Preliminary Skylake Mobile SKU Lineup

[witeken]

Wouldn't that be Gen8? https://intel.activeevents.com/sf14/connect/sessionDetail.ww?SESSION_ID=1312

 

[mikk]

http://www.cpu-world.com/news_2014/2014101501_More_details_on_Skylake_GT1_GT2_and_GT3_graphics.html

GT1 12 EUs
GT2 24 EUs
GT3 48 EUs
GT4 72 EUs

So there is no EU count change from GT1-GT3, they just added a third slice for GT4 variants. GT1-GT3 performance differences must come from architectural GPU improvements.

 

[III-V]

http://www.cpu-world.com/news_2014/2014101501_More_details_on_Skylake_GT1_GT2_and_GT3_graphics.html

GT1 12 EUs
GT2 24 EUs
GT3 48 EUs
GT4 72 EUs

So there is no EU count change from GT1-GT3, they just added a third slice for GT4 variants. GT1-GT3 performance differences must come from architectural GPU improvements.

The biggest factor in IGP performance is undoubtedly with the move from DDR3 to DDR4. Also, I believe Crystalwell will be making a more widespread appearance.

Although I do wonder what Gen9 will bring. I was hoping for more aggressive changes with Gen8, but it seems to be limited to basically re-balancing things. Which is fine, since the performance improvement form that will be fairly significant, but I think Intel still has a lot of catching up to do.

 

[witeken]

http://www.cpu-world.com/news_2014/2014101501_More_details_on_Skylake_GT1_GT2_and_GT3_graphics.html

GT1 12 EUs
GT2 24 EUs
GT3 48 EUs
GT4 72 EUs

So there is no EU count change from GT1-GT3, they just added a third slice for GT4 variants. GT1-GT3 performance differences must come from architectural GPU improvements.

There was a presentation saying there would be a GT with 96EUs, quite a shame they don't double the EU count again. Which is easily possible with 14nm.

The biggest factor in IGP performance is undoubtedly with the move from DDR3 to DDR4. Also, I believe Crystalwell will be making a more widespread appearance.

Although I do wonder what Gen9 will bring. I was hoping for more aggressive changes with Gen8, but it seems to be limited to basically re-balancing things. Which is fine, since the performance improvement form that will be fairly significant, but I think Intel still has a lot of catching up to do.

Have you watched/read Gen8 information, it seems quite a major update to me, although there aren't fundamental changes to the architecture.

 

[III-V]

Have you watched/read Gen8 information, it seems quite a major update to me, although there aren't fundamental changes to the architecture.

Well, that's basically what I was saying. What I was hoping for was a big architectural update, but that's not what we got. Still, the work that they did do should make Gen8 far more appealing, but I still feel like they'll be about a generation behind.

 

[Dufus]

In particular it's funny to see them talk about AVX 3.2, while the official name has been known for more than a year and it's AVX-512.

Before then IIRC, with Larrabee.

FWIW Intel's emulator only gives AVX512 for the Skylake Xeons, not for standard Skylake emulation.

 

[DrMrLordX]

Wouldn't that be Gen8? https://intel.activeevents.com/sf14/connect/sessionDetail.ww?SESSION_ID=1312

Hmm looks like you're right. So we could see some iGPU compute action as early as Broadwell? Hmm! AMD better hurry their butts up on the software side.

 

[mikk]

According to Vr-zone Intel started sampling of Skylake-S ES1.

http://chinese.vr-zone.com/131337/i...latform-will-on-schedule-as-2015-q2-10192014/

They still expect the Greenlow platform for Q2 2015 whilst notebook Skylake for Q4. I think Q3 is a more realistic target for desktop but we will see.

 

[witeken]

Unlikely, since Intel and Brian Krzanich stated multiple times that Skylak would ramp and launch in the second half of the year.

 

[ShintaiDK]

Arcodring to vr-zone the IGP part runs at twice the clockspeed of previous

 

[mikk]

Unlikely, since Intel and Brian Krzanich stated multiple times that Skylak would ramp and launch in the second half of the year.

It's unclear what he exactly means, it could be that he mainly referred to mobile and not to the entry server Skylake or desktop. The Roadmap indicated Q2, I think there is a delay into Q3 because Broadwell-K switched from Q2 into Q3 as well. In that case Vr-zone's informations are a bit outdated, it would explain their Q2 claim.

Arcodring to vr-zone the IGP part runs at twice the clockspeed of previous

I don't see this part in my google translation.

 

[witeken]

That does not pass my sniff test.

When the two were sending samples of the processor clock is 2.3GHz (Turbo Boost can to 2.9GHz) and 2.2GHz (Turbo Boost can to 2.4GHz), which 2.3GHz processor with GT2 Graphics, while the 2.2GHz version does not match any Intel HD Graphics.

 

[witeken]

It's unclear what he exactly means, it could be that he mainly referred to mobile and not to the entry server Skylake or desktop. The Roadmap indicated Q2, I think there is a delay into Q3 because Broadwell-K switched from Q2 into Q3 as well. In that case Vr-zone's informations are a bit outdated, it would explain their Q2 claim.

I thought the IDF presentation was quite clear.

https://www.youtube.com/watch?v=EMSdor65BBY

 

[mikk]

I thought the IDF presentation was quite clear.

https://www.youtube.com/watch?v=EMSdor65BBY

Production in second half means launch in Q4, I doubt Skylake desktop and entry Server is that late.

 

[ShintaiDK]

That does not pass my sniff test.

Ye, it may simply be that the 2.3Ghz is GT2 and the 2.2Ghz is GTx(4?). And no mention of speed.

 

[mikk]

No there is a GT2 and another without GPU. Clock frequencies are CPU related not GPU, it's quite clear from the translator. There is no GPU clock info.

 

[positivedoppler]

What is Intel up to?
http://wccftech.com/intel-14nm-skylakes-engineering-samples-spotted/
http://www.kitguru.net/components/c...r-with-disappointing-frequencies-to-partners/

I'm going to take a while stab and guess maybe Intel is trying to out HSA AMD and allocated a huge chunk of silicon to the GPU in an attempt to wipe out low to mid tier video cards for Nvidia and AMD.

 

[Enigmoid]

What is Intel up to?
http://wccftech.com/intel-14nm-skylakes-engineering-samples-spotted/
http://www.kitguru.net/components/c...r-with-disappointing-frequencies-to-partners/

I'm going to take a while stab and guess maybe Intel is trying to out HSA AMD and allocated a huge chunk of silicon to the GPU in an attempt to wipe out low to mid tier video cards for Nvidia and AMD.

Its an ES so nothing is sure.

IMO it could be a couple things.

Its simply an ES and the 14nm process isn't primed for high frequencies. Or perhaps the frequencies are low because of AVX3 (similar to HW xeons with different boost frequency depending on the instructions) and the global core clock is simply set to match.

What are the chances that Intel has radically designed the core and widened it substantially, a la Core 2, trading clockspeed for IPC?

Honestly until we see some performance at known clocks its impossible to be sure. If its the typical 10% IPC boost and shipping clocks are similar then I would say this is a disappointment.

 

[Arachnotronic]

What are the chances that Intel has radically designed the core and widened it substantially, a la Core 2, trading clockspeed for IPC?

That'd make sense. Core is long overdue for a widening

 

[jpiniero]

Intel shipping at low clock speeds is not unusual for an early ES but you would think the TDP would reflect that. It would not surprise me if they are dealing with heat density problems, but not that bad.

 

[witeken]

What is Intel up to?
http://wccftech.com/intel-14nm-skylakes-engineering-samples-spotted/
http://www.kitguru.net/components/c...r-with-disappointing-frequencies-to-partners/

I'm going to take a while stab and guess maybe Intel is trying to out HSA AMD and allocated a huge chunk of silicon to the GPU in an attempt to wipe out low to mid tier video cards for Nvidia and AMD.

Why are Kitguru and Wccftech already reviewing CPUs which they don't have, which they don't know anything about, which are not even close to being released and are not final?

The relatively low frequencies of the Skylake-S samples is a disturbing news since in many cases low clock-rate means low performance. It is unclear whether Intel will manage to boost frequencies of the Skylake-S chips by the time they hit the market.

[...]if this is true then Intel is well on its way to becoming a monopolistic tyrant.

That does not pass my sniff test.

 

[erunion]

Intel shipping at low clock speeds is not unusual for an early ES but you would think the TDP would reflect that. It would not surprise me if they are dealing with heat density problems, but not that bad.

I wouldn't read into the TDP figure. Its likely not representing the chip's actually parameters, merely distinguishes the chip as an ES for a 95W-class product.

 

[witeken]

What is Intel up to?
http://wccftech.com/intel-14nm-skylakes-engineering-samples-spotted/
http://www.kitguru.net/components/c...r-with-disappointing-frequencies-to-partners/

I'm going to take a while stab and guess maybe Intel is trying to out HSA AMD and allocated a huge chunk of silicon to the GPU in an attempt to wipe out low to mid tier video cards for Nvidia and AMD.

BTW, couldn't it be possible that those are laptop chips? Intel said at IDF that the developer laptops would be shipping in Q1, IIRC.

 

[Arachnotronic]

BTW, couldn't it be possible that those are laptop chips? Intel said at IDF that the developer laptops would be shipping in Q1, IIRC.

LGA 115x chips usually don't go into laptops

 

[ShintaiDK]

"Low" frequency ES samples is nothing new. Just look at history.