below

http://www.*************/tablet-pcs/pp_332586.html Teclast X3 Pro 2 $477.13
11.6" 8GB RAM 128GB SSD
Skylake Core m3-6Y30

http://www.aliexpress.com/store/pro...0-1080-8GB-128GB-HDMI/104718_32659083868.html


http://www.*************/tablet-pcs/pp_312556.html Cube i9 531.54$
12.2" Skylake Core M3-6Y30 Dual Core 1.51GHz 4GB RAM 128GB ROM Dual Cameras Type-C

http://www.*************/tablet-pcs/pp_246302.html Onda V919 3G Core M 9.7" 357.92$

Windows 8 + Android 5.0 Broadwell Core M 5Y10 64bit Dual Core 2.0GHz 9.7 inch
4GB RAM 64GB ROM

http://www.*************/tablet-pcs/pp_263828.html Onda V116w Core M 3G $407.18
11.6" Windows 10 broadwell core-m 5Y10C 64bit Dual Core 2.0GHz 4GB RAM 128GB ROM


http://www.*************/tablet-pcs/pp_252601.html Teclast X2 Pro Broadwell Core M 5Y10 350$
http://www.aliexpress.com/store/pro...0-1080-8GB-128GB-HDMI/104718_32659083868.html

http://www.*************/tablet-pcs/pp_227132.html Cube i7 $344.54
Broadwell Core M 1.0GHz 4GB RAM 64GB ROM FHD Screen

EDIT: gearbestdotcom is replaced by '*'s for some reason

As far as it is known Core-M tabs weren't falling under the "Contra-Revenue" .
Other lower performing chips had to get 7-9" tabs mostly which was always a struggle(not to say contra-revenue wasn't having any effect). Plus the tablet market didn't prove to be a very high profit market for most small and medium OEMs in recent times.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
We know almost nothing about Beema's characteristics. :/

They actually were not competitive from a power use standpoint. They were on par or slightly better compared to Haswell but Bay Trail used 1/2 the power of Haswell as a platform.

http://www.anandtech.com/show/7974/...hitecture-a10-micro-6700t-performance-preview

Kabini was claimed by AMD to achieve 11 hour battery life on a 45WHr battery in idle, 9 hours in web browsing, and 7 hours in 720p video playback: http://images.anandtech.com/doci/6976/Screen%20Shot%202013-05-23%20at%2012.24.51%20AM_575px.png

http://www.anandtech.com/show/7974/...hitecture-a10-micro-6700t-performance-preview

Haswell: http://www.blogcdn.com/www.engadget.com/media/2013/05/intel-10.jpg

Comparing the Kabini's results from both of AMD's slides, Kabini's results are roughly comparable. 20-30% improvement on an SoC level AMD was claiming wouldn't have changed anything. The overall power management technology is on par with 2010's Oak Trail platform.

My Dell Venue Pro 8 Tablet achieves ~8 hour battery life from a meagre 15WHr battery. The entire platform with screen on idles at 1.4W. You'll see from Atom Tablet reviews that 1080p HD or 1440p HD screen devices achieve 8-10 hour on a 30WHr battery. They are at a different class.
 
Last edited: