Intel Broadwell BDW-H delayed May 2015

[witeken]

Yes, it does. Intel has a Tick-Tock model with a 1 year cadence.

 

[mikk]

Yes, it does. Intel has a Tick-Tock model with a 1 year cadence.

Like Haswell>Broadwell

Dream on.

 

[witeken]

Like Haswell>Broadwell

Dream on.

14nm had a 3 month delay because of yield issues. And Intel has 2 major architectures; Atom is now equally important as Core. And for mobile, for Broadwell-U/Y and Atom, 14nm is much more important than desktop, for which a tock is most important. I think that 14nm was just an exception.

 

[mikk]

14nm had a 3 month delay because of yield issues. And Intel has 2 major architectures; Atom is now equally important as Core. And for mobile, for Broadwell-U/Y and Atom, 14nm is much more important than desktop, for which a tock is most important. I think that 14nm was just an exception.

6 months delay. Broadwell was planned for Q2 2014 initially. 24-30 months is a realistic target for the next shrink, surely not 18 months.

 

[Homeles]

6 months delay. Broadwell was planned for Q2 2014 initially. 24-30 months is a realistic target for the next shrink, surely not 18 months.

Q1, actually.

Whoops, never mind. Bad memory.

 

[Fjodor2001]

Yes, it does. Intel has a Tick-Tock model with a 2 year cadence.

Fixed that for you. On desktop we have ~2 years for Haswell -> Broadwell-K.

 

[CHADBOGA]

It is a very confusing time to be sifting through the Tea Leaves of Intel's roadmaps. :whiste:

 

[witeken]

6 months delay. Broadwell was planned for Q2 2014 initially.

Yes, end of Q2. Unless you would cherry pick; the people who claim those incredibly long times between tick-tocks also say that it takes something like 9 months to go from production to launch, so early Q2 wouldn't be possible at all.


So still about 3 months delay.

24-30 months is a realistic target for the next shrink, surely not 18 months.

Intel stated volume production inQ4'15, so same Q as 14nm was planned which means a Q2 launch and a delta of less than 2 years.

 

[witeken]

Fixed that for you. On desktop we have ~2 years for Haswell -> Broadwell-K.

Intel's tick-tock doesn't say anything about specific SKUs. Haswell -> Broadwell will be 1 years and some months.

 

[Fjodor2001]

Intel's tick-tock doesn't say anything about specific SKUs. Haswell -> Broadwell will be 1 years and some months.

If it should be sustainable in the long run, it has to be valid per SKU category too. So if desktop Haswell -> Broadwell-K is 2 years, desktop Skylake has to be 0 days. Is that likely?

 

[SAAA]

If it should be sustainable in the long run, it has to be valid per SKU category too. So if desktop Haswell -> Broadwell-K is 2 years, desktop Skylake has to be 0 days. Is that likely?

Apparentely it's happening (http://forums.anandtech.com/showthread.php?t=2384656)... Still this leave me some dubts about the improvements in Skylake, if any at all.
Or at least not the much larger than 10% that many were wishing, unless Broadwell overclocks like mad and offsets any larger gap.

 

[Homeles]

Apparentely it's happening (http://forums.anandtech.com/showthread.php?t=2384656)... Still this leave me some dubts about the improvements in Skylake, if any at all.
Or at least not the much larger than 10% that many were wishing, unless Broadwell overclocks like mad and offsets any larger gap.

A separate team handles every other tock. E.g., Hillsboro is responsible for Nehalem, Haswell, and probably whatever's after that, whereas Haifa is responsible for Conroe, Sandy Bridge, and Skylake. In addition, Haswell -> Skylake is still a ~two year gap, so it's not as if Skylake's team has been given less time than they traditionally are.

 

[Fjodor2001]

Apparentely it's happening (http://forums.anandtech.com/showthread.php?t=2384656)... Still this leave me some dubts about the improvements in Skylake, if any at all.
Or at least not the much larger than 10% that many were wishing, unless Broadwell overclocks like mad and offsets any larger gap.

Well if that is true, tick-tock is dead. It's tick and tock at the same time from now on?

 

[Idontcare]

Well if that is true, tick-tock is dead. It's tick and tock at the same time from now on?

Tick-tock died a while ago, you know this because Intel stopped talking about it.

When was the last time Intel officially referenced tick-tock in any of their marketing, analyst or conference call/earnings presentations?

Last time I saw Intel officially refer to its tick-tock roadmap was circa 2012. It is a nice conceptual model though so it lives on in a useful capacity in forum chatter, but as a business strategy it died a couple years ago once it became obvious that a "Haswell Refresh" was going to become necessary as a 14nm tick delay was unavoidable.

 

[witeken]

Well if that is true, tick-tock is dead. It's tick and tock at the same time from now on?

Broadwell tock: late/early Q3/Q4
Skylake: Q2
Canonlake: Skylake + 1 year
Cannonlake sucessor: Canonlake + 1 year.

 

[witeken]

Tick-tock died a while ago, you know this because Intel stopped talking about it.

When was the last time Intel officially referenced tick-tock in any of their marketing, analyst or conference call/earnings presentations?

Tick, tock, tick, tock.

Last time I saw Intel officially refer to its tick-tock roadmap was circa 2012. It is a nice conceptual model though so it lives on in a useful capacity in forum chatter, but as a business strategy it died a couple years ago once it became obvious that a "Haswell Refresh" was going to become necessary as a 14nm tick delay was unavoidable.

How did Intel know in 2012 that it would delay 14nm? If Intel didn't screw up (around the launch of Haswell), there wouldn't be a delay.

 

[shady28]

Broadwell tock: late/early Q3/Q4
Skylake: Q2
Canonlake: Skylake + 1 year
Cannonlake sucessor: Canonlake + 1 year.

And that's not likely.

SB announced in Jan 2011, with availability in March / April 2011. I know because that's when I bought my iMac, and it was one of the very first systems available with SB.

Then we have April 2012 with Ivy Bridge @ 22nm. I should point out that within a month or so, pretty much everything IB is available.

IRC, Haswell was around June 2013 @ 22nm, so here we have a 2-3 month slip. Again pretty much everything available at this point.

Now in May 2014 we have "Haswell refresh".

Intel is stating "before Christmas" for the 14nm process tech Broadwell. In my book that means November(ish). Plus it appears this will not be a "full" release but will focus on mobile version, with other version releasing over the following 3-4 months.

I'm not debating the logic behind that, but my point being - the 12 month tick/tock died in 2012 with IB. Intel has not been able to maintain it post SB. It's not slipping in nice little 6 month increments as we like to think in, but it's steadily slipping.

It's gone from 12 moths with SB, to 15 with IB, to 16 with Haswell, and now to 18-24 with Broadwell.

I tend to think Broadwell will be one of the longest lived chips since the P3, because that 2016 launch of Skylake is likely to look a lot more like "late 2016 into Q1 2017". I bet Intel spends most of 2015 ramping up 14nm Broadwell and Cherry Trail etc etc. That means Broadwell would be current for a solid 2 years.

 

[mikk]

I tend to think Broadwell will be one of the longest lived chips since the P3, because that 2016 launch of Skylake is likely to look a lot more like "late 2016 into Q1 2017". I bet Intel spends most of 2015 ramping up 14nm Broadwell and Cherry Trail etc etc. That means Broadwell would be current for a solid 2 years.

Skylake-S is coming in Q2-Q3 2015. Broadwell is nonexistent for basically the entire desktop segment, hence why saying the longest lived chip since P3 sounds funny.

 

[shady28]

Skylake-S is coming in Q2-Q3 2015. Broadwell is nonexistent for basically the entire desktop segment, hence why saying the longest lived chip since P3 sounds funny.

Well that's going to be a nice mess. Broadwell and Skylake launching at the same time for different models? That would only make sense if there's no appreciable performance difference, however I did find this :

" Broadwell desktop i5/i7 CPUs will be also released in Q2 2015. What is interesting is that all unlocked processors at launch will be Broadwell based, and all locked parts will be built on Skylake-S core. At this time, Intel does not plan to transition Celeron, Pentium and Core i3 processors to Skylake architecture before second half of 2015."

It sounds to me like the desktop cpu is about to fade into obscurity.

 

[Homeles]

I tend to think Broadwell will be one of the longest lived chips since the P3, because that 2016 launch of Skylake is likely to look a lot more like "late 2016 into Q1 2017". I bet Intel spends most of 2015 ramping up 14nm Broadwell and Cherry Trail etc etc. That means Broadwell would be current for a solid 2 years.

You'd probably benefit from glossing over the top 10 thread titles in this forum -- you're a bit out of touch when it comes to Skylake.

Nevermind, looks like you've been brought up to speed.

 

[mikk]

Well that's going to be a nice mess. Broadwell and Skylake launching at the same time for different models? That would only make sense if there's no appreciable performance difference, however I did find this :

" Broadwell desktop i5/i7 CPUs will be also released in Q2 2015. What is interesting is that all unlocked processors at launch will be Broadwell based, and all locked parts will be built on Skylake-S core. At this time, Intel does not plan to transition Celeron, Pentium and Core i3 processors to Skylake architecture before second half of 2015."

It sounds to me like the desktop cpu is about to fade into obscurity.

As a non OC user I like it. By mid next year I can choose between a SKL non-K Quadcore or Broadwell-K with a bigger iGPU and edram if all goes according to plan.

 

[NTMBK]

How did Intel know in 2012 that it would delay 14nm? If Intel didn't screw up (around the launch of Haswell), there wouldn't be a delay.

When did Intel publicly announce the shelving of its Arizona fab? And when did they internally start considering this, and its ramifications on product timelines?

 

[IntelUser2000]

Tick-tock died a while ago, you know this because Intel stopped talking about it.

I don't think that's true.

Intel does this, I believe in part to reduce the time it takes to bring a new process.

They did that back in Nehalem, when it was thought that 32nm Westmere would launch in March-April. That would mean a delay of 3-4 months. But because they had literally no "mainstream" desktops CPUs based on Nehalem, they brought 32nm Clarkdale/Arrandale in January. Remember that the mainstream desktop/laptop chip was Havendale/Auburndale.

-Havendale/Auburndale(45nm): August/September 2009
-So a 32nm Clarkdale/Arrandale would be August/September 2010
-And 32nm Westmere was scheduled for March/April 2010
-Because Havendale/Auburndale had problems, they cancelled it and pulled Clarkdale in

Now, they are saying 14nm is delayed. It would affect 10nm launch even more if they went by +1 year schedule. So they take away Broadwell for the mainstream Desktop, allowing Skylake to launch earlier.

-Haswell was mid-2013
-With the 14nm delay, that would mean Broadwell in Q1 2015, Skylake in Q1 2016, and 10nm even if it sticks to 12 month schedule, it'll be Q1 2017, and maybe Q2 2017 if its a 15 months schedule
-They thus cancelled Broadwell for regular Desktops(non-K), allowing Skylake to come in Q2 2015. That allows 10nm at Q3 2016, even with 15 months

 

[witeken]

As a non OC user I like it. By mid next year I can choose between a SKL non-K Quadcore or Broadwell-K with a bigger iGPU and edram if all goes according to plan.

Why do you think Skylake, with its GT4 and Gen9 architecture, will have inferior graphics?

 

[mikk]

Why do you think Skylake, with its GT4 and Gen9 architecture, will have inferior graphics?

I haven't said this. It is pretty clear the initial Skylake models are all GT2 based. I wouldn't expect SKL-GT4e before 2016. Maybe a year after Broadwell GT3e.