Intel Skylake / Kaby Lake

[IntelUser2000]

That iPad Pro graphics, though...

Embarrassing for Intel.

Looking at preliminary Broxton results, its doing pretty close to the iPad Pro. The situation still seems to be inconclusive.

Benchmarking is at its worst nowadays. We can't normalize for anything.

 

[selni]

The ST Geekbench score of the iPad in the Arstechnica review is also quite good; I'd not say it is embarrassing, but seeing it above Skylake is nonetheless impressive.

Sweepr, there are some better ST results for Skylake Core M, though the one you showed definitely is better at MT. This one for instance: https://browser.primatelabs.com/geekbench3/6235640

Geekbench ST/MT looks to be very heavily distorted by the hash/encryption tests/instruction sets. A9X is still impressive, but looks a lot worse when you see the effect the sha1/2 results have (and to be fair, skylake and AES should go too).

 

[Nothingness]

Geekbench ST/MT looks to be very heavily distorted by the hash/encryption tests/instruction sets. A9X is still impressive, but looks a lot worse when you see the effect the sha1/2 results have (and to be fair, skylake and AES should go too).

Yeah, too much crypto in there.

Removing all crypto this gives:
https://browser.primatelabs.com/geekbench3/compare/5517741?baseline=6235640
iPad Pro ST/Int 2949 (with crypto/hash 3096)
MacBook m7-6Y75 ST/Int 3133 (with crypto/hash 3268)

Still quite close.

But enough off topic

 

[Sweepr]

Desktop Skylake-S GT4e (Iris Pro 580) Listed: Meet Core i7-6785R, Core i5-6685R & Core i5-6585R

BGA (?) Skylake-S GT4e now available to system builders. Just a matter of time till the first LGA models now, hopefully it doesn't take too long.

Full price list.

 

[frozentundra123456]

What are these intended for, and what is the TDP? I would like to see either a laptop chip or a full on high clocked desktop chip. If these are intended for NUC type devices, seems like again Intel is relegating their best product to a niche device.

 

[Glo.]

What are these intended for, and what is the TDP? I would like to see either a laptop chip or a full on high clocked desktop chip. If these are intended for NUC type devices, seems like again Intel is relegating their best product to a niche device.

All-in-One. Like iMac 4K.

 

[Arachnotronic]

All-in-One. Like iMac 4K.

Exactly.

 

[witeken]

Weird. According to the much-respected SemiEngineering, Intel plans 7nm for 2018-2019, which is at most 2 years from 10nm.

http://semiengineering.com/10nm-versus-7nm/

Meanwhile, Intel will move into 10nm production by mid-2017, with 7nm slated for 2018 or 2019, sources said.

Also from the article:

“Not all 10nm technologies are the same,” said Mark Bohr, a senior fellow and director of process architecture and integration at Intel. “It’s now becoming clear that what other companies call a ‘10nm’ technology will not be as dense as Intel’s 10nm technology. We expect that what others call ‘7nm’ will be close to Intel’s 10nm technology for density.”

 

[Sweepr]

JagatOC tried overclocking Intel's HD Graphics 530 iGPU.

JagatOC - Bench Log: Overclocking Intel HD Graphics 530 (Aircooling + Extreme Cooling)

- GPU @ 1350MHz + DDR4-2133 (Air)

- GPU @ 1350MHZ + DDR4-3000 (Air)

- GPU @ 1600MHZ + DDR4-3866 (LN2)

- GPU @ 1850MHz + DDR4-3733 (LN2)

www.jagatoc.com/2016/04/bench-log-overclocking-intel-hd-graphics-530-aircooling-extreme-cooling

 

[Cali3350]

Lenovo now seems to be offering the Yoga 900 with the 6560U and PCI-E SSD. No idea when this happened, but its a nice update? :O

 

[jpiniero]

Weird. According to the much-respected SemiEngineering, Intel plans 7nm for 2018-2019, which is at most 2 years from 10nm.

Intel feeling it a little bit? That feels a little early if they have to use EUV at 7 nm.

 

[frozentundra123456]

Intel feeling it a little bit? That feels a little early if they have to use EUV at 7 nm.

Whistling in the dark, perhaps???? But with the minimal gains we are seeing at each node shrink, not sure it matters too much.

 

[Arachnotronic]

Weird. According to the much-respected SemiEngineering, Intel plans 7nm for 2018-2019, which is at most 2 years from 10nm.

http://semiengineering.com/10nm-versus-7nm/



Also from the article:

7nm in 2018? Bwahahahaha! No.

 

[Shmee]

Any news on Skl E? when is the next E platform out?

 

[witeken]

7nm in 2018? Bwahahahaha! No.

Yeah, Intel is unfortunately not a company that gives light guidance and then overshoots their predictions, instead consistently failing at their goals.

Maybe they should connect their own virtuous IoT gadgets with the data center to improve their yield learning.

If they prove that they can get Moore's Law from 3 to 1 year (10->7 from '17->'18) with IoT, maybe then that will start fueling their IoT business on hot fire, which will accelarate the data center to 20% CAGR !

 

[Arachnotronic]

Yeah, Intel is unfortunately not a company that gives light guidance and then overshoots their predictions, instead consistently failing at their goals.

Maybe they should connect their own virtuous IoT gadgets with the data center to improve their yield learning.

If they prove that they can get Moore's Law from 3 to 1 year (10->7 from '17->'18) with IoT, maybe then that will start fueling their IoT business on hot fire, which will accelarate the data center to 20% CAGR !

Intel's execution these days is just plain bad. I think Krzanich should be shown the door and somebody from TSMC installed at the top.

 

[witeken]

Intel's execution these days is just plain bad. I think Krzanich should be shown the door and somebody from TSMC installed at the top.

BK's big mistake was mobile. It cost them a lot of money and until 2014 it was all "ultra mobilez!!!1", but then they went silence and abandoned the battlefield.

So the contra-revenue was for nothing. The 2 partnerships (Spreadtrum and the other) for nothing. Foundry going slow and not taking off. The complete Atom roadmap disappointed. Remember, back in 2013 BK was super proud of their speed when they announced the "creative" act of improving TTM with Atom on 28nm.

I think no one expected this failure. Intel's track record is superb. I remember using this article in discussions a long time ago on AT: http://seekingalpha.com/article/2025261-history-suggests-intel-will-succeed-in-mobile.

BK's thing was always that he wanted to make the company more agile, super fast and responsive, not missing the big stuff like mobile and IoT. But BK's track record isn't all that good, to be honest. And sure, I understand that Intel wants (and should) "if it computes and connects, it does it best with Intel".

But I just can't understand how they failed so badly. Bad decisions were definitely made by people high on the chain at Intel.

Now it is all "cloud + IoT". No more words about mobile except 5G.

BK shouldn't change his vision so often.

But maybe he was just forced to abandon? Because if you look at it, fundamentally, what caused all of this? The lack of good technology*. Even though the R&D spending rate is astonishing. The super slow time to market / execution. They didn't have compelling technology, so then for BK the game ends and. He lost his bet -- possibly the whole reason he became CEO -- and now also with the PC in free fall (despite Skylake, Windows 10) has to go after the profit.

* Okay, so it all started with the 14nm disaster. Not BK's fault. Maybe some hubris / slow reaction from the manufacturing folks. And this just killed everything beceause it slowed down the WHOLE company. The whole Tick-Tock beating heart of Intel. The 14nm modem. Integrated (low-end) Atom. High-end Atom. Core m still too costly for mobile. [I mean, if you look at this, Intel was just not ready for mobile when BK took over, not prepared at all to fight in the cost sensitive space, unlike the PC.] Broadwell(-E) and Xeon 14nm delay. Skylake wireless vision cancelled. Xeon Phi delay. Silicon photonics delay. Delays all over the place and bad products with the good stuff always 18 months on the horizon.

Last, but least, 10nm delay (!!!!!).

If I were BK, I would stop saying the meaningless words "we'll always strive to get back to 2 year blabla", and instead DO it. I'd slow down the buyback and take the annual 1.4B savings they got now and use all of it to invest in the beating heart of Intel. Because a faster beating heart will result in all the sweet technology (which is better than the competition because they're beating faster), and thus the ROI investors want.

Edit: I don't think they need people from TSMC high at the chain. They should take people from Apple (while they're at it, they can also take Anand Shimpi and Brian Klug).

Edit: You can't put all the blame on Otellini, since BK has done quite a reshuffle of the management, so maybe he made the company just worse in some respects. I dunno.

I just know that they're right with focusing on DCG and HPC and probably IoT and hopefully foundry (but that hasn't materialized since the first talk of opening up in 2013 Investor meeting, and companies probably very scared with the recent track record and lack of experience) and non-volatile memory.

The growth stuff must enable them to continue invest to invest in process thech more and more and 450mm².

 

[frozentundra123456]

Yea, what happened to the wireless charging? I see phones advertising it now. I really dont know what is going on. Even before the "14 nm disaster" they seemed to just throw money at gimmick after gimmick to see what would stick, but very little ever did.

The worst part about 14nm was not only the delay, but the products were not compelling really when they finally did come out. Cherry trail was a disaster. And Broadwell and Skylake gave middling performance gains on the high end, and only modest power savings and performance gains in mobile. I mean you still have throttling issues and I dont see any huge increases in battery life either. The only really nice gains were in igpu, but they still have a long ways to go there as well. It would help if they made edram readily available on more than Macs and super high end devices though.

 

[IntelUser2000]

* Okay, so it all started with the 14nm disaster. Not BK's fault. Maybe some hubris / slow reaction from the manufacturing folks.

We don't know but we can guess!

They focused way too much on density with 14nm. You can see from ALL segments(memory, CPU, flash) that there's a brick wall that's physical limitation that they can't breach.

Memory/Flash reached it first because it reached its limits first. It's just smaller than CPU/Logic/GPU.

They were too proud with what they accomplished with process. Now they are paying for it. In fact I think this is in a minor parallel with decision makers at companies like RIM(Blackberry). They panicked trying to slower their decent, which accelerated their decent.

 

[monstercameron]

Anyword on the performance level of the hd580, super excited about it but ive never owned big core intel nor iris graphics.

 

[Arachnotronic]

EDIT: nevermind! witeken, ygpm.

 

[IntelUser2000]

BK's big mistake was mobile. It cost them a lot of money and until 2014 it was all "ultra mobilez!!!1", but then they went silence and abandoned the battlefield.

I wonder if he did anything wrong other than that he might have been instrumental in bringing 14nm delays.

Obviously CEOs don't know the detail, nor they make decisions on it, but they are supposed to make the big decisions. It may be that "density" focus was BK's idea. Intel has made really good choices in balancing out being conservative versus aggressive in bringing out new tech. Remember Copper Interconnect versus AMD at 0.18u? What about Gate-Last vs. Gate-First? Making the right decisions at that point was what really widened their lead.

Their focus was starting to be killed when mobile "threat" was seemingly unsurmountable. They shouldn't have been so aggressive on density on 14nm. By being too aggressive they sacrificed yields which indirectly impacts performance of the product. Better yields in general allow top bin to be higher, meaning higher clocks. They can also be more aggressive in circuitry designs because they know the process can better handle it.

Now they have disappointed us thoroughly. Airmont in 2014 @ 2.7GHz might not have made a difference, but maybe 3GHz Goldmont in 2015 would have. At least they'd have given everyone the illusion of "hope" for mobility a little longer. Knights Landing in late 2015 would have been the absolute best class. Now its practically 2017, with significantly lower peak FP performance, even lower average performance, and similar perf/watt to Nvidia's Pascal. What about Broadwell/Skylake's performance(or lack thereof)? What about the practical non existence of Broadwell C, or Skylake's availability for purchase?

If they get to the point where Blackberry/Nokia is at, we would point to14nm failures when they really stumbled. People would say "mobile killed them". The reasons would have been complex, but if you can explain it using one word, it would be true. They started falling when they decided to change their core competency, which is process, to get into mobile. Now not only their chances in mobile don't exist anymore, their bread and butter PC business is under threat too.

 

[Sweepr]

Exclusive: Core i3-7100U & Core m7-7Y75 (Kabylake ULT/ULX) GFXBench Results

- Intel® Core™ i3-7100U CPU @ 2.40GHz

- Intel® Core™ m7-7Y75 CPU @ 1.30GHz

Engineer sample. 100MHz higher base clock compared to Skylake.

 

[Head1985]

Will kabylake have L4 on K CPUs?

 

[Arachnotronic]

Will kabylake have L4 on K CPUs?

No.