Mobile World Congress 2015

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
IDF 2007.
You can't take Paul Otellini's mobile thing seriously, can you? If he had been serious, he had put Atom in Tick-Tock mode from the beginning, he would have invested in low BOM and across the board (low to high-end) offerings, he would have bought Infineon earlier, he would have invested in 22nm SoC flavor from the beginning and put Atom on bleeding edge node to exploit their process technology, he would have switched the modem to Intel's fabs from the beginning, he would have done all the things that Intel is now doing in 2014-2015.

So please wake me up in Q1'16 when 14nm Atom (Broxton + SoFIA) is launched as it should have been in Q4'14 (notice how Intel's smartphone strategy is 1.5 years behind where it should have been, and will hopefully still be competitive). Until then, I don't see why I should get excited about Atom -- the testament of Intel's pre-"if anything computes and is connected, it is best with Intel" era.
 

III-V

Senior member
Oct 12, 2014
678
1
41
They were aiming for 2.7GHz. Not that this would have done much to fundamentally change the competitive landscape, but the lack of uplift is pretty bad. Parametric yield issues?
I'm trying to nail down what's going on with 14 nm. I think neglecting the high-power part of the spectrum is biting them in the rear.

For everything sub 0.7 V, 14 nm is a remarkable improvement over 22 nm. The only area they've regressed, according to their paper, is with DIBL. For everything over 0.7 V... well... that data is omitted.
 
Last edited:

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
I'm trying to nail down what's going on with 14 nm. I think neglecting the high-power part of the spectrum is biting them in the rear.

For everything sub 0.7 V, 14 nm is a remarkable improvement over 22 nm. For everything over... well... that data is omitted.

So Intel can't make a high performance process anymore? I'm going to wait until BDW-K at IDF before jumping on any bandwagon.
 

III-V

Senior member
Oct 12, 2014
678
1
41
So Intel can't make a high performance process anymore? I'm going to wait until BDW-K at IDF before jumping on any bandwagon.
It's not that they can't. It's that they didn't. They didn't spend the money doing so, because it was perceived as not being worth it. That may have been a costly mistake on their part... hopefully it is being rectified with 10 nm, but I don't have much faith. They did make some moves explicitly for high performance, though -- like the thick metal layers they added -- but so far, 14 nm has not translated to the real world gains you'd think there'd be.

I do not have high hopes for BDW-K at all, despite my previous, outspoken optimism.
 
Mar 10, 2006
11,715
2,012
126
So please wake me up in Q1'16 when 14nm Atom (Broxton + SoFIA) is launched as it should have been in Q4'14 (notice how Intel's smartphone strategy is 1.5 years behind where it should have been, and will hopefully still be competitive). Until then, I don't see why I should get excited about Atom -- the testament of Intel's pre-"if anything computes and is connected, it is best with Intel" era.

We'll see. Still pretty bummed that Intel delayed Broxton.
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
Looking at the performance, it does look like Intel will not be able to sell even the Z8700 into high-end. If pricing is right, they might still be able to shift a few SoCs, in particular into the Windows ecosystem, where generally x86 seems to be preferred over ARM (Windows RT).
What we currently are seeing is the ARM competition closing the technology gap (moving from planar to 3d) Given the architectural advantages of ARM over x86 it is easy to see that a narrow technology gap is dangerous for Intel.
 

ninaholic37

Golden Member
Apr 13, 2012
1,883
31
91
Atom may be forever the cheapo-brand.
I think this was a special case, where they needed to make Cherry Trail cost way less than Bay Trail so they don't lose money again in mobile. I'm not surprised if it doesn't perform much better, their goal was probably just to make it perform just as good and cost less money to make, then maybe work on performance after they get the costs down. Considering how much they lost on Bay Trail-T, nailing both lower cost and higher performance in one round seems a little out there. That's my theory anyway, feel free to rip it apart if you believe otherwise.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
So Intel can't make a high performance process anymore? I'm going to wait until BDW-K at IDF before jumping on any bandwagon.

Broadwell-K AFAIK is a 65W which is probably a LGA version of Haswell R chips' GT3e GPU and clock pretty low for the CPU. Considering how Skylake-K is coming at the same time with GT2 GPU instead.

The problem across their entire product line indicates the issue is due to the process.

SA got it right.

http://semiaccurate.com/2014/07/11/intel-castrates-broadwell-gutting-performance/

I don't think even Skylake will turn out to be the panacea that's hyped out to be. If we take the best-case scenario of 20-30% gain in Y class talked about here in the forum by a Chinese guy, that's still pretty poor. 2.3 Cinebench R11.5 score on a brand-spanking-new top-of-the-line 2016 product. Whoopee-doo!

You can't take Paul Otellini's mobile thing seriously, can you?
Who cares? They screwed up no matter what. Brian K got elected more than 1.5 years ago, and surely had bigger responsibilities for a longer time. When Paul O. was doing right in his days with Core 2, we saw the benefits immediately. Company as overall is tanking.

For everything sub 0.7 V, 14 nm is a remarkable improvement over 22 nm. The only area they've regressed, according to their paper, is with DIBL. For everything over 0.7 V... well... that data is omitted.
What product runs at 0.7V for its "performance" frequency? Pretty much nothing if you see product voltage ranges. That means whatever performance gains touted in 22nm and 14nm was all hype because in the ever more important 1+ volts it had zero gains.

leading CPU and GPU performance with Broxton -- moving to 10nm in H2 -- and modem on par with QQ.
I don't believe this will happen either. The entire responsibility is on Intel's hands to PROVE that they can execute, rather than promising for 8 years. Look at how well Silvermont was perceived in terms of performance. Remember how the lead Silvermont had was beaten by Apple in a week?

In 1.5 years I think they need to be at least 60% faster in CPU. That's in Core M territory. Do I think Core chip available in that time will gain that much? No. That means either cannibalization happens or artificial crippling of Broxton.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
http://cdn2.wccftech.com/wp-content/uploads/2015/02/Intel-Cherry-Trail_Atom-x7-3D-Performance.jpg

How in the world do they expect 2x gain in GFXBench 2.7 T-Rex HD(Offscreen) and 50% gains in 3DMark Ice Storm Unlimited to justify anything other than a mid-range product?

3DMark Ice Storm Unlimited Graphics(Offscreen)-
A8X: 31.8k
Tegra K1: 36.7k
Atom(Best score): 17.2k
Cherry Trail: 25.8k

GFXBench 2.7 T-Rex HD(Offscreen)-
A8X: 70.5 fps
Tegra K1: 66 fps
Atom(Best score): 27.3 fps
Cherry Trail: 54.6 fps

"Devices based on Atom x5 and x7 processors are due out in the first half of this year."

So Cherry Trail, loses to a current generation when by the time its released, just 6 months later, successors like A9X and X1 will be released. The latter of which will improve performance by 50-100%.

Here's how GFXBench 3.0 T-Rex HD stands:
A8X: 32.4 fps
Tegra K1: 31.2 fps
Atom(Best score): 16 fps
Cherry Trail?

X1, which by the way, doubles that to 63.6 fps and beats Core M graphics in that benchmark and overall performs similar.
 
Last edited:

kpkp

Senior member
Oct 11, 2012
468
0
76
Sofia on Windows 10 for phones is now official.
Microsoft, Intel join forces on low-cost Windows 10 phones

The 2 titans from the PC world, try to do it again on phones. The only logical advantage that could bring is if you can dock the phones, but focusing exclusively on the low end, I don't see that being a viable idea or if I want to be optimistic, they are saving the big announcements for Build.
 
Last edited:

Fjodor2001

Diamond Member
Feb 6, 2010
3,990
440
126
It's not that they can't. It's that they didn't. They didn't spend the money doing so, because it was perceived as not being worth it. That may have been a costly mistake on their part... hopefully it is being rectified with 10 nm, but I don't have much faith. They did make some moves explicitly for high performance, though -- like the thick metal layers they added -- but so far, 14 nm has not translated to the real world gains you'd think there'd be.

I do not have high hopes for BDW-K at all, despite my previous, outspoken optimism.

Hmmm.... what to make of that?

Intel could have made 14 nm a high performance node, but chose to prioritize low power mobile instead?

Are we sure it's not just that 14 nm provided more benefits in the low power range, regardless of what Intel would have liked? I.e. even if they'd chosen to prioritize high performance over low power, the intrinsic properties of 14 nm did not allow that?

Also, more importantly, is this a trend we'll see going forward to 10, 7, 5 nm? I.e. it'll not primarily improve high performance and frequencies? :hmm:
 

podspi

Golden Member
Jan 11, 2011
1,982
102
106
Sofia on Windows 10 for phones is now official.
Microsoft, Intel join forces on low-cost Windows 10 phones

The 2 titans from the PC world, try to do it again on phones. The only logical advantage that could bring is if you can dock the phones, but focusing exclusively on the low end, I don't see that being a viable idea or if I want to be optimistic, they are saving the big announcements for Build.
A dockable phone would be amazing. We're really almost at that point where it'd be reasonable.
 

Khato

Golden Member
Jul 15, 2001
1,249
321
136
Are we sure it's not just that 14 nm provided more benefits in the low power range, regardless of what Intel would have liked? I.e. even if they'd chosen to prioritize high performance over low power, the intrinsic properties of 14 nm did not allow that?

Also, more importantly, is this a trend we'll see going forward to 10, 7, 5 nm? I.e. it'll not primarily improve high performance and frequencies? :hmm:

If they actually make use of the potential density improvements, yeah, the trend may very well continue. I'm guessing that whatever increase in drive current Intel obtained from the better fin structures in 14nm may well have been completely negated/overridden by the increased parasitic parameters of the denser interconnect. And that they may continue that trend in order to stay on Moore's law rather than relax density to increase maximum performance.
 

jdubs03

Senior member
Oct 1, 2013
951
594
136
I don't think even Skylake will turn out to be the panacea that's hyped out to be. If we take the best-case scenario of 20-30% gain in Y class talked about here in the forum by a Chinese guy, that's still pretty poor. 2.3 Cinebench R11.5 score on a brand-spanking-new top-of-the-line 2016 product. Whoopee-do

In 1.5 years I think they need to be at least 60% faster in CPU. That's in Core M territory. Do I think Core chip available in that time will gain that much? No. That means either cannibalization happens or artificial crippling of Broxton.

If they can get 20-30% on top of Broadwell-Y I think that would be satisfactory (even at that min value), anything less its trouble, anything more we should throw a party. 2.3pts isn't good enough, 2.6pts is (which is 1.3x the 2.0 of 5y71) though I'd be more interested in single-thread, if it is near 1.3-1.4, we should be satisfied.

On the 2nd point, goldmont better make up for the no gain from airmont. airmont is a joke, and goldmont has to up the ante to at least 1600 in gb3sc. and that is by no means certain (even then they're going to be considerably behind on sc vs. next years exynos, kryo, p1, and a10/x).

core m is the saving grace for intel at premium priced tablets, and the margin on those is likely to decrease.

--

my prediction is that next year the new exynos probably won't be a huge performer in relation to this years version. kryo i think has the chance to bring qcomm back near the top, but neither i anticipate will challenge apple in perf. nvidia will probably be a beast too.
 

III-V

Senior member
Oct 12, 2014
678
1
41
Hmmm.... what to make of that?

Intel could have made 14 nm a high performance node, but chose to prioritize low power mobile instead?

Are we sure it's not just that 14 nm provided more benefits in the low power range, regardless of what Intel would have liked? I.e. even if they'd chosen to prioritize high performance over low power, the intrinsic properties of 14 nm did not allow that?

Also, more importantly, is this a trend we'll see going forward to 10, 7, 5 nm? I.e. it'll not primarily improve high performance and frequencies? :hmm:
22 nm was the same way.

If you were to backport the improvements made with 14 nm -- air-gapped interconnects, taller, rectangular fins -- and put it on 32 nm, you'd have something that'd clock much better than 32 nm did back then.

I am not entirely sure it's a flaw with the process, so much as it's an issue with making things smaller -- this article highlights a lot of the issues that were appearing back in 2004, before some significant developments were made that prolonged frequency scaling for a few more nodes (e.g. Hi-K Metal Gate).

Intel also really pushed density -- something they previously did not do. And hey, it worked. They have the densest node on every reported metric. It's also fantastic for low power. But effort spent in those areas could have been spent on performance -- it's a "can't have your cake and eat it too" problem. Density is also inverse to performance these days.

However, I think in this instance, it's more of an issue of outdated ways of thinking, in regards to CPU architecture, than anything else. Certainly, making things as wide as possible is not going to work in the long term.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
Hmmm.... what to make of that?

Intel could have made 14 nm a high performance node, but chose to prioritize low power mobile instead?

Are we sure it's not just that 14 nm provided more benefits in the low power range, regardless of what Intel would have liked? I.e. even if they'd chosen to prioritize high performance over low power, the intrinsic properties of 14 nm did not allow that?

Also, more importantly, is this a trend we'll see going forward to 10, 7, 5 nm? I.e. it'll not primarily improve high performance and frequencies? :hmm:

Yes, that is precisely how these things happen.

The engineers develop a product (in this case a process node) within a set of constraints (timeline, cost, etc) while being directed to maximize a set of priorities (higher density, lower leakage, etc.) and what comes out of that is something that reflects the priorities and direction set by the decision makers within management.

What I think you are mistakenly doing is assuming that there is such a thing as "intrinsic properties of 14 nm". The intrinsic properties certainly do exist, but they exist by intentional creation by the engineers themselves.

As a baker, when I bake pumpkin pie the final product (the pumpkin pie) is most certainly going to have the intrinsic property of pumpkin. Not by virtue of it being a pumpkin pie, but by virtue of me (the baker) intentionally putting pumpkin into the pie when I created it.

10nm and subsequent nodes will continue to target whatever electrical properties Intel's decision makers feel the market is most willing to pay for.

When Apple reports to its shareholders that the market was willing to enrich Apple with $18B of profits in just one business quarter, that is a compelling reason for Intel to pursue similar markets.

No business right now is standing in front of their shareholders telling them they just banked $18B profit in 90 days by selling high-clockspeed desktop CPUs. So why would they make it a priority when there are clearly much more profitable markets to go after?
 

DrMrLordX

Lifer
Apr 27, 2000
22,117
11,783
136
I mean in a real market with competition you cant just make those former soviet 10 years plan and just truck on.

IN SOVIET RUSSIA, 10-year plans makes YOU!

ah ah ah, I love thees country.

You guys should rename this thread "Intel world congress."

Intel may not have a lot to show at MWC, but honestly I'd rather read about Intel's foray into the phone/tablet realm than read about a bunch of ARM stuff. If x86 could somehow crush ARM and push them out of the phone/tablet space, it would be hilarious. More people crying about a bad ISA dominating everything. Mua ha haaaa!
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
Intel may not have a lot to show at MWC, but honestly I'd rather read about Intel's foray into the phone/tablet realm than read about a bunch of ARM stuff. If x86 could somehow crush ARM and push them out of the phone/tablet space, it would be hilarious. More people crying about a bad ISA dominating everything. Mua ha haaaa!

I don't care if x86 or ARM is the winner, what I care about is the veritable explosion in inexpensive and readily accessible apps that has come with the ARM ecosystem.

With ARM, the app store is the killer app.
 

Mk pt

Member
Nov 23, 2013
67
17
81
Intel may not have a lot to show at MWC, but honestly I'd rather read about Intel's foray into the phone/tablet realm than read about a bunch of ARM stuff. If x86 could somehow crush ARM and push them out of the phone/tablet space, it would be hilarious. More people crying about a bad ISA dominating everything. Mua ha haaaa!

Intel shares owner/investor?

Because only for profit one would like to see a monopoly in mobile SoC market like he have right now in the desktop cpu business.
 

Nothingness

Diamond Member
Jul 3, 2013
3,134
2,145
136
Intel may not have a lot to show at MWC, but honestly I'd rather read about Intel's foray into the phone/tablet realm than read about a bunch of ARM stuff. If x86 could somehow crush ARM and push them out of the phone/tablet space, it would be hilarious. More people crying about a bad ISA dominating everything. Mua ha haaaa!
Yes can't wait for that to happen... and see you cry because you've lost the possibility to choose and have to pay what Intel wants and accept no progress for generations of similar chips.

x86 *is* a bad ISA. Only someone with zero knowledge of its assembly language and another assembly language would claim otherwise.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
Intel shares owner/investor?

Because only for profit one would like to see a monopoly in mobile SoC market like he have right now in the desktop cpu business.

Yes can't wait for that to happen... and see you cry because you've lost the possibility to choose and have to pay what Intel wants and accept no progress for generations of similar chips.

x86 *is* a bad ISA. Only someone with zero knowledge of its assembly language and another assembly language would claim otherwise.

C'mon guys, surely you both caught to the sarcasm in his post, right? He even ended it with the standard maniacal laugh (you know, because it would be maniacal to want Intel to have a monopoly, etc.). Pretty sure he gets it if he is making satirical fun of it
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |