Skylake/Broadwell Roadmap Update @Vr-zone

[mikk]

Roadmap starts in Q3 2014, so roughly half a year old. Vr-zone Roadmap starts in Q1 2015. I noticed that SKL-H 4+4e for mobile was planned for November 2015 - January 2016....only 1-2 months after SKL-H 4+2.

 

[Fjodor2001]

News: Apparently there will be Skylake Core M chips too. As expected, but I don't think it was confirmed until now.

http://vr-zone.com/articles/intel-core-m-skylake-chips-set-arrive-h2-year/86859.html

Intel CEO Brian Krzanich today confirmed that new Intel Core M chips based on the Skylake processor microarchitecture will be arriving in the second half of the year.
[...]
The Skylake based Core M chips are expected to bring a significant improvement in both application as well as graphics performance. Currently the Core M chips are seen to be weaker performers than the earlier Haswell ULV chips that powered ultrabooks and tablets.
[...]
Other major feature of the upcoming Skylake Core M processors will be support for second generation of Intel’s RealSense 3D camera technology. The technology allows people to interact with their notebook or tablet using gestures and facial recognition.

I don't know what to expect from "significant improvement in both application as well as graphics performance" though. Lately that has meant anything from 5-50%...

 

[mikk]

Broadwell is really short-lived. Given that BDW wasn't really groundbreaking this is understandable.

 

[witeken]

Take performance claims with a grain of salt. They just say "are expected", so it's just speculation.

We already knew SKL-Y would be coming. We've known for a long time now that it's 4W.

I think SKL-Y could get a better reputation, since now people are comparing against HSW or even BDW-U. It will be interesting to see how SKL performance compares to BDW-Y, since it will apparently lack 2nd gen FIVR, and it will have lower 4W TDP. CNL-Y will be even more mature Core M and then it should really become a good option (certainly if it goes to lower price points, as I expect to happen).

 

[frozentundra123456]

Take performance claims with a grain of salt. They just say "are expected", so it's just speculation.

We already knew SKL-Y would be coming. We've known for a long time now that it's 4W.

I think SKL-Y could get a better reputation, since now people are comparing against HSW or even BDW-U. It will be interesting to see how SKL performance compares to BDW-Y, since it will apparently lack 2nd gen FIVR, and it will have lower 4W TDP. CNL-Y will be even more mature Core M and then it should really become a good option (certainly if it goes to lower price points, as I expect to happen).

Sorry, I really want x86 to compete well with ARM in mobile, but wasn't Broadwell supposed to be the magic bullet, and now we are speculating out two more generations, and saying, well maybe then?

Intel needed to take advantage of their process lead right now and strike a serious blow against ARM. Instead we got a delayed, underwhelming product, both in core and atom lines. This just allows ARM to get even more firmly entrenched and improve their performance as well.

 

[Khato]

Broadwell is really short-lived. Given that BDW wasn't really groundbreaking this is understandable.

Ayup, the only 'groundbreaking' point with respect to Broadwell would be the 14nm process... which is also kinda the reason why none of the rest of it is really groundbreaking.

Wouldn't surprise me if the "significant improvement in both application as well as graphics performance" for Skylake actually is significant. Since it will effectively reap the benefits of both tick and tock compared to Haswell/Broadwell. (As reviews of Broadwell thus far have shown that the expected process technology 'tick' improvements really only manifest in decreased die size, not so much the promised efficiency.)

 

[ShintaiDK]

Sorry, I really want x86 to compete well with ARM in mobile, but wasn't Broadwell supposed to be the magic bullet, and now we are speculating out two more generations, and saying, well maybe then?

Intel needed to take advantage of their process lead right now and strike a serious blow against ARM. Instead we got a delayed, underwhelming product, both in core and atom lines. This just allows ARM to get even more firmly entrenched and improve their performance as well.

ARM cant compete performance wise with big cores at all.

And in that segment the battle is Atoms. It never was to be Broadwell.

ARM already lost close to 25% of the tablet segment.

 

[witeken]

Sorry, I really want x86 to compete well with ARM in mobile, but wasn't Broadwell supposed to be the magic bullet, and now we are speculating out two more generations, and saying, well maybe then?

Intel needed to take advantage of their process lead right now and strike a serious blow against ARM. Instead we got a delayed, underwhelming product, both in core and atom lines. This just allows ARM to get even more firmly entrenched and improve their performance as well.

SKL-Y or CLK-Y won't change BDW-Y's performance, competitiveness or adoption. I'm just saying that we're just now seeing the first fanless Core M designs. Those will improve, Core M will improve and the price points should also improve within 2 years. Just look at Ivy Bridge. It was a good CPU for ultrabooks and convertibles etc., but it really took Haswell.

 

[witeken]

Ayup, the only 'groundbreaking' point with respect to Broadwell would be the 14nm process... which is also kinda the reason why none of the rest of it is really groundbreaking.

BDW is actually quite a major overhaul of HSW, with lots of iterative improvements across the board.

http://files.shareholder.com/downlo...6-4AB7-B35F-BD98AB44B43F/Intel_14nm_Aug11.pdf

 

[Arachnotronic]

ARM cant compete performance wise with big cores at all.

And in that segment the battle is Atoms. It never was to be Broadwell.

ARM already lost close to 25% of the tablet segment.

I'm sure an ARM licensee could build a very fast "big core". Just as X86 isn't inherently unsuitable for low power, ARM isn't unsuitable for high performance

 

[kimmel]

I'm sure an ARM licensee could build a very fast "big core". Just as X86 isn't inherently unsuitable for low power, ARM isn't unsuitable for high performance

As long as TSMC/Samsung can get their high speed/power process nodes out. GPUs have been stuck for awhile on 28nm.

 

[frozentundra123456]

ARM cant compete performance wise with big cores at all.

And in that segment the battle is Atoms. It never was to be Broadwell.

ARM already lost close to 25% of the tablet segment.

I was talking about phones primarily, and tablets. And yes that is where intel needed 14nm Cherry Trail to be on time and knock it out of the park on performance per watt. So what have we gotten so far? A MIA product, and based on broadwell results, I am expecting far less improvement than was touted. Where is all the power saving of 14nm? Where is the great graphical improvement from Gen 8?

I won't go crazy like a certain poster here and try to prove that 14nm is a regression. But Intel needed it to be a big step forward in gpu and power savings, and instead it has seemed so far to be just another incremental improvement.

And yes, ARM lost tablet share, but lets see if Intel can maintain that foothold and how rapidly they can decrease contra revenue. Again, I cant see all the delays helping here either.

As I have said before, all the problems with 14nm could not have possibly come at a worse time for intel.

 

[III-V]

BDW is actually quite a major overhaul of HSW, with lots of iterative improvements across the board.

http://files.shareholder.com/downlo...6-4AB7-B35F-BD98AB44B43F/Intel_14nm_Aug11.pdf

Saying "iterative improvements" and "major overhaul" are rather contradictory statements.

That being said, on the CPU side of things, it's a bigger deal than Ivy Bridge was. It seems the graphics side is as well, however bandwidth limitations hold it back a fair bit.

Now, if only the production side lived up to Ivy Bridge's legacy...

 

[Enigmoid]

So what have we gotten so far? A MIA product, and based on broadwell results, I am expecting far less improvement than was touted. Where is all the power saving of 14nm? Where is the great graphical improvement from Gen 8?

I won't go crazy like a certain poster here and try to prove that 14nm is a regression. But Intel needed it to be a big step forward in gpu and power savings, and instead it has seemed so far to be just another incremental improvement.

14nm is pretty good.

http://www.notebookcheck.net/Acer-Aspire-V3-371-Notebook-Review.135831.0.html

We use the stress test to examine the inner temperature development in an extreme scenario. The CPU and GPU are loaded via Prime95 and FurMark, and they are observed for one hour. It is remarkable that the CPU's clock rate never dropped to below 2500 MHz, and thus it remained even over the base speed. The CPU immediately clocked up to its maximum of 2.7 GHz directly after the stress test and achieved the same 3DMark 11 scores as in a cold start.

2.5 ghz CPU + 900 mhz igp at ~18W package power is amazing compared to haswell-U. Broadwell-U can load max turbo SOLID under the most strenuous loads (package power is a little higher than 15W but I'm not sure how accurate HWinfo is here with graphics power at 0.8W).

Its even better when you see the total notebook power consumption is barely over 30W.

 

[Arachnotronic]

I was talking about phones primarily, and tablets. And yes that is where intel needed 14nm Cherry Trail to be on time and knock it out of the park on performance per watt. So what have we gotten so far? A MIA product, and based on broadwell results, I am expecting far less improvement than was touted. Where is all the power saving of 14nm? Where is the great graphical improvement from Gen 8?

I won't go crazy like a certain poster here and try to prove that 14nm is a regression. But Intel needed it to be a big step forward in gpu and power savings, and instead it has seemed so far to be just another incremental improvement.

And yes, ARM lost tablet share, but lets see if Intel can maintain that foothold and how rapidly they can decrease contra revenue. Again, I cant see all the delays helping here either.

As I have said before, all the problems with 14nm could not have possibly come at a worse time for intel.

I think the phenomenon you are describing is really just architectural/product planning failures on Intel's part re: Cherry Trail for tablets.

Intel wildly underestimated what the ARMy would bring to bear in terms of CPU/GPU performance, IMO.

 

[III-V]

14nm is pretty good.

http://www.notebookcheck.net/Acer-Aspire-V3-371-Notebook-Review.135831.0.html

2.5 ghz CPU + 900 mhz igp at ~18W package power is amazing compared to haswell-U. Broadwell-U can load max turbo SOLID under the most strenuous loads (package power is a little higher than 15W but I'm not sure how accurate HWinfo is here with graphics power at 0.8W).

Its even better when you see the total notebook power consumption is barely over 30W.

Wow, Broadwell-U looks pretty amazing in that review.

 

[witeken]

Saying "iterative improvements" and "major overhaul" are rather contradictory statements.

No, it's like evolution. Lots of micro- eventually results in macroevolution.

*Improved density by 2x to 2.2x
*14nm power decrease and performance increase for much improved performance per watt
*Slightly improved BDW (micro)architecture
*Gen 8 architecture with more cache, more shaders and higher sampler throughput
*HSA, DCC, media improvements, H.256
*3DL, 2nd gen FIVR
*Reduced PCB footprint
*Improved PCH

 

[Fjodor2001]

No, it's like evolution. Lots of micro- eventually results in macroevolution.

*Improved density by 2x to 2.2x
*14nm power decrease and performance increase for much improved performance per watt
*Slightly improved BDW (micro)architecture
*Gen 8 architecture with more cache, more shaders and higher sampler throughput
*HSA, DCC, media improvements, H.256
*3DL, 2nd gen FIVR
*Reduced PCB footprint
*Improved PCH

Yes, perhaps lots of small improvements so it looks impressive when listed like that.

But what about the actual end result, which is what matters? At least when it comes to performance, from what has been seen so far, I think the benchmarks have been quite disappointing. Both on the CPU and iGPU.

 

[Fjodor2001]

*14nm power decrease and performance increase for much improved performance per watt

Is that true for all Broadwell models? I remember there being some disagreement on that for the U-models over in the Broadwell thread, but I'm not sure what the conclusion was.

*HSA, DCC, media improvements, H.256

I guess it's a typo, so you mean H.265. But note this:

Broadwell’s video decode capabilities will also be increasing compared to Haswell. On top of Intel’s existing codec support, Broadwell will be implementing a hybrid H.265 decoder, allowing Broadwell to decode the next-generation video codec in hardware, but not with the same degree of power efficiency as H.264 today. In this hybrid setup Intel will be utilizing both portions of their fixed function video decoder and executing decoding steps on their shaders in order to offer complete H.265 decoding. The use of the shaders for part of the decoding process is less power efficient than doing everything in fixed function hardware but it’s better than the even less optimal CPU.

 

[NTMBK]

I think the phenomenon you are describing is really just architectural/product planning failures on Intel's part re: Cherry Trail for tablets.

Intel wildly underestimated what the ARMy would bring to bear in terms of CPU/GPU performance, IMO.

To be fair, if Cherry Trail had arrived on schedule it would probably have been pretty competitive. But the delays mean that their competitors are now are year ahead of where they would have been in the original plan.

 

[krumme]

To be fair, if Cherry Trail had arrived on schedule it would probably have been pretty competitive. But the delays mean that their competitors are now are year ahead of where they would have been in the original plan.

That migt be a factor.
But when i think 2 years back i bought a nook hd+ tablet for one of the kids with a 1.5ghz ti 4470 core 2x arm A9. When released nearly 3 years ago that cpu was the fastest on the arm market.
Now we got 14nm samsumg quad A57.
The improvement in 3 years is insane. Sweepr shows that brilliantly.
We tend to judge Intel compared to that and that makes the improvement seem small.
But the situation is also quite different. Instead of competing against amd intel now face samsung and apple.
The difference here is even far far greater than what the performance differences indicate.
Still i think its relevant to judge eg bw perf on its own merits. And yeaa its not something that makes me personally buy a new ultrabook but for new owners its an small improvement that is kind of free so thats just nice. We have to remember the level was very high before.

 

[coercitiv]

14nm is pretty good.

2.5 ghz CPU + 900 mhz igp at ~18W package power is amazing compared to haswell-U. Broadwell-U can load max turbo SOLID under the most strenuous loads (package power is a little higher than 15W but I'm not sure how accurate HWinfo is here with graphics power at 0.8W).

Its even better when you see the total notebook power consumption is barely over 30W.

The more reviews come in, the stranger this Broadwell vs. Haswell comparison becomes: how is it that despite having lower idle power consumption and seemingly being able to sustain higher turbo clocks at same TDP, Broadwell and Haswell equipped units have the same stamina while browsing the Internet?

If Haswell U lowered idle power usage while keeping the same TDP as Ivy Bridge U and managed to bring significant improvements in both idle and browsing stamina, why does Broadwell U improve in idle power usage, keep the same TDP, yet fail to improve in browsing times?

 

[Enigmoid]

That migt be a factor.
But when i think 2 years back i bought a nook hd+ tablet for one of the kids with a 1.5ghz ti 4470 core 2x arm A9. When released nearly 3 years ago that cpu was the fastest on the arm market.
Now we got 14nm samsumg quad A57.
The improvement in 3 years is insane. Sweepr shows that brilliantly.
We tend to judge Intel compared to that and that makes the improvement seem small.
But the situation is also quite different. Instead of competing against amd intel now face samsung and apple.
The difference here is even far far greater than what the performance differences indicate.
Still i think its relevant to judge eg bw perf on its own merits. And yeaa its not something that makes me personally buy a new ultrabook but for new owners its an small improvement that is kind of free so thats just nice. We have to remember the level was very high before.

Dual A9 to quad A57 is an awesome improvement however it needs to be taken in context. Power consumption has jumped through the roof (its interesting to note that the new A57 Exynos shows basically no efficiency improvements in AT's tests despite a node shrink). We have also gone from 45nm to 14 nm.

 

[witeken]

Maybe because browsing battery life is mainly determined by screen power consumption (size, resolution and screen technology).

 

[CakeMonster]

I'm confused about the video decoding/encoding options (and I've pretty much always been). Does any video player utilize it? Or is it like its only taken advantage of using WMP? In that case I couldn't care less about the hardware support.