[VR-Zone]Skylake-K coming out only 1.5 months after Broadwell-K

[Sweepr]

Who says it does?

 

[witeken]

Who says it's caused by Cherry Trail, was my question...

 

[frozentundra123456]

OP seems a bit harsh. I would view a short transition to skylake as a positive, if the rumors are true. Broadwell is water under the bridge, a "sunk cost" so to speak. In fact it initially was planned to be skipped altogether on the desktop anyway.

It all comes down to the performance of skylake. If it performs well, intel will be in an even stronger position in desktop. If OTOH, the 14nm process itself is fundamentally flawed, then that is a problem, because skylake will we underwhelming as well.

I mean you can rip on Broadwell all you want, but their prime x86 competitor is still on 28nm. The problem for intel is they needed Broadwell to be outstanding in order to make bigger inroads against ARM.

 

[Sweepr]

Who says it's caused by Cherry Trail, was my question...

What else would cause the delay? Back in February/March when Cherry Trail was still expected to be out in late 2014 Broxton was a mid-2015 product.

http://www.pcmag.com/article2/0,2817,2453954,00.asp

 

[erunion]

BDW-K targets a different market, really.

One can't overstate that point.

As I pointed out in the previous thread about this roadmap, everything suggests that the entire broadwell desktop line was cancelled outright. With the exception of quad core iris pro:
http://www.legitreviews.com/wp-content/uploads/2014/03/broadwell-iris-pro.jpg

 

[witeken]

What else would cause the delay? Back in February/March when Cherry Trail was still expected to be out in late 2014 Broxton was a mid-2015 product.

http://www.pcmag.com/article2/0,2817,2453954,00.asp

What, do you think, caused BDW-E's delay?

 

[Sweepr]

What, do you think, caused BDW-E's delay?

Lack of competition? Certainly not the case with Bay Trail/Cherry Trail or Broxton, especially in the crowded Android mobile SoC arena.

 

[witeken]

Lack of competition?

Then why isn't SKL delayed?

So missing that by doing something else with Broadwell would have been a mistake. And I think getting that volume is a good thing. We think we’ve managed the transition on the number of SKUs as Broadwell will have and how we will transition the market to Skylake now moving forward from a margin or COGS standpoint. But remember, they are on the same technology, the same piece of silicon, it’s the same factory. All we do is change the piece of glass in the scanner to get a different product. So there is not a change or revamp of our factories that needs to occur for this.

Stacy Smith - Chief Financial Officer
Yeah. That last point is important. They are both 14-nanometer products for us. So it doesn’t change our factory profile. And just generally, the faster we bring out new features and cool stuff to the market, the better off we are. So we are not planning to slow down Skylake, if that was at the heart of your question.

My hypothesis is because of the yield problems. Defect density issues only get worse with bigger dies.

Certainly not the case with Bay Trail/Cherry Trail or Broxton, especially in the crowded Android mobile SoC arena.

Then why not release Broxton for phones in mid-'15?

 

[Mondozei]

OP seems a bit harsh. I would view a short transition to skylake as a positive, if the rumors are true.

Oh, don't get me wrong, so do I! I don't like artificial delays. Intel, by doing this, is essentially acknowleding that Broadwell has been a flop, due to endless delays(which is about the difficulty of the node transition, essentially).

Broadwell is water under the bridge, a "sunk cost" so to speak. In fact it initially was planned to be skipped altogether on the desktop anyway.

Yeah, I still don't think it makes sense to even release Broadwell-K when Skylake-K comes 1.5 months after that. Of course, I don't have inside info to Intel's business plans, so maybe there is a pricing play at work that they are planning, basically going after AMD with a brick in a dark backalley

AMD, after all, if it competes at all, it competes on price. Getting a Broadwell-K desktop CPU for a price very close to where FX-8350 is would essentially bleed AMD red and if Zen would be inadequate, then that's it...

Which brings me to..

It all comes down to the performance of skylake. If it performs well, intel will be in an even stronger position in desktop. If OTOH, the 14nm process itself is fundamentally flawed, then that is a problem, because skylake will we underwhelming as well.

I mean you can rip on Broadwell all you want, but their prime x86 competitor is still on 28nm. The problem for intel is they needed Broadwell to be outstanding in order to make bigger inroads against ARM.

We can't say anything about 14 nm, but if Intel is indeed going as fast on Skylake-K, it should speak volumes about their confidence at this point.

The second point you raise is interesting. I view this play as primarily anti-AMD. As we know, Zen is coming out in 2016. Intel is skipping artificial delays by pushing ahead as fast as possible and "welcoming" Zen with as strong hand as they can deliver.

When it comes to ARM, Intel is still pursuing a loss-leader strategy or also known as "contra revenue"(basically Intel pay OEMs to use their stuff as to drive adoption). That isn't sustainable and so long as Qualcomm's chips have a better all-integrated solution SoC, Intel will continue to suffer.

Finally, this is about desktop, not mobile. And even if ARM has rattled the saber on the server enterprise solution, it's still overwhelmingly x86 and will continue to be so.

If Intel was really as serious about mobile as they claim, then I agree with previous commentators who think that Broxton shouldn't be artificially delayed to 2016 as it now looks to be, unlike Skylake.

 

[Sweepr]

My hypothesis is because of the yield problems. Defect density issues only get worse with bigger dies.

Are you suggesting the same problem is affecting a much smaller Broxton and that's the reason for its delay?

Then why not release Broxton for phones in mid-'15?

You started questioning that CT's delay wasn't the reason for Broxton delay, now tell us your hyphothesis.

 

[witeken]

Are you suggesting the same problem is affecting a much smaller Broxton and that's the reason for its delay?

No, I was talking about BDW-E.

You started questioning that CT's delay wasn't the reason for Broxton delay, now tell us your hyphothesis.

Because your claim does not make sense to me (and there's no evidence for), so I made some comments to let your reevaluate your statement; my hypothesis is the null one. I don't know (well, to be honest, I just think Broxton's taking longer than anticipated like SoFIA 14nm, or maybe BK's definition of readiness on that roadmap wasn't time to market).

 

[Sweepr]

Because your claim does not make sense to me (and there's no evidence for), so I made some comments to let your reevaluate your statement; my hypothesis is the null one. I don't know (well, to be honest, I just think Broxton's taking longer than anticipated like SoFIA 14nm, or maybe BK's definition of readiness on that roadmap wasn't time to market).

There's no evidence for a non-existent hyphothesis either, so I'll stick with mine till someone at Intel comes clean about the reason for CT and Broxton's delay (and if they are related).

 

[witeken]

There's no evidence for a non-existent hyphothesis either, so I'll stick with mine till someone at Intel comes clean about the reason for CT and Broxton's delay (and if they are related).

I like this one: http://www.explainxkcd.com/wiki/index.php/1478:_P-Values. This is simply how science works. You may think that CT's delay is the reason, but I contend you don't have a lot of certainty you idea is right. Just because we don't know the answer doesn't mean it's some random idea (ever heard of God of the gaps?). There's got to be some reason of course, but I don't know.

 

[frozentundra123456]

Oh, don't get me wrong, so do I! I don't like artificial delays. Intel, by doing this, is essentially acknowleding that Broadwell has been a flop, due to endless delays(which is about the difficulty of the node transition, essentially).



Yeah, I still don't think it makes sense to even release Broadwell-K when Skylake-K comes 1.5 months after that. Of course, I don't have inside info to Intel's business plans, so maybe there is a pricing play at work that they are planning, basically going after AMD with a brick in a dark backalley

AMD, after all, if it competes at all, it competes on price. Getting a Broadwell-K desktop CPU for a price very close to where FX-8350 is would essentially bleed AMD red and if Zen would be inadequate, then that's it...

Which brings me to..



We can't say anything about 14 nm, but if Intel is indeed going as fast on Skylake-K, it should speak volumes about their confidence at this point.

The second point you raise is interesting. I view this play as primarily anti-AMD. As we know, Zen is coming out in 2016. Intel is skipping artificial delays by pushing ahead as fast as possible and "welcoming" Zen with as strong hand as they can deliver.

When it comes to ARM, Intel is still pursuing a loss-leader strategy or also known as "contra revenue"(basically Intel pay OEMs to use their stuff as to drive adoption). That isn't sustainable and so long as Qualcomm's chips have a better all-integrated solution SoC, Intel will continue to suffer.

Finally, this is about desktop, not mobile. And even if ARM has rattled the saber on the server enterprise solution, it's still overwhelmingly x86 and will continue to be so.

If Intel was really as serious about mobile as they claim, then I agree with previous commentators who think that Broxton shouldn't be artificially delayed to 2016 as it now looks to be, unlike Skylake.

Is there any info on what graphics Skylake K will have? It is possible that despite the K designation, Broadwell and Skylake will target somewhat different markets. If the TDP is limited to 65 watt, and and that includes a large igp, Broadwell K may target high end all in ones, desktop macs??, and small brix like devices, with better gpu than the rest of the lineup. Now why that would be unlocked is a good question, but a gt3e would seem to be better suited for those than for high end desktops which will usually have a dgpu anyway.

 

[Sweepr]

I like this one: http://www.explainxkcd.com/wiki/index.php/1478:_P-Values. This is simply how science works. You may think that CT's delay is the reason, but I contend you don't have a lot of certainty you idea is right. Just because we don't know the answer doesn't mean it's some random idea (ever heard of God of the gaps?). There's got to be some reason of course, but I don't know.

Do you think that fact that CT was pushed from 2014 to H1/2015 (probably late H1, mid-2015 for real devices) had absolutely no impact in Broxton's delay? It might not be the only reason (there might be some 14nm difficulties too, or something else) but it could have played an important role. Consumers would be confused if both CT and Broxton devices were released @ mid-2015.

 

[witeken]

Do you think that fact that CT was pushed from 2014 to H1/2015 (probably late H1, mid-2015 for real devices) had absolutely no impact in Broxton's delay? It might not be the only reason (there might be some 14nm difficulties too, or something else) but it could have played an important role. Consumers would be confused if both CT and Broxton devices were released @ mid-2015.

Considering how aggressive Intel wants to be, I'd say that Cherry Trail would simply have been canceled because when it hits the market, it can already be replaced with an improved version without BOM delta, Gen9 graphics, Goldmont architecture and all the rest.

 

[Arachnotronic]

Roadmaps just haven't acknowledged its existence, so far. I'd be a bit surprised if they didn't have "K" SKUs at launch, though. BDW-K targets a different market, really.

It's right on this image from ShintaiDK...

"Enthusiast Quad Core"

 

[rtsurfer]

If BW-K was meant to give 1150 and DDR3 owners an upgrade path, it makes sense to launch it along Skylake.

But if it doesn't support most DDR3 sticks, and older 1150 mobos can't run it, the whole thing just puzzles me.

It does support both. But there is a catch. And its 1.35V DDR3 only. Another catch may be its UniDIMMs only, but thats still to be seen.

Considering that there are people OverClocking DDR4 on Haswell-E & running up to 1.5V through it. I wouldn't be surprised if people can stretch the 1.5v limit to 1.65V unless Intel has MOBO makers disable it with some kind of microcode.

 

[jpiniero]

It's right on this image from ShintaiDK...

"Enthusiast Quad Core"

We don't know when the 4+4e Skylake will be ready though.

 

[Abwx]

Then why isn't SKL delayed?


My hypothesis is because of the yield problems. Defect density issues only get worse with bigger dies.

Then why not release Broxton for phones in mid-'15?

I selected the relevant sentence in the text you quoted....

But remember, they are on the same technology, the same piece of silicon, it’s the same factory. All we do is change the piece of glass in the scanner to get a different product.

 

[Arachnotronic]

We don't know when the 4+4e Skylake will be ready though.

Meh, for enthusiast "K" SKUs, who cares?

 

[Arachnotronic]

I selected the relevant sentence in the text you quoted....

Pretty sure Broxton was not intentionally delayed. Intel had no issues keeping Skylake on schedule, and they needed Broxton ASAP to be competitive in terms of perf/power/BoM.

 

[erunion]

We don't know when the 4+4e Skylake will be ready though.

true, which might be why Broadwell Iris Pro is being released. Because Skylake gt4e isn't coming in 2015.

 

[ElFenix]

working out the kinks of a new process node using an old uarch design so that they wouldn't also be debugging a new uarch is the whole point of the tick-tock strategy. the fact that intel is executing on it doesn't mean broadwell is a failure, it means that tick-tock is working.

 

[NTMBK]

working out the kinks of a new process node using an old uarch design so that they wouldn't also be debugging a new uarch is the whole point of the tick-tock strategy. the fact that intel is executing on it doesn't mean broadwell is a failure, it means that tick-tock is working.

They have only tick-tocked on the CPU. The GPU (which is the majority of the Broadwell chip) is a new arch on a new process node.