Intel Skylake / Kaby Lake

[IntelUser2000]

In Cinebench, the 2.77 for Llama Mountain is probably correct. The XPS13 gets 2.83 with the same silicon but 3MB L3 vs 4MB in the 5Y70. Considering Llama Mountain was running at 6W that kind of score seems reasonable if it was kept cool and could boost to 2.6GHz consistently.

You must be one of those people that think that it only uses 6W but it can "dissipate 15W" because "TDP and power usage is different". At one point a 15W dissipated has to go somewhere, and in a closed space like in a ultra-thin Notebook you can say 15W Power = 15W thermals. Unless you find an application where it arbitrarily runs the CPU full for 10 mins and stops for 10 mins, eventually a 15W CPU that's boosted and using 25W has to come back down to 15W no matter whether it has thermal system to indefinitely sustain 25W. And the correlation between TDP that Intel claimed and power that real CPU used was close... until the Core M chip showed up.

XPS 13 uses a Broadwell U silicon which is a 15W part.

Supposedly Intel claims Broadwell U and Y are different: http://images.anandtech.com/doci/8355/BDW-14nm.png

Unfortunately for Intel they claimed *MUCH* higher performing part than their false hyped claims.

 

[IntelUser2000]

If Skylake Core M performs in both CPU and Games like this: http://www.anandtech.com/show/8983/dell-xps-13-review/3

And you can run for few hours and you still get the same score, then I'll call it successful, and I bet you more manufacturers use it and more customers buy it.

I find it amazing how Intel released Core M in such a poor state. It literally is featured in systems that cost the same as U-class chips.

http://www.anandtech.com/show/9160/the-dell-venue-11-pro-7000-review/4

Dell XPS 13 with Core i5 5200U outperforms the Venue 11 Pro 7000's Core M 5Y71 by 30% in CPU and 50-60% in GPU while it has 19% advantage in Light and identical in Heavy workloads when you normalize for WHr. How can you justify the latter? You can't.

 

[SAAA]

If Skylake Core M performs in both CPU and Games like this: http://www.anandtech.com/show/8983/dell-xps-13-review/3

And you can run for few hours and you still get the same score, then I'll call it successful, and I bet you more manufacturers use it and more customers buy it.

I find it amazing how Intel released Core M in such a poor state. It literally is featured in systems that cost the same as U-class chips.

That's actually the largest problem of core-M: not enough performance lead for the price of the units you pay. Yes they are portable, fanless at times but any ultrabook end up winning it for me, especially at similar prices.

http://www.anandtech.com/show/9160/the-dell-venue-11-pro-7000-review/4

Dell XPS 13 with Core i5 5200U outperforms the Venue 11 Pro 7000's Core M 5Y71 by 30% in CPU and 50-60% in GPU while it has 19% advantage in Light and identical in Heavy workloads when you normalize for WHr. How can you justify the latter? You can't.

That may be due to race to sleep? The XPS can keep the higher clocks longer and end any spiky workload before the Venue that has to throttle at some point.
Oh I wish we had some kind of Llama Mountain tablet selling, then we could compare it with other devices, especially that poor Lenovo part who sunk the chip reputation even more...

I admit it was also Intel's fault in this case, hyping the chip too much and maybe being excessive with the 4.5W TDP.
Haswell was 11 and had many good devices, why don't keep it close: 8-9W and give the SDP (or cTPD down) of 4.5? Who knows, they probably expected 14nm to work better from start but maybe we'll finally get those performance from Skylake.

 

[coercitiv]

That may be due to race to sleep?

It may also be that, contrary to popular belief, passive cooling is not as power efficient as active cooling under certain usage scenarios (continuous load). I don't consider this to be the main cause, but rather something to keep in mind. It also adds up to the race to sleep scenario - CoreM had to work longer and at higher temps. (eg: 1.9Ghz at ~90C versus 2.5Ghz at ~80C)

Venue 11

running Prime95 alone, the processor clocked with 1.7 to 1.9 GHz. A benchmark performed directly afterward resulted in the same scores as in a cold start. The SoC heated up to 90 °C (~194 °F), but settled to around 88 °C (~190 °F) in the stress test.

XPS 13

the XPS 13 aced the CPU stress test. Under 100% CPU load (using Prime95), we monitored clock rates and saw nothing less than 2496 MHz (the max turbo frequency) the entire duration of the test. The maximum temperature recorded by HWiNFO was just 78 °C

 

[MrTeal]

You must be one of those people that think that it only uses 6W but it can "dissipate 15W" because "TDP and power usage is different". At one point a 15W dissipated has to go somewhere, and in a closed space like in a ultra-thin Notebook you can say 15W Power = 15W thermals. Unless you find an application where it arbitrarily runs the CPU full for 10 mins and stops for 10 mins, eventually a 15W CPU that's boosted and using 25W has to come back down to 15W no matter whether it has thermal system to indefinitely sustain 25W. And the correlation between TDP that Intel claimed and power that real CPU used was close... until the Core M chip showed up.

XPS 13 uses a Broadwell U silicon which is a 15W part.

Supposedly Intel claims Broadwell U and Y are different: http://images.anandtech.com/doci/8355/BDW-14nm.png

Unfortunately for Intel they claimed *MUCH* higher performing part than their false hyped claims.

Nice subtle ad hominem. I don't make any assumptions about the power draw of the chip (either 5Y70 or 5200U) from the TDP, and it's not assured that BD-U would draw 15W running Cinebench as well. It is extremely likely that the two chips draw the same amount of power at their boosted frequency if they can both maintain it since they are the same die.

The ability to run Core M at full boost for hours isn't part of the chip. It's a function of the system that it's in, not an inherent limit to the chip. Toss a 5Y70 into an XPS13 (ignoring the difference in package) and it will perform as well as the original 5200U in single threaded benchmarks. The Venue 11 is performance limited by decisions that Dell made wrt thermals.

Your linked image doesn't say what you think it does. That was from when Core M was launching (much before BD-U) and is comparing the scaling Intel got on their 14nm process to traditional scaling. It's not a comparison to BD-U. The GT2 BD-U chips and BD-Y use the same 24 EU, 82mm^2, 1.3B transistor die.

 

[witeken]

René James made this comment on Moore's Law and 10nm:

Brian Krzanich - Chief Executive Officer
The next question from bridge. What is Intel’s long-term plan as we begin to push up against limitations of smaller transistor geometries? That one, I turn to Renée.

Renée James - President
As you heard from Brian, it’s the 50th anniversary of Moore’s Law. And we feel that we’re in a normal 10-year horizon of development and the near in five-year we’re in execution. And this year we will take our 14-nanometer process, will crossover to be our predominant high volume process. The thing that follows 14 is 10-nanometer and the 10-nanometer development is progressing very well. You’ll start to see start-up cost in the second half of the year on 10-nanometer. And we will talk about the timing of that later this year, the early part of next year about when that will happen.

After 10, comes 7-nanometer and we feel very confident that we have line of sight of how to manufacture that note. So overall, we think we’re in very good shape.

Start-up in H2 sounds like early 2017 products, maybe EOY '16.

 

[Ajay]

René James made this comment on Moore's Law and 10nm:



Start-up in H2 sounds like early 2017 products, maybe EOY '16.

"Progressing very well" is a bit of vague response. Based on some of IDontCare's earlier comments, I would think 2016 is out of the question. I think will will have to wait till later for Intel to nail down a better estimate.

 

[dahorns]

Intel Corporation Skylake Client platform
Intel Core m7-6Y75 @ 2.40 GHz
http://browser.primatelabs.com/geekbench3/2637682


A first Geekbench entry of Skylake-Y. Does it mean base frequency will be 2.40 Ghz for m7-6Y75?

Looks like Acer has been playing with Skylake-Y:

http://browser.primatelabs.com/geekbench3/2344229

http://browser.primatelabs.com/geekbench3/2344032

Also, it looks like some of the earlier links were definitely a 4-core 8 thread mobile chip. Acer has an entry listed as an Acer Aspire E5-491G. As far as I know, the Acer Aspire E5 series have all been laptops.

http://browser.primatelabs.com/geekbench3/2641571

 

[Sweepr]

Looks like Acer has been playing with Skylake-Y:

http://browser.primatelabs.com/geekbench3/2344229

http://browser.primatelabs.com/geekbench3/2344032

Also, it looks like some of the earlier links were definitely a 4-core 8 thread mobile chip. Acer has an entry listed as an Acer Aspire E5-491G. As far as I know, the Acer Aspire E5 series have all been laptops.

http://browser.primatelabs.com/geekbench3/2641571

1.6GHz, 2.3GHz and 2.4GHz base clocks. That's quite a jump from Broadwell-Y's 0.8-1.2GHz if those are really Skylake-Y parts.

 

[Shehriazad]

If Skylake is actually comptetitively priced vs the 4K series...I might get a 6500.

But somehow I can't shake the feeling that the pricing will be in no actual relation to performance once it hits the market. But it might flush some more 4K chips ashore that I can toy around with.


Anyway...those leaked benchmarks from a few hours (days? I just saw them) back look interesting, I kind of expected more...but OCing on the K chips is still not known..so yea.

 

[bradley]

I'm actually getting a little excited about Skylake. So basically (in theory) my low profile 1.35V DDR3 should work fine with Skylake's memory controller. Because that's a huge selling point for me. I'll miss my Asus Z-97 Deluxe terribly, but will manage.

 

[Dave2150]

I'm actually getting a little excited about Skylake. So basically (in theory) my low profile 1.35V DDR3 should work fine with Skylake's memory controller. Because that's a huge selling point for me. I'll miss my Asus Z-97 Deluxe terribly, but will manage.

DDR3 DIMMS are not compatible with Skylake, not even the low voltage ones.

The only DDR3 that is compatible with skylake are DDR3L SO-DIMMs, though I doubt many motherboard manufacturers will bother to make specific DDR3L SO-DIMM slots versions.

Also it would be rather silly to invest in a DDR3L SO-DIMM only Z170 motherboard. You'd have to upgrade your motherboard if you ever wanted DDR4 etc.

 

[bradley]

DDR3 DIMMS are not compatible with Skylake, not even the low voltage ones.

The only DDR3 that is compatible with skylake are DDR3L SO-DIMMs, though I doubt many motherboard manufacturers will bother to make specific DDR3L SO-DIMM slots versions.

Also it would be rather silly to invest in a DDR3L SO-DIMM only Z170 motherboard. You'd have to upgrade your motherboard if you ever wanted DDR4 etc.

Basically Intel now makes chips for mobile devices and large scaling servers, the only two who will see power improvements from DDR4. If the desktop isn't dead, someone is sure trying to kill it with all these socket changes. I do believe motherboard manufacturers will address these needs by kludging a solution together, but maybe you're right and the desktop market has shrunken too much to bother.

 

[jpiniero]

As I mentioned before, the Skylake DDR3 support is really for OEMs who don't want to pay the prices for DDR4. Cannonlake supports only DDR4 btw, so they will have to switch eventually.

I think it's fair to say that there are doubts whether Intel will continue to do mainstream line socket models. It's not it's really necessary now since OEMs are obviously fine with doing BGA.

 

[ShintaiDK]

DDR3 DIMMS are not compatible with Skylake, not even the low voltage ones.

The only DDR3 that is compatible with skylake are DDR3L SO-DIMMs, though I doubt many motherboard manufacturers will bother to make specific DDR3L SO-DIMM slots versions.

Also it would be rather silly to invest in a DDR3L SO-DIMM only Z170 motherboard. You'd have to upgrade your motherboard if you ever wanted DDR4 etc.

SO-DIMM or DIMM is completely irrelevant from the CPUs perspective.

I dont think you gonna see any Z170 boards that doesnt use DDR4.

 

[ShintaiDK]

I think it's fair to say that there are doubts whether Intel will continue to do mainstream line socket models. It's not it's really necessary now since OEMs are obviously fine with doing BGA.

Lets not start another round of this fearmongering rumourmill. It was bad enough last time.

 

[crashtech]

Yeah, best to have a link from someone at Intel actually saying it out loud before making such an assertion.

But then again most users here like to act as if the CPU can't/shouldn't come out of the socket, so few would cry if it really was soldered on.

 

[Dufus]

XPS 13

the XPS 13 aced the CPU stress test. Under 100% CPU load (using Prime95), we monitored clock rates and saw nothing less than 2496 MHz (the max turbo frequency) the entire duration of the test. The maximum temperature recorded by HWiNFO was just 78 °C

Strange that they managed to run prime at full tilt yet the Cinebench 11.5 result (which usually draws less power) was only 2.33 instead of above 2.8 indicating an equivalent speed of less than 2.2GHz for the 5200U. Strange also that they doctored the CPU current power draw (whited out) in their cpu stress HWiNFO screen shot.

 

[iSkylaker]

DDR3 DIMMS are not compatible with Skylake, not even the low voltage ones.

The only DDR3 that is compatible with skylake are DDR3L SO-DIMMs, though I doubt many motherboard manufacturers will bother to make specific DDR3L SO-DIMM slots versions.

Also it would be rather silly to invest in a DDR3L SO-DIMM only Z170 motherboard. You'd have to upgrade your motherboard if you ever wanted DDR4 etc.

DDR4 is getting cheap anyway, Crucial already have some budget kits already:
http://www.amazon.com/dp/B00MTSWEQE

Probably by the time Skylake launch, DDR4 kits will be priced at about the same as current high-quality DDR3 kits.

 

[coercitiv]

Strange that they managed to run prime at full tilt yet the Cinebench 11.5 result (which usually draws less power) was only 2.33 instead of above 2.8 indicating an equivalent speed of less than 2.2GHz for the 5200U.

That's a really odd result. The only thing I noticed about Cinebench is it tends to hit the current limit (is it PP0?) even at relatively low package power numbers compared to Prim95. For example, when playing with Haswell U I noticed it didn't go past 15-16W in Cinebench even with PL1 = PL2 = 25W. Crank up max current and frequency/power went up as well.

Nevertheless, Cinebench result is fishy.

Strange also that they doctored the CPU current power draw (whited out) in their cpu stress HWiNFO screen shot.

It is strange indeed, but having the values from the Maximum column kinda defeats the purpose.

PS: thx for the post on TDP and Power Limiting (Haswell) over at notebookreview forums.

 

[IntelUser2000]

Your linked image doesn't say what you think it does. That was from when Core M was launching (much before BD-U) and is comparing the scaling Intel got on their 14nm process to traditional scaling. It's not a comparison to BD-U. The GT2 BD-U chips and BD-Y use the same 24 EU, 82mm^2, 1.3B transistor die.

I don't know how you can conclude it otherwise:

"Broadwell-Y Design/Process Optimizations..."
"A new process flavor for fanless optimization point for Broadwell-Y"

They screwed up, if you are genuinely interested in looking at how Core M devices perform, you'll see they fall very much short of the hype shown at pre-release demonstrations. Now the excuse is that its "thermally dependent". You can only excuse it for so long.

We have seen actual good products. They don't need to hype it, because the product speaks for itself. Core 2 performed better than we expected, it overclocked really well, it consumed less power, it ran cooler, and it beat the competition regardless of whether it was 64-bit mode, synthetic, or widely used applications and code.

The ability to run Core M at full boost for hours isn't part of the chip. It's a function of the system that it's in, not an inherent limit to the chip.

Surely that justifies for the fact that it performs far less than the U chip and throttles like mad even though it costs the same. "Short-burst temporary use" wasn't what Intel hyped on their pre-release launches and performance benchmarks:

http://www.pcper.com/files/imagecache/article_max_width/review/2014-08-11/broadwell-12.jpg
http://cdn0.mos.techradar.futurecdn.net//art/pc_processors/Intel/Intel%20Core%20M%20processor/Intel%20Core%20M%20processor%20LLPT%20Final%20Distribute%20r1-page12-580-90.jpg
http://www.androidcentral.com/sites...announcement_computex_slide.jpg?itok=ZB8c4D6c

Anand and the media did not help by helping to spread the misleading information straight to its readers:

But even that’s still not enough, and for Core M Intel went so far as to give Broadwell-Y its own die and design a low-power optimized version of their 14nm process just for it.

By doing this Intel was able to further reduce power consumption in all of the major areas over what would be a traditional 14nm Intel process.

 

[witeken]

"Progressing very well" is a bit of vague response. Based on some of IDontCare's earlier comments, I would think 2016 is out of the question. I think will will have to wait till later for Intel to nail down a better estimate.

Which comments?

 

[mikk]

In Intels current graphics driver a new GPU naming scheme appeared and I'm pretty sure it belongs to Skylake.

iSKLULTGT1 = "Intel(R) HD Graphics 510"
iSKLULTGT15 = "Intel(R) HD Graphics 510"
iSKLULTGT2 = "Intel(R) HD Graphics 520"
iSKLULTGT3 = "Intel(R) HD Graphics 535"
iSKLULTGT3e = "Intel(R) Iris(TM) Graphics 540"
iSKLULTGT328W = "Intel(R) Iris(TM) Graphics 550"
iSKLULXGT1 = "Intel(R) HD Graphics"
iSKLULXGT2 = "Intel(R) HD Graphics 515"
iSKLDTGT15 = "Intel(R) HD Graphics 530"
iSKLDTGT1 = "Intel(R) HD Graphics 510"
iSKLDTGT2 = "Intel(R) HD Graphics 530"
iSKLDTGT4 = "Intel(R) Iris(TM) Pro Graphics 570/580"
iSKLHaloGT4 = "Intel(R) Iris(TM) Pro Graphics 580"
iSKLHaloGT2 = "Intel(R) HD Graphics 530"
iSKLSRVGT4 = "Intel(R) Iris(TM) Pro Graphics P580"
iSKLWSGT2 = "Intel(R) HD Graphics P530"
iSKLWSGT4 = "Intel(R) Iris(TM) Pro Graphics P580"

http://catalog.update.microsoft.com/v7/site/Search.aspx?q=10.18.15.4212

 

[Fjodor2001]

Looks like Acer has been playing with Skylake-Y:

http://browser.primatelabs.com/geekbench3/2344229

http://browser.primatelabs.com/geekbench3/2344032

Also, it looks like some of the earlier links were definitely a 4-core 8 thread mobile chip. Acer has an entry listed as an Acer Aspire E5-491G. As far as I know, the Acer Aspire E5 series have all been laptops.

http://browser.primatelabs.com/geekbench3/2641571

Are any of those Skylake Y Core M? I couldn't find any indication of that in the test results.

 

[dahorns]

Are any of those Skylake Y Core M? I couldn't find any indication of that in the test results.

I think Skylake-Y, like Broadwell-Y, is Core-M. They are one and the same.

I'm assuming the products I linked are Sky-lake Y because they share the same model as this one: http://browser.primatelabs.com/geekbench3/2637682

The only difference is the stepping (2 vs 3). The scores are also very comparable.