Intel intentionally delayed release of Broadwell?

[Fjodor2001]

http://news.techeye.net/chips/chipzilla-speeds-up-broadwell-release

Intel claimed that the decision to postpone mass production of Broadwell was because of slow demand for personal computers in general and microprocessors and yields that were below Intel's comfortable level.

I know others here on the forum has claimed such intentional delays by Intel due to factors other than unexpected design/manufacturing problems to be just fantasy. But now it seems like they delayed Broadwell intentionally partly due to "slow demand" after all. To me it sounds like they could release Broadwell earlier if they wanted to, but intentionally decided not to. Interesting... :hmm:

 

[mikk]

Shortly after Intel made its official announcement, slides from the company’s roadmap made it to the Internet and revealed that the chip giant only plans to start rolling-out its mainstream Broadwell microprocessors for desktops and laptops only in late Q4 2014.

Shilov is dumb as usual. There is no laptop Roadmap with late Q4 2014 for BDW.

 

[dave_the_nerd]

Why make a product that won't sell? They don't owe you a faster CPU every 6 months.

 

[MrK6]

Not really surprising, and it's not like Intel has much competition in the PC market anyway. Shame because I'm looking towards Broadwell for my next build, maybe not now.

 

[ShintaiDK]

Shilov is dumb as usual. There is no laptop Roadmap with late Q4 2014 for BDW.

^^ This.

 

[rgallant]

http://forums.anandtech.com/showthread.php?t=2364700
when those that think a 920 x58 is state of the art for surfing the net/gaming -what gaming ??? pong
yet does not list any off his equipment for gaming . but yet might spend $500.00 a year for a smart phone that does the same nothing the year before and the year before that.

 

[mrmt]

Not really surprising, and it's not like Intel has much competition in the PC market anyway. Shame because I'm looking towards Broadwell for my next build, maybe not now.

Not really surprising? It would be very surprising if Intel took that decision. Why the heck they would delay the introduction of a new node, you know, a node that should give them better economics than the previous one?

 

[GreenChile]

Intel has never and will never intentially delay the transition to a new node due to slow demand. They will simply scale back output to match the demand. Broadwell was clearly stated to be delayed due to yield defect issues.

 

[ninaholic37]

Why the heck they would delay the introduction of a new node, you know, a node that should give them better economics than the previous one?

Maybe they made too many Ivy Bridge / Haswell chips and need to sell them off first?

 

[mrmt]

Maybe they made too many Ivy Bridge / Haswell chips and need to sell them off first?

That would be reflected in the balance sheet

 

[Xpage]

They will make broadwell chips on time, just not mass produced. They will do low number runs to improve yields so that the initial reduction of profit from the transition to a new node with lower yields will not occur or will be shortened.


1- Do low # runs until yield increases or demand increased
2- ????
3- profit more

 

[positivedoppler]

Why make a product that won't sell? They don't owe you a faster CPU every 6 months.

I wish some of the Intel Zealots who wish death to AMD would realize this lol.

 

[Fjodor2001]

My guess is that Intel has made $X billion investment in Haswell/IvyBridge, so they want to sell a certain amount of chips to maximize return on investment. Looking at the percentage of PCs and laptops being sold that still has Ivy Bridge chips in them this seems likely.

Hence they choose to not release Broadwell until the previous investment has paid off as much as possible. They just intentionally keep it on the shelf ready to be released when it suits Intel.

The only reason they can act like this is lack of competition.

 

[Torn Mind]

Seems like the semiconductor companies need to node shrinks in order to cut costs quite a bit, but they can't do that if no one buys the stuff made on the previous node first. For Intel or any semiconductor company, their ideal would be accelerating their node cadence, not decelerating it since the benefits from node shrinks (more speed, less power consumption, less silicon used) would outweigh any increase in short term profits.

AMD sure would like GloFlo and TSMC to get a move on it with their new nodes and reap the benefits, but they're behind. Likewise, Intel wants to win mobile and if they had 7nm or whatever node length after 14 is, they'd have had a huge impact if they got to that node sooner. But that is not the case.

 

[meloz]

That statement is rubbish, Intel act as if demand is not affected by the product they are selling, or its price. I do not buy that yield talk either, they are reducing the pace of migration to 14nm. If the yields from the node were bad enough to matter they would actually need more capacity to fulfill the Atom and Broadwell notebook demand, not less.

If AMD were competitive on the desktop front, you can be sure we would get Broadwell for desktop in Q2 2014, instead we get Haswell with 100 MHz bump.

 

[Torn Mind]

That statement is rubbish, Intel act as if demand is not affected by the product they are selling, or its price. I do not buy that yield talk either, they are reducing the pace of migration to 14nm. If the yields from the node were bad enough to matter they would actually need more capacity to fulfill the Atom and Broadwell notebook demand, not less.

If AMD were competitive on the desktop front, you can be sure we would get Broadwell for desktop in Q2 2014, instead we get Haswell with 100 MHz bump.

If AMD hadn't intentionally delayed their node shrink prior to Conroe's release, they would've been more competitive with Intel. Likewise, they'd kill to have GloFlo get their next process node up and running to save costs and get better performing chips.

Intel might profit more in the short-term, but they'd be destroyed in the long-term if 14nm is delayed to the point TSMC and GloFlo get their next node shrinks up and running. They do not have any major advantages in the market where the money is, which is in phones and tablets. If lowering power consumption while increasing performance is the name of the game, which it is in mobile, the node shrink is absolutely critical for achieving such goals. But since developing a node requires immense investment in research, design, and building fabs, selling goods made from that node is needed to pay off those and other costs.

 

[Pheesh]

the passage you quoted mentions yield issues. If they had awesome yield right from the jump they would be rushing it to market as fast as possible in certain segments.

 

[MrK6]

Not really surprising? It would be very surprising if Intel took that decision. Why the heck they would delay the introduction of a new node, you know, a node that should give them better economics than the previous one?

You're taking your own assumptions as fact, which they aren't. Each node shrink has a much higher cost of entry as well as higher R&D and refinement. The days of happily plunging into a new node aren't as we get closer to the limits of silicon.

the passage you quoted mentions yield issues. If they had awesome yield right from the jump they would be rushing it to market as fast as possible in certain segments.

This and more. Intel is a company of experts at what they do. I've seen them make very few mistakes and this move is planned like any other.

 

[witeken]

You're taking your own assumptions as fact, which they aren't. Each node shrink has a much higher cost of entry as well as higher R&D and refinement. The days of happily plunging into a new node aren't as we get closer to the limits of silicon.

Transistor cost is still getting lower at an exponential rate, and Intel is in fact going to bring it down even faster than their current rate.

Since Intel has the necessary money, they're still happily plunging into new nodes while the others are getting further behind.

 

[MrK6]

Transistor cost is still getting lower at an exponential rate, and Intel is in fact going to bring it down even faster than their current rate.

Since Intel has the necessary money, they're still happily plunging into new nodes while the others are getting further behind.

The slide you posted shows the benefits of going to a new node, there's nothing new there. What it doesn't show is the cost of going to a new node, which is what we're discussing. It's true Intel has more money than the rest to keep pushing technology, however, at each point there is a cost/benefit analysis to estimate the ROI. When that timeline or ratio becomes unfavorable, they'll hold off just like their competition has.

 

[mrmt]

You're taking your own assumptions as fact, which they aren't. Each node shrink has a much higher cost of entry as well as higher R&D and refinement. The days of happily plunging into a new node aren't as we get closer to the limits of silicon.

Nope, that's not my own assumption, that's THE industry standard. That's Moore's law itself. If your next node doesn't have better economics than the previous node, you screwed up and have to go back to the drawing board.

 

[GreenChile]

The slide you posted shows the benefits of going to a new node, there's nothing new there. What it doesn't show is the cost of going to a new node, which is what we're discussing. It's true Intel has more money than the rest to keep pushing technology, however, at each point there is a cost/benefit analysis to estimate the ROI. When that timeline or ratio becomes unfavorable, they'll hold off just like their competition has.

Do you need glasses? The top left graph of the slide shows the normalized cost/transistor. That IS the cost of going to a new node. This conspiracy of Intel intentionally delaying 14nm due to slow demand is baseless without supporting evidence. 14nm will give them an added edge, especially in mobile where it is needed the most and if it was an option I'm certain they release it sooner rather than later.

 

[MrK6]

Do you need glasses? The top left graph of the slide shows the normalized cost/transistor. That IS the cost of going to a new node. This conspiracy of Intel intentionally delaying 14nm due to slow demand is baseless without supporting evidence. 14nm will give them an added edge, especially in mobile where it is needed the most and if it was an option I'm certain they release it sooner rather than later.

I think you need to learn how to read graphs rather than insult other members because you don't understand the argument. Companies don't magically push a button and *poof* a new node drops. It takes a lot of R&D cost as well as refinements/respins to get the silicon where the need it.

 

[kimmel]

I think you need to learn how to read graphs rather than insult other members because you don't understand the argument. Companies don't magically push a button and *poof* a new node drops. It takes a lot of R&D cost as well as refinements/respins to get the silicon where the need it.

The cost of the transistor includes the R&D costs normalized across the expected volume. If this graph didn't work like this Moores law wouldn't exist. The reason the industry moves forward is that it's actually cheaper to build devices on newer nodes with the shrink. Counter-intuitive but true.

 

[GreenChile]

I think you need to learn how to read graphs rather than insult other members because you don't understand the argument. Companies don't magically push a button and *poof* a new node drops. It takes a lot of R&D cost as well as refinements/respins to get the silicon where the need it.

And that is all built into the cost. Development and capital equipment costs are all calculated into the cost per transistor. I'm sorry if you can't understand this point or simply don't believe it but it is true. That graph is representative of the overall cost reduction by going to 14nm.