What is Intel's 2016 strategy?

[SOFTengCOMPelec]

So let me get this right, the massively underfunded Zen/K12 will somehow turn up as a direct miracle about equal to the reborn of christ.

In the same time Intel will have huge delays because some Wikipedia article may state so? Not that I even see this delay in the article.

And somehow this will make AMD competitive?

This thread is made up of fantasy.

The problem (I've got), is that the "official" information (or even rumors), lack relatively vital information, in which to reliably respond to your comments.

The key unanswered questions in my mind are:

How much faster is Skylake going to be (including IPC, any clock speed improvements and instruction set speedups, such as AVX3), compared to the existing range of cpus ?

Is AMD Zen going to come out on time, and be as good as it can be, given its budget and other constraints ?
I.e. How fast (IPC) is it going to be ? How many cores (max) ? and what frequency will it go up to ?

Despite Intels claims to the contrary, is the progress (towards Canonlake 10nm) going to be on time and as trouble free, as they are hoping ?

Since I don't know the answers to these questions, until closer to 2016/17, I can't easily agree or disagree with you.

 

[Arachnotronic]

*Your article of 10nm in H1'17 has been debunked by Intel.
*Intel has said 10nm production would start in 2015.
*Intel is at this very moment spending (see CFO commentary) as if HVM would start in Q4 or early next year.
*Also see the 10nm timing in this silde: http://www9.pcmag.com/media/images/409118-intel-technology-roadmap.jpg?thumb=y
*Mark Bohr said he doesn't expect problems for the 10nm node.

So Intel strategy for 2016 will be to
*Further reduce mobile loss
*Gain smartphone and tablet footprint with
- Broxton
- SoFIA MID
- SoFIA 2
*Announce 10nm successors for these product with early '17 launch
*Launch Cannonlake for desktop, mobile and Core M (with good availability in Q4)
*Increase silicon photonics sales
*Continue 'disruptive' 3D NAND strategy
*Sell Knights Landing, announce more about Knights Hill for H2'17 launch
*15% CAGR in DCG
*Something with IoT
*Continue no-cables initiative, NUC, Next-NUC, Compute Stick, fanless strategy
*Maintain flat or slight growth in PC
*Overtake Qualcomm in connectivity
*Announce Core for smartphone ^^ (possibluh for iPhone 7)
*Announce/use EUV?

The items in bold aren't likely.

 

[witeken]

The items in bold aren't likely.

Always better to aim high, though . If Intel doesn't have the internal goal to overtake Qualcomm (at least, technology wise) or develop a product that would make Apple excited to at least take a lot at it, that wouldn't be a sign of ambition.

 

[oobydoobydoo]

Qualcomm has tripped up with S810, so at least Intel has a better chance to catch up than they did in 2014 With bay Trail vs S805. I really hope to see this line of Altera products on intel's 14nm process because that will offer us a lot of insight into their whole process. Never before (IIRC) has intel allowed others to use its process so there may be a lot of new info.

I get the feeling intel is hiding something, and I don't think it's good. The 5Y71 benchmark results in the T300Chi are encouraging. That is, until you notice that battery life goes down to 4-5 hours of movie playback... which means less than 4 hours real world use for most people. That's just not acceptable in 2015 for a $700+ device. It appears they solved the heat dissipation issues, which allowed them to sustain much higher clocks... but that also causes the 5Y71 to have significantly worse battery life than the much poorer performing 5Y10 and 5Y70 competitors.

 

[witeken]

It appears they solved the heat dissipation issues, which allowed them to sustain much higher clocks... but that also causes the 5Y71 to have significantly worse battery life than the much poorer performing 5Y10 and 5Y70 competitors.

If you don't use challenging workloads, the clock speed will be just as high or low as the 2 other SKUs you mention. It's called energy efficiency: joules per instruction. If you have the same amount of instructions and the same efficiency (BDW+14nm), then the amount of joules consumed by the CPU will be the same.

 

[Arachnotronic]

I get the feeling intel is hiding something, and I don't think it's good. The 5Y71 benchmark results in the T300Chi are encouraging. That is, until you notice that battery life goes down to 4-5 hours of movie playback... which means less than 4 hours real world use for most people. That's just not acceptable in 2015 for a $700+ device. It appears they solved the heat dissipation issues, which allowed them to sustain much higher clocks... but that also causes the 5Y71 to have significantly worse battery life than the much poorer performing 5Y10 and 5Y70 competitors.

Poor battery life in video playback sounds like a poor media engine implementation to me. You should NOT be fully loading your CPU playing back video in a modern mobile device.

 

[oobydoobydoo]

Poor battery life in video playback sounds like a poor media engine implementation to me. You should NOT be fully loading your CPU playing back video in a modern mobile device.

My point is that I have yet to see an exhaustive battery life study done on Core M (or any Broadwell product) like there was when Haswell was released. I think Anandtech had a big article about Haswell efficiency and FIVR, but no such thing has ever been done with broadwell. The closest we get to knowing how efficient it is, is to look at wifi browsing times for different devices. Often the only reviews we can compare are those with totally different processors.


Why the dearth of any battery life analysis? This is why I think intel is hiding something... the review sites won't touch Broadwell for any kind of exhaustive analysis. You may be right about other system components causing this, but we just don't know! Why don't we know the performance specifications to a processor that has been available in retail (albeit in very limited quantities) since November 2014?

 

[dahorns]

Poor battery life in video playback sounds like a poor media engine implementation to me. You should NOT be fully loading your CPU playing back video in a modern mobile device.

The testing may show poor implementation or a driver issue. But you also need to remember that the device they tested had a 12.5 in. WQHD screen and only a 30Wh battery. It wouldn't be completely surprising to see lower battery life in that configuration, regardless of the SOC. There is a 1080P panel option that I'm sure would see significantly longer battery life.

Here is what google translate said:

In the endurance part, the official claimed T300 Chi can be used continuously for eight hours, Ada to "balance the effectiveness" of the brightness and volume continuous playback 1080p MP4 format video, up to 10% of the low-battery warning, in fact, about 3 hours and 40 minutes, endurance performance and official claims there is a difference:

Ada also tested to "power saving mode" continuous play 720P movie to test endurance, about 4 hours and 20 minutes of pure Internet and paperwork should be able to level 5 to 6 hours:

The Asus website claims 8 hours 1080p playback.

Another review reported a PCMark 8 battery time of 3:24 to 3:54, which I think is good score for that test (SP3 scores 2:38).

In sum, the reviews on battery time are mixed.

 

[dahorns]

My point is that I have yet to see an exhaustive battery life study done on Core M (or any Broadwell product) like there was when Haswell was released. I think Anandtech had a big article about Haswell efficiency and FIVR, but no such thing has ever been done with broadwell. The closest we get to knowing how efficient it is, is to look at wifi browsing times for different devices. Often the only reviews we can compare are those with totally different processors.


Why the dearth of any battery life analysis? This is why I think intel is hiding something... the review sites won't touch Broadwell for any kind of exhaustive analysis. You may be right about other system components causing this, but we just don't know! Why don't we know the performance specifications to a processor that has been available in retail (albeit in very limited quantities) since November 2014?

Uh . . . other than testing devices, how else do you want them to do battery life tests? We have ample evidence that Broadwell offers superior battery life in standby and load compared to nearly identical systems with Haswell.

 

[Arachnotronic]

Why the dearth of any battery life analysis? This is why I think intel is hiding something... the review sites won't touch Broadwell for any kind of exhaustive analysis. You may be right about other system components causing this, but we just don't know! Why don't we know the performance specifications to a processor that has been available in retail (albeit in very limited quantities) since November 2014?

I own a Dell XPS 13 and battery life seems fine to me (although I haven't bothered to do any rigorous battery life testing).

 

[sm625]

Intel needs to strip out AVX and a few other useless features that make broadwell cores really huge, and then call it the new atom. Then they need to combine ARM and x86 decoders into one megadecoder. Then it will be able to execute ARM and x86 instructions at the same time, which would allow for a true dual boot dual mode device, ideal for both tablets and chromebooks, and hell even some phones. By the time 10nm rolls around, the chips should be low power enough to serve as desktop PC in a 5" phone form factor. They would do this in the hope that microsoft will smarten up and make a compelling windows phone that can seamlessly toggle between WP and windows. This sort of thing is definitely coming so they mas as well start rolling it out now.

 

[oobydoobydoo]

Uh . . . other than testing devices, how else do you want them to do battery life tests? We have ample evidence that Broadwell offers superior battery life in standby and load compared to nearly identical systems with Haswell.

Then why aren't the actual devices lasting any longer in practice? I've seen several comparisons, one with a Haswell system compared to a broadwell, and the haswell system won 1/2 the tests!


If Broadwell was efficient, the devices would be too. I would like them to do a comparison between a haswell and broadwell ultrabook with similar clocks and similar battery sizes, which is easy to do, and then use the bios to clock them the same. Run a test suite... is this new to you? It's pretty standard to do these kinds of exhaustive tests on a new architecture. That is, unless that architecture is Broadwell.

 

[oobydoobydoo]

I own a Dell XPS 13 and battery life seems fine to me (although I haven't bothered to do any rigorous battery life testing).

Exactly, nobody has bothered. Doesn't that seem odd to you? I don't think broadwell is inefficient necessarily, but I think the 5Y71 implementation is benchmarking high at the cost of efficiency... which is important if we want to understand the differences between broadwell and Haswell.

 

[witeken]

My point is that I have yet to see an exhaustive battery life study done on Core M (or any Broadwell product) like there was when Haswell was released.

Intel has studied Broadwell's battery life, but they are probably biased, aren't they.

 

[Dave2150]

Same as their 2015 strategy: more profits. :sneaky:

Well so far their 2015 stategy has been very poor - no product launches for desktop and only delays in sight. Broadwell-K delayed or cancelled, Skylake also delayed.

Intel's 14nm process seems incapable of being viable for CPU's with greater than 2 cores at the moment, since all we've seen are the Broadwell-U parts.

 

[III-V]

Well so far their 2015 stategy has been very poor - no product launches for desktop and only delays in sight. Broadwell-K delayed or cancelled, Skylake also delayed.

They've still been achieving record revenues and profits. Intel's doing fine.

Intel's 14nm process seems incapable of being viable for CPU's with greater than 2 cores at the moment, since all we've seen are the Broadwell-U parts.

That's a silly conclusion to make. Product launches are almost always staggered. It makes sense for Intel to target mobile first, and it's a bit odd to see someone criticize their business strategy, while asking for them to prioritize a stagnant market segment.

Uh . . . other than testing devices, how else do you want them to do battery life tests? We have ample evidence that Broadwell offers superior battery life in standby and load compared to nearly identical systems with Haswell.

Broadwell's strength with idle power is very clear, but I don't feel the same way about load. That's where things get muddy.

 

[dahorns]

Then why aren't the actual devices lasting any longer in practice? I've seen several comparisons, one with a Haswell system compared to a broadwell, and the haswell system won 1/2 the tests!


If Broadwell was efficient, the devices would be too. I would like them to do a comparison between a haswell and broadwell ultrabook with similar clocks and similar battery sizes, which is easy to do, and then use the bios to clock them the same. Run a test suite... is this new to you? It's pretty standard to do these kinds of exhaustive tests on a new architecture. That is, unless that architecture is Broadwell.

Easy enough, go to notebookcheck and look at their review of the new broadwell systems. You can then compare the numbers to last years model featuring haswell. Most of the products are refreshes, so they are literally the same system with the new processor. Performance is up 5-10%, battery life is substantially better at idle and at load, smaller improvements or the same in general wifi-fi testing.

 

[ShintaiDK]

Peak load is quite obvious. Compare the 2 generation NUCs using U SKUs. (Removes most of the laptop platform difference.)

24W to 21W at peak load.

 

[dahorns]

They've still been achieving record revenues and profits. Intel's doing fine.
That's a silly conclusion to make. Product launches are almost always staggered. It makes sense for Intel to target mobile first, and it's a bit odd to see someone criticize their business strategy, while asking for them to prioritize a stagnant market segment.

Broadwell's strength with idle power is very clear, but I don't feel the same way about load. That's where things get muddy.

Everything I've seen has shown heavy load is much more efficient with broadwell. It's the in between states that see equivalent or more modest improvement. No idea why that is.

 

[III-V]

Everything I've seen has shown heavy load is much more efficient with broadwell. It's the in between states that see equivalent or more modest improvement. No idea why that is.

See, I was browsing notebookcheck today and it didn't look so great in the reviews I read. There were a couple of Dells that had similar performance, better idle life, but worse load stats. I've seen stellar performance increases from Broadwell, and embarrassing regressions. It's kind of all over the place, and that seems to be primarily an OEM issue. Wonder if it's got something to do with power delivery.

 

[Pheesh]

My point is that I have yet to see an exhaustive battery life study done on Core M (or any Broadwell product) like there was when Haswell was released. I think Anandtech had a big article about Haswell efficiency and FIVR, but no such thing has ever been done with broadwell. The closest we get to knowing how efficient it is, is to look at wifi browsing times for different devices. Often the only reviews we can compare are those with totally different processors.


Why the dearth of any battery life analysis? This is why I think intel is hiding something... the review sites won't touch Broadwell for any kind of exhaustive analysis. You may be right about other system components causing this, but we just don't know! Why don't we know the performance specifications to a processor that has been available in retail (albeit in very limited quantities) since November 2014?

Anyone can go out and purchase a broadwell system right now and do whatever tests they want. If an exhaustive analysis about one particular facet of a CPU does not exist yet it's simply because no reviewer has gone through the significant effort to do so. Time for you to get the equipment and fill the gap!

 

[Arachnotronic]

Exactly, nobody has bothered. Doesn't that seem odd to you?

Meh, there's a lot of stuff from Notebookcheck that people keep posting here, so professionals have done so.

 

[tweakboy]

Is skylake a atom and whategver ? its special cpu for mobile ? let me know.