Intel Broadwell-K & Skylake (non-K) desktop CPUs to launch in Q2-2015

[Mand]

Man, how confusing is this? Desktop Broadwell and Skylake to be released at the same time?

Why would Intel ever want to do something like that? Are ticks and tocks released simultaneously now?

This.

Can someone explain to me why people think Skylake is coming as soon as it has ever been scheduled on a roadmap chart? I mean, it's seeming like people are expecting them to basically skip Broadwell entirely, and that doesn't make sense to me.

 

[mikk]

This.

Can someone explain to me why people think Skylake is coming as soon as it has ever been scheduled on a roadmap chart? I mean, it's seeming like people are expecting them to basically skip Broadwell entirely, and that doesn't make sense to me.

Skylake was planned for H1 2015 a year ago: http://chinese.vr-zone.com/60095/ha...ddr4-memory-supposed-arrive-2014-q4-04192013/

And yes Broadwell basically skips desktop completely, there is one exception which is the unlocked 4+3e part.

 

[Mand]

And can someone explain how Broadwell skipping desktops makes any sense at all?

How exactly did Intel's development schedule run over itself?

 

[Homeles]

And can someone explain how Broadwell skipping desktops makes any sense at all?

How exactly did Intel's development schedule run over itself?

Want to explain how it doesn't make sense?

Westmere did the same.

Intel's development schedule "ran over itself" because of the 14nm process. Broadwell's clearly been ready to go for quite some time. They had working demos last fall, and it was initially supposed to launch in the first half of this year. The process both Broadwell and Skylake use has been delayed, however that has nothing to do with the architectural development. The process barely came online in time for Broadwell to not be skipped over entirely, since Skylake is basically ready.

 

[Fjodor2001]

Westmere did the same.

http://en.wikipedia.org/wiki/Intel_Tick-Tock

Nehalem, November 17, 2008
Westmere, January 4, 2010
Sandy Bridge, January 9, 2011

 

[SAAA]

Actually Westmere is a wonderful example: quad core Nehalem chip were "upgraded" on the same 45nm process just like Haswell refresh, then they skipped the shrink for mainstream desktop parts and moved directly into 32nm Sandy (both a tick and tock, wow!)

 

[AtenRa]

Actually Westmere is a wonderful example: quad core Nehalem chip were "upgraded" on the same 45nm process just like Haswell refresh, then they skipped the shrink for mainstream desktop parts and moved directly into 32nm Sandy (both a tick and tock, wow!)

Only quads skipped the shrink, 6-cores and duals + HT(First Core i3) were released at 32nm for the desktop.

 

[SAAA]

Only quads skipped the shrink, 6-cores and duals + HT(First Core i3) were released at 32nm for the desktop.

Yes I missed to say quad only, still it's not that different: 6-cores are coming next year at 14nm, but will use Broadwell architecture, also dual-cores starting from Y/U chips. So quads are either mobile Broadwell or Skylake desktop, which makes sense considering they will increase TDP at the same process size with the tock, that's not a problem for desktops. the only outsider at this point are the K chips, will Intel release them on the older tech? Or maybe they are just planning to move the overclocking community to 6-cores with the 5820K and successors.

 

[Fjodor2001]

Still nobody has has answered the question why it makes sense to release desktop Broadwell-K and Skylake at the same time. All that has been mentioned is that "it's been done before" (which really isn't true either as mentioned here and here), but that's no explanation of why it would be reasonable to do so.

 

[mikk]

Broadwell-K is a prestige object for Intels GPU department and not aimed for the mass because of its edram (and high price most likely). Skylake GT2 is aimed for a different clientele.

 

[frozentundra123456]

So if Broadwell-K is not for the "mainstream" overclocker and Skylake is locked, where does that leave the rest who want to overclock, back on Haswell or Devil's Canyon? Still seems like they are giving the shaft to desktop overclockers, in contrast to their stated intention to emphasize performance desktop. Personally though, I would reserve judgement on this, because I am not sure that they will in fact release them simultaneously.

 

[mikk]

Of course Devil's Canyon or Broadwell-K (till Skylake-K arrives, we don't know when this happens).

 

[cbn]

Still nobody has has answered the question why it makes sense to release desktop Broadwell-K and Skylake at the same time. All that has been mentioned is that "it's been done before" (which really isn't true either as mentioned here and here), but that's no explanation of why would be reasonable to do so.

I think it might have to do with Intel optimizing the new uarch for mobile first.

(Example: Look what Intel did with Haswell. Even though the 22nm process was more mature we actually got lower maximum overclocks. Some say this was done because Haswell had 22nm xtors that were more optimized for mobile compared to Ivy Bridge)

 

[Homeles]

Still nobody has has answered the question why it makes sense to release desktop Broadwell-K and Skylake at the same time. All that has been mentioned is that "it's been done before" (which really isn't true either as mentioned here and here), but that's no explanation of why would be reasonable to do so.

As usual, that's not what was said.

And can someone explain how Broadwell skipping desktops makes any sense at all?

Want to explain how it doesn't make sense?

Westmere did the same.

Westmere, like Broadwell, mostly skipped the mainstream desktop.

 

[PPB]

As usual, that's not what was said.
Westmere, like Broadwell, skipped the mainstream desktop.

So robbing a bank makes sense if you did it before?

What Intel did back then didnt make sense, and still doesnt now.

 

[Homeles]

So robbing a bank makes sense if you did it before?

What Intel did back then didnt make sense, and still doesnt now.

http://en.wikipedia.org/wiki/False_analogy

Also, what about it doesn't make sense to you?

 

[frozentundra123456]

So robbing a bank makes sense if you did it before?

What Intel did back then didnt make sense, and still doesnt now.

I guess it does if you got away with it!!!

 

[IntelUser2000]

Hmm at this point my only hope is for 3/4-way SMT or more... 2 core 6/8 threads yeah! The 2 core 2 threads really need to be relegated as celeron or lower end parts in 2015...

2+2 means 2 CPU cores + GT2 GPU cores, not 2 cores/threads.

 

[IntelUser2000]

So robbing a bank makes sense if you did it before?

What Intel did back then didnt make sense, and still doesnt now.

Consider this: When Intel says they want to bring a new chip out, their customers(which are business customers and system integrators since they don't directly sell to consumers) want a fixed amount of time to sell the previous generation products. That's roughly 12 months. Intel sells more than 300 million chips that go into Desktops and Laptops. They can't simply ignore them.

There's a common thing to both 32nm transition and the 14nm transition. They were both delayed. Well, the 32nm was supposed to be delayed, but they pulled back in 32nm Clarkdale chips.

Original schedule:
-Auburndale/Havendale: 2H 2009, 45nm process, dual core + iGPU
-Lynnfield/Clarksfield: 2H 2009, 45nm process, quad core
-Westmere EP: March 2010, 32nm process, 6 cores

Then this happened. Auburndale/Havendale got delayed. It had serious issues. So they pulled Clarkdale/Arrandale in, which is a 32nm dual core + iGPU chip, in January 2010.

Now forward a year to Sandy Bridge. Which chips were refreshed? Dual cores and quad core, replacement to Clarkdale/Arrandale and Lynnfield/Clarksfield.

Now, let's say if Auburndale/Havendale worked fine. And Lynnfield/Clarksfield got 32nm successors. That would have meant
-Clarkdale/Arrandale: Q3/Q4 2010
-32nm quad core successor to Lynnfield/Clarksfield: Q3/Q4 2010

Remember, that OEMs and system integrators want ~12 months for the system to be sold.

That would make Sandy Bridge Q3/Q4 2011. At the earliest Q2/Q3 2011, but that's unlikely.

In 14nm, another situation is a catalyst for cancelling a line to minimize the process delay(Skylake being later means 10nm comes later). PC sales are shrinking. Especially traditional beige box desktops. So they get Haswell Refresh, and get Skylake first.

 

[frozentundra123456]

Of course Devil's Canyon or Broadwell-K (till Skylake-K arrives, we don't know when this happens).

Yea, but why not just skip Broadwell-K and bring out Skylake and Skylake-K at the same time? Still doesnt make sense to me if Skylake is any improvement over broadwell, why bring out the highest performing chips (-K models) on the inferior architecture.

 

[John Connor]

Question: I have heard of Q1 and Q2, etc but what does H1 mean? I know Q is quarter.

 

[AtenRa]

Question: I have heard of Q1 and Q2, etc but what does H1 mean? I know Q is quarter.

First Half(H1) and second Half(H2) of the year. January to June(H1) or Q1 + Q2 and July to December(H2) or Q3 to Q4.

 

[John Connor]

Cool thanks!

 

[Idontcare]

Yea, but why not just skip Broadwell-K and bring out Skylake and Skylake-K at the same time? Still doesnt make sense to me if Skylake is any improvement over broadwell, why bring out the highest performing chips (-K models) on the inferior architecture.

Which is probably why it is a good idea to give consideration towards discounting this "leak" just like we should have done with the whole "ZOMG! Intel is EOL'ing LGA socketed processors for the desktop! For realzies!" silly nonsense of a leak about broadwell when it made the rounds.

This one doesn't pass the sanity test, so apply some Occam's razor and let's move on

 

[Sweepr]

Yea, but why not just skip Broadwell-K and bring out Skylake and Skylake-K at the same time? Still doesnt make sense to me if Skylake is any improvement over broadwell, why bring out the highest performing chips (-K models) on the inferior architecture.

Broadwell 4C+GT3e SKUs might not be ready till early Q2/2015. Perhaps Intel will launch Skylake 4C+GT2 for desktops in late Q2/2015 (or early H2) and all other versions will be launched a few months later. If Skylake-K is a 4C+GT4e part it could be launched much later than Skylake GT2, and they certainly need a new unlocked mainstream chip with competitive iGPU performance to replace Devil's Canyon by the time Excavator and AMD's 2016 APUs arrive.