Intel Broadwell BDW-H delayed May 2015

[TreVader]

It looks like samsung will be on 16nm before intel is on 14nm. Maybe AMD can pull an upset and get GloFo to use their process.

 

[CHADBOGA]

It looks like samsung will be on 16nm before intel is on 14nm.

What are you basing that on?

 

[Homeles]

It looks like samsung will be on 16nm before intel is on 14nm. Maybe AMD can pull an upset and get GloFo to use their process.

You clearly don't understand what's being delayed here.

 

[bullzz]

@Fjodor2001 - U 2+3 is also a 15W CPU. its used in MBA. apple isnt going to like this

 

[Fjodor2001]

Perhaps Apple will be releasing the next MBA as a fanless 12" version using the Y model (see this)? That's the first Broadwell model to be released...

 

[jpiniero]

Perhaps Apple will be releasing the next MBA as a fanless 12" version using the Y model (see this)? That's the first Broadwell model to be released...

Yeah, Broadwell-Y. Obviously Broadwell-U is still too much to be fanless.

 

[Fjodor2001]

Yes, that's interesting! And what's even more interesting is that if you look at the release schedule, all the low power & low frequency models are released first. Then the other models are released at later times, in order of increasing power & frequency.

And the time difference between the first released product (Y-model) and the last model (H/K-models) is ~9 months (!).

You could argue that this is because mobile gets priority nowadays, if production capacity is limited. But this is an unusually large time difference between release of models, something we're not used to seeing. And all the top end mobile/laptop chips will be released last, which is also kind of strange, if mobile/laptop should be prioritized.

So I'm instead suspecting that this could mean that Intel are having problems reaching high frequencies at 14 nm, and are hoping to perfect the process to allow for higher frequencies as time passes. :hmm:

I'll add one more observation: The theory above could also fit in with Intel previously having been said to considering skipping Broadwell-K (and LGA versions of Broadwell in general) completely. I.e. if they at that time were worried that they would not be able to reach sufficiently high clock speeds in time for top end Broadwell desktop LGA versions. So then waiting until Skylake would give them more time to perfect the 14 nm process tech for high speed desktop models...?

Anyone with deeper process tech knowledge than me that would like to comment on that? How much is it usually possible to increase top frequency, from the early stages when a process tech is introduced to the later stages when the process has matured? 10%, 30%, ...?

 

[jpiniero]

If it was simply clock speed issues, they could just simply only release the lower clocked versions.

I still think it's an economics issue more than anything else - that the anticipated cost savings of going to 14 nm is taking much longer to achieve than hoped. Obviously they need to get Broadwell-U and Y out this year so if it's close enough they will just suck it up; but everything else will just have to wait.

Would be very surprised if they end up skipping Broadwell-H and go straight to Skylake.

 

[Ajay]

So I'm instead suspecting that this could mean that Intel are having problems reaching high frequencies at 14 nm, and are hoping to perfect the process to allow for higher frequencies as time passes. :hmm:

Could just be a market timing issue, since Intel is releasing Haswell Refresh CPUs. Of course, then the only reason to release BW H/K models would be to fulfill OEM pre-orders. Skylake LGA is coming out to soon after to bother with Intel Retail CPUs, IMHO.

 

[AtenRa]

That's economic nonsense. As soon as 14nm yields are high enough, Intel's priority will be to switch to the new process to take advantage of the increased margins per unit. $X (22nm) is already sunk and $Y (14nm) mostly spent, so the only way to make more short-term profit on 22nm is to halt development on 14nm. That's not going to happen.

I bet Intel will produce and sell more 22nm SKUs than 14nm in 2015.

 

[witeken]

I'll add one more observation: The theory above could also fit in with Intel previously having been said to considering skipping Broadwell-K (and LGA versions of Broadwell in general) completely. I.e. if they at that time were worried that they would not be able to reach sufficiently high clock speeds in time for top end Broadwell desktop LGA versions. So then waiting until Skylake would give them more time to perfect the 14 nm process tech for high speed desktop models...?

Anyone with deeper process tech knowledge than me that would like to comment on that? How much is it usually possible to increase top frequency, from the early stages when a process tech is introduced to the later stages when the process has matured? 10%, 30%, ...?

To me your hypothesis doesn't make sense. As far as I can see, Core is a suitable architecture for higher clock speeds, and 14nm isn't worse in an way than 22nm, certainly not in a meaningful way that it makes Intel consider skipping Broadwell-K..

 

[SAAA]

But what's important is the original source, which is Chinese VR-Zone, and they are usually correct.

Anyway, here's the release plan:

Just noticed: isn't that 2+3e chip new? In the previous leak only Skylake had 2+3e SKUs and they are supposed to be -U chips, not -H BGA ones...

 

[Arachnotronic]

It looks like samsung will be on 16nm before intel is on 14nm. Maybe AMD can pull an upset and get GloFo to use their process.

Typical pipe-dream, anti-Intel post from you. Seriously...just...okay.

Samsung is struggling with massive yield problems on 20nm but yes, moving to FinFETs is gonna be a cakewalk and they're going to beat Intel to it, the company that has shipped tons upon tons of FinFET silicon.

http://blogs.barrons.com/techtrader...-white-hot-iphone-6-opportunity-says-bluefin/

 

[TreVader]

What are you basing that on?

You clearly didn't read my post. Try again.

 

[TreVader]

Typical pipe-dream, anti-Intel post from you. Seriously...just...okay.

Samsung is struggling with massive yield problems on 20nm but yes, moving to FinFETs is gonna be a cakewalk and they're going to beat Intel to it, the company that has shipped literally megatons of FinFET silicon.

http://blogs.barrons.com/techtrader...-white-hot-iphone-6-opportunity-says-bluefin/

Intels process advantage looks to be mostly FUD at this point. On the only product where they have tough competition (bay trail) they can't match even 28nm ARM.


I hope Samsung allows gloflo to use 16nm. Be funny to see people try to rationalize a supposed two year lead with that info lol.

 

[witeken]

You clearly didn't read my post. Try again.

I read your post again but I still have the same question as CHADBOGA.

 

[TreVader]

I read your post again but I still have the same question as CHADBOGA.

Ssssshhhhh read it a third time.

 

[witeken]

Intels process advantage looks to be mostly FUD at this point. On the only product where they have tough competition (bay trail) they can't match even 28nm ARM.


I hope Samsung allows gloflo to use 16nm. Be funny to see people try to rationalize a supposed two year lead with that info lol.

Intel's been shipping 22nm FinFETs since 2012 in massive volumes, worth multi-dozens of billion dollars. Doesn't sound FUD to me. Here's some proof:

 

[witeken]

Ssssshhhhh read it a third time.

If multiple people have to read it multiple times, I think that means there's something wrong with the post. Maybe you could try again and clarify your post, because 14nm is in volume production right now, while even Samsung's 20nm FinFETs are still a few years away.

 

[TreVader]

Intel's been shipping 22nm FinFETs since 2012 in massive volumes, worth multi-dozens of billion dollars. Doesn't sound FUD to me. Here's some proof:

They are on a different process, but how is that practically an advantage? Look at 22nm vs 28nm. Bay Trail vs S805 for example, they have zero practical performance advantage to the process.


Every difference in performance between 22nm and 28nm could be chalked up to a different Uarch. How is that an advantage? Or is their Uarch so woefully inadequate that it can only compete with a 2yr manufacturing advantage.

They have a different process. They have no advantage over 28nm tho, some over 32.

 

[TreVader]

If multiple people have to read it multiple times, I think that means there's something wrong with the post. Maybe you could try again and clarify your post, because 14nm is in volume production right now, while even Samsung's 20nm FinFETs are still a few years away.

Apparently I replied to the wrong post with that post. I was supposed to reply to homeless and sent a nonsensical reply to Chad

 

[Arachnotronic]

Intels process advantage looks to be mostly FUD at this point. On the only product where they have tough competition (bay trail) they can't match even 28nm ARM.


I hope Samsung allows gloflo to use 16nm. Be funny to see people try to rationalize a supposed two year lead with that info lol.

You are confusing design issues with process issues. Intel at 22nm implements the FinFET device which at the transistor level offers pretty substantially more performance than any other node available today.

Just because Intel's design teams mucked it up doesn't mean the process lead is any less real, but such subtlety apparently escapes you.

 

[witeken]

Silvermont is a very efficient architecture, certainly combined with Intel's 22nm process. You won't see embarrassing things like the clock speed decreasing to a fifth within 3 seconds:

 

[TreVader]

Silvermont is a very efficient architecture, certainly combined with Intel's 22nm process. You won't see embarrassing things like the clock speed decreasing to a fifth within 3 seconds:

Silvermont and bay trail D are both architectures that are competitive with 28nm ARM processors. In some cases, they preform substantially worse.


I'm saying that either 22nm doesn't really offer much over 28nm, or intels uarch is just very inefficient and poor. One or the other is the case.

You can't claim the best Uarch and a process advantage over your competitors and then simultaneously be outperformed by them without something being amiss.

 

[Khato]

They are on a different process, but how is that practically an advantage? Look at 22nm vs 28nm. Bay Trail vs S805 for example, they have zero practical performance advantage to the process.

Okay, so zero practical performance advantage for how much lower power? Trying to say that BayTrail has no advantage over the ARM competition would be the same as my saying that Haswell Y obliterates everything ARM has to offer without mentioning the fact that it has the same level of power delta as Baytrail does compared to performance-competitive ARM cores.