• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Mobile World Congress 2015

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Another paper launch like Core M last year. It would have been nice if we could get at least a prototype preview from Cherry Trail. Also Intel is silent about CPU improvements in their slides (Brian as well), this speaks for itself. x7-8700 might be 5-10% faster if we are lucky (Geekbench results weren't really faster than Z3770 or Z3795).
 
an altogether incremental information keynote. somewhat disappointing but I guess at this point it makes sense to have tempered expectations.

I just really dont get why intel can't put together a competent roadmap and get it out on time in the mobile segment.

maybe once they get their own integrated soc's on their 14nm/10nm fabs they can start plugging away at the market.
 
Arachnotronic said:
They can put together a competent roadmap. They just have trouble executing to it.
Click to expand...

Its more deliberate seen from here. They have also rightfully reduced capex. If we look at the crazy accelerating competition that is happening on mobile now i think it is right to ajust plans to protect profitability.
I mean in a real market with competition you cant just make those former soviet 10 years plan and just truck on. Its far more dynamic and about continually ajust plans.
The critique of Intel here is imo wrong. What Intel just do is showing they are agile.
 
x7-8700 clocks with 2.4 Ghz says Intel, is this burst frequency or base clock?

Maybe 1.6 Ghz base and 2.4 Ghz burst.
 
It means there is basically zero IPC improvement in Airmont based on the Geekbench scores. Only a slighty higher burst clock under multithreading usage.
 
Intel is really disappointing with atom/airmont. It was already such a small core before the die shrink. They could have added ooe and kept the size the same or maybe even just slightly smaller than silvermont. And the samsung process is very interesting, but too few details as yet.
 
sm625 said:
Intel is really disappointing with atom/airmont. It was already such a small core before the die shrink. They could have added ooe and kept the size the same or maybe even just slightly smaller than silvermont. And the samsung process is very interesting, but too few details as yet.
Click to expand...

Silvermont supports OOE (at least for integer...FPU is in-order).

I'm kind of shocked that ARM has done: A12/A17, A57, A72, A53 in the same time it has taken Intel to bring out Airmont.

What is Intel spending all of that mobile R&D money on?
 
Nothingness said:
And probably more time at higher clock speed on a single core.
Click to expand...

I'm starting to wonder if the reason Intel didn't do an Airmont CPU architecture disclosure is that they didn't actually change anything in going from Silvermont to Airmont. Even most "ticks" have *some* micro-architectural enhancements!

Thank goodness for 14nm bringing perf/watt gains to the table!
 
Arachnotronic said:
I'm starting to wonder if the reason Intel didn't do an Airmont CPU architecture disclosure is that they didn't actually change anything in going from Silvermont to Airmont. Even most "ticks" have *some* micro-architectural enhancements!
Click to expand...
That's all the more odd that Silvermont was a new micro-arch and hence probably has a lot of low hanging fruits. Apple improved IPC by almost 10% on A8 while changing process at the same time. Intel certainly could have done the same, but chose not to.
 
Nothingness said:
That's all the more odd that Silvermont was a new micro-arch and hence probably has a lot of low hanging fruits. Apple improved IPC by almost 10% on A8 while changing process at the same time. Intel certainly could have done the same, but chose not to.
Click to expand...

Probably because they thought Goldmont would be in the market in mid-2015 and Airmont was supposed to be a mid-2014 deal, IMO.
 
1.5 years after Bay Trail and no CPU improvement with its successor is really poor. So we have to wait another year or 1.5 years for a CPU improvement with Broxton. This is insane.
 
Arachnotronic said:
I'm kind of shocked that ARM has done: A12/A17, A57, A72, A53 in the same time it has taken Intel to bring out Airmont.

What is Intel spending all of that mobile R&D money on?
Click to expand...

FYI, SVM and A12 were announced at the same time, with the latter still not found in SoCs. A few keywords:

ARM: IP reuse, low efficiency, big.little, architecture development, licensing.

Intel: HDR, high efficiency (from little to big in 1 core), modem, SoC development* and deltaBOM elimination (Broxton) and process porting (SoFIA), Tick-Tock development (Airmont, Goldmont, Gen8, Gen9, Gen10), 10nm development, software development (Android for x86 etc.).

*They're still entering the market, acquiring partners and pursuing them to use their SoCs. It isn't like they've been here for a decade. Tick-Tock is a good strategy, but they have to get more experience and get better and faster in reacting in this space. Once they get it going they should not run into more such issues with big delays etc. Just look at Intel's entering of the foundry market. It takes some time (like getting the SoC flavor on par TTM wise, for mobile), but in the end it will be top-notch because Intel is committed to it.

I'm not worried. Intel's good at spending R&D (maybe except for graphics?), it may just take a few years for everything to work out.
 
witeken said:
FYI, SVM and A12 were announced at the same time, with the latter still not found in SoCs. A few keywords:
Click to expand...

A17 (revision of A12) is in shipping SoCs today. A57 is, too, and it looks like A72 is imminent.

Silvermont showed up in BYT in late 2013 and now we're getting a clock-slugged Airmont which offers minimal perf/clock enhancement built on Intel's 14nm process in 1H 2015.

It isn't like they've been here for a decade.
Click to expand...

Almost...

iphone10.jpg


IDF 2007.
 
Last edited:
Arachnotronic said:
This. Absolutely poor execution on Intel's part.
Click to expand...

That's how Tick-Tock works, and it isn't like QQ, Nvidia, Apple or ARM are any faster or better at arch development. SVM was this weird architecture that was developed with the non-committed P. Otellini. I'm sure Intel is right now working on 10nm Atom Tock, under control by BK, which should blow the competition away if executed well (even if it was for the sole fact that it's 10nm), which I think will be the case since 10nm Tick-Tock won't be hindered by BOM deltas, contra-revenue, 10nm delays, development of common chassis, development of a new unified tablet-smartphone platform, establishing mobile footprints, persuading OEMs of Intel's commitment, etc.

It was a good thing that P.O. retired.
 
Arachnotronic said:
This. Absolutely poor execution on Intel's part.
Click to expand...

It seems like Intel has invested a lot more in pushing its big cores down the stack than in improving its small cores. I'm starting to think we are more likely to see a Core-M derivative in a flagship smart phone before anything Atom based.

Atom may be forever the cheapo-brand.
 
Arachnotronic said:
Silvermont supports OOE (at least for integer...FPU is in-order).

I'm kind of shocked that ARM has done: A12/A17, A57, A72, A53 in the same time it has taken Intel to bring out Airmont.

What is Intel spending all of that mobile R&D money on?
Click to expand...
It's a bit easier to make small iterations of your architecture, and then go and license it, without ever actually having to worry about the timing of fabs and holidays and such.

All of those are fairly minor revisions to preexisting designs. They're still in low-hanging fruit paradise, so those minor revisions can bring them good boosts in performance and performance/watt.

That's not to say that Intel isn't taking their sweet time. They need to figure out how to lower their TTM without impacting their legendary QA. ARM seems so nimble in comparison, even (or especially?) Apple. Intel cannot afford to stray from reiterating once a year.

They're probably pretty strained, though. They're moving a lot of products to 14 nm -- Broadwell, Skylake, Cherry Trail, Broxton, SoFIA LTE-2, XMM 7460 -- and that's just what they've stated thus far. There hasn't been any point in the past where they've had so many products on their leading edge node. While individual product lines may be moving slow, as a whole, they've become much more aggressive.

All of this will have very positive implications on their cost structure. Intel had previously been missing the rising handheld tide that was lifting all of their competitors' boats. They're doing just about everything the need to be doing... but Cherry Trail and Broxton are definitely behind schedule. Merrifield and Moorefield were super late too.

Also, yeah, Cherry Trail is just unacceptably slow for the year and a half wait. Maybe they were planning on clock speed boosts from 14 nm to be larger in magnitude...
dahorns said:
It seems like Intel has invested a lot more in pushing its big cores down the stack than in improving its small cores. I'm starting to think we are more likely to see a Core-M derivative in a flagship smart phone before anything Atom based.

Atom may be forever the cheapo-brand.
Click to expand...
That's fine, as long as they make Atom cheap. They've burned over a billion as a result of not doing so -- you'd think accelerating Cherry Trail's introduction would be of high importance.
 
witeken said:
That's how Tick-Tock works, and it isn't like QQ, Nvidia, Apple or ARM are any faster or better at arch development. SVM was this weird architecture that was developed with the non-committed P. Otellini. I'm sure Intel is right now working on 10nm Atom Tock, under control by BK, which should blow the competition away if executed well (even if it was for the sole fact that it's 10nm), which I think will be the case since 10nm Tick-Tock won't be hindered by BOM deltas, contra-revenue, 10nm delays, development of common chassis, development of a new unified tablet-smartphone platform, establishing mobile footprints, persuading OEMs of Intel's commitment, etc.

It was a good thing that P.O. retired.
Click to expand...

We'll see. A lot of Intel's Silvermont architects have left the company and are now at Apple, Qualcomm, ARM, etc.

Heck, even Per Hammarlund just left Intel -- and he was a big shot on the company's Core stuff!
 
Last edited:
III-V said:
Also, yeah, Cherry Trail is just unacceptably slow for the year and a half wait. Maybe they were planning on clock speed boosts from 14 nm to be larger in magnitude...
Click to expand...

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?
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top