(Semiwiki) Intel 14nm Delayed Again?

[witeken]

On the desktop market, they are competing only with themselves. Nothing on this roadmap would make me want to move on from 22nm if I was an Intel shareholder or executive whose bonus structure is based off of earnings, not new shiny benchmarks:

It's really simple: smaller transistors -> more transistors/area -> cheaper transistors -> highers margins. But AMD doesn't force them to do anything, so it makes sense to use the 14nm fab capacity first for new, important markets as the tablet and smartphone markets. But economically, as long as you can pay it, it's always better to go to a smaller process node.

 

[witeken]

Just for the record, Semiwiki is TSMC paid PR.

Just like Semiaccurate?

 

[Headfoot]

That author just exudes bias. I wouldn't trust anything they said even if their information is accurate. They will just spin it and spin it

 

[mikk]

First TSMC 28nm product(FPGA) was released in Q4 2011(October), AMD officially released Tahiti(HD7970) in January 2012.
First TSMC 20nm products will be released this quarter (Q1 2014). That make it 28 months (February) or 29 months if they will release in March 2014.

What CPU/GPU/SoC based on TSMC 20nm will be available this quarter? All I heard is that somewhere in H2 we might see the first 20nm products if we are lucky while the bigger cores are unlikely to be available this year. Maxwell for example will be on 28nm this year. First HD7970 cards were available in December 2011. TSMC 20nm is more a 2015 thing like Intels 14nm and therefore 3 years between TSMC 28-->20nm is more accurate.

 

[AtenRa]

What CPU/GPU/SoC based on TSMC 20nm will be available this quarter? All I heard is that somewhere in H2 we might see the first 20nm products if we are lucky while the bigger cores are unlikely to be available this year. Maxwell for example will be on 28nm this year. First HD7970 cards were available in December 2011. TSMC 20nm is more a 2015 thing like Intels 14nm and therefore 3 years between TSMC 28-->20nm is more accurate.

Well it seams i was wrong, TSMC already have shipped 20nm products in Q4 2013.

Xilinx Ships Industry's First 20nm All Programmable Product

Nov 11, 2013
SAN JOSE, Calif., Nov. 11, 2013 /PRNewswire/ -- Xilinx, Inc. (NASDAQ: XLNX) today announced first customer shipment of the semiconductor industry's first 20nm product manufactured by TSMC, and the PLD industry's first 20nm All Programmable device. Xilinx UltraScale™ devices deliver an ASIC-class advantage with the industry's only ASIC-class programmable architecture coupled with the Vivado® ASIC-strength design suite and recently introduced UltraFast™ design methodology. The UltraScale devices enable 1.5X – 2X more realizable system-level performance and integration for customers, equivalent to a generation ahead of the competition.

So TSMCs 20nm first product is only 25 months after 28nm product (October 2011).

 

[Tristor]

I don't know, maybe I'm not in the bulk of enthusiasts, but there is no such thing in my mind as too fast. On the other hand, even I am starting to get to a point where there's simply nothing out there that's really pushing hardware in a way that's meaningful to me. Games would typically be that application, but the console portification of games has pretty much destroyed their ability to reliably drive hardware forward in the PC market. After that, no other application I do on a regular basis requires anything close to the power I had even 2 years ago.

Despite that, I don't continue to stop building a new PC every year for a long time, so I'll be one enthusiast at least who plans to help float the hardware industry. Intel is getting my dollars now, but I hope AMD does something non-stupid for once once bulldozer has run its course. Either way, I just don't see the "PC" market drying up. Maybe things are shifting to be more power/budget conscious over being top-performance conscious, but that's a shift in the market, not a reduction in the market. Smaller processes can help drive that market segment by increasing volume yields and lowering power requirements, so I really don't see the need to keep tick/tock decreasing anytime soon for Intel.

 

[mikk]

Well it seams i was wrong, TSMC already have shipped 20nm products in Q4 2013.

Xilinx Ships Industry's First 20nm All Programmable Product



So TSMCs 20nm first product is only 25 months after 28nm product (October 2011).

Shipments are irrelevant, what matters to us is when I can buy it.

 

[Fjodor2001]

How funny. Really puts the whole debate regarding how to interpret the earlier "We have confidence the [14 nm yield] problem is fixed because we have data it is fixed," statement by Intel's Krzanich from October 2013 in a new light. As discussed earlier in Intel 14 nm delayed thread.

 

[MisterMac]

So... there's a lot of Intel reasoning towards this could only in worst case scenario be a small speedbump.

The interesting is historic data - he was right before even if he's a loose-cannon that will blog anything anti-intel.


I find it funny even Intel17 had to immediately make a SeekingAlpha post.... just in case.



Is every single Intel fanboi(including myself) going to say it's nothing... IF IT is delayed?

And if it includes mobile as well?

 

[Homeles]

If it's delayed, it's not a small deal. It's a big one. Intel needs 14nm, and they need it now. They can definitely compete with ARM without it, but they'd be fighting an uphill battle. With 14nm, and the cost savings and performance benefits that come with it, not to mention all the products it enables, it would make Intel's lineup the easy choice for OEMs.

 

[Khato]

So TSMCs 20nm first product is only 25 months after 28nm product (October 2011).

More like 32 months - http://press.xilinx.com/2011-03-18-...elopment-Platform-for-Next-Generation-Systems

 

[bullzz]

I dont think its a question of If intel will be able to ship broadwell in H2'14. Its a question of if they hit say 50-60% yield. we dont know what yield tsmc is hitting with 20nm. there is no way they can tell apple or ms they dont enuf dies. even with 10% yield they will give apple/ms priority and push products out

 

[AtenRa]

More like 32 months - http://press.xilinx.com/2011-03-18-...elopment-Platform-for-Next-Generation-Systems

hmm you are right, 32 months from first 28nm shipment to first 20nm shipment.

 

[OCGuy]

If it's delayed, it's not a small deal. It's a big one. Intel needs 14nm, and they need it now. They can definitely compete with ARM without it, but they'd be fighting an uphill battle. With 14nm, and the cost savings and performance benefits that come with it, not to mention all the products it enables, it would make Intel's lineup the easy choice for OEMs.

Intel so far has only confirmed delaying desktop, which I don't think anyone can argue they "need" to release at this point..

14nm SoC has already been shown working, and is in a competitive market segment.

 

[ShintaiDK]

Intel so far has only confirmed delaying desktop, which I don't think anyone can argue they "need" to release at this point..

14nm SoC has already been shown working, and is in a competitive market segment.

Desktop delayed? How so, Broadwell was never intended to be a desktop chip. The Broadwell-K is just a newly exception.

Atoms are simply deemed more important than desktops for new nodes. So desktop is down to a 4th place in the queue for a new node.

 

[krumme]

Shipments are irrelevant, what matters to us is when I can buy it.

Tell that to the ones selling the cpu. For them the oem aib whatever is the market. The added time is just because there is value added. Eg. The cpu is integrated in a soc. The soc is put into a television. Its a chain. And we also buy an entire tv not only a cpu in a tray.

 

[krumme]

Avoiding Capex danger. Cost reductions. Intel adapts. You cant sell atom at a huge loss and keep new proces tech comming on the same pace. You got a take hour choice. Sounds reasonable to me.

At least we avoid in this thread the idiotic bashing about someone beeing incompetent. Thank you.

 

[mikk]

Tell that to the ones selling the cpu. For them the oem aib whatever is the market. The added time is just because there is value added. Eg. The cpu is integrated in a soc. The soc is put into a television. Its a chain. And we also buy an entire tv not only a cpu in a tray.

At least he should compare shipment against shipment and not shipment against availability or launch. His comparison is nonsense.

 

[krumme]

At least he should compare shipment against shipment and not shipment against availability or launch. His comparison is nonsense.

You are probably right. I didnt and will not read the source.

 

[AtenRa]

At least he should compare shipment against shipment and not shipment against availability or launch. His comparison is nonsense.

I already acknowledged that 28nm shipment to 20nm shipment is 32 months. First products are being expected for Q1 2014 release and that put products to products even quicker and in no way at 3 years as was being implied.

 

[Arachnotronic]

So... there's a lot of Intel reasoning towards this could only in worst case scenario be a small speedbump.

The interesting is historic data - he was right before even if he's a loose-cannon that will blog anything anti-intel.


I find it funny even Intel17 had to immediately make a SeekingAlpha post.... just in case.



Is every single Intel fanboi(including myself) going to say it's nothing... IF IT is delayed?

And if it includes mobile as well?

If it's delayed, then Intel had better hope that the ARMy runs into similar problems with its 16FF, otherwise the playing field could become much more "equal" re: process tech, which eliminates a key long-term advantage of Intel's.

 

[jpiniero]

If Broadwell really does get delayed into 2015 then you almost have to wonder if they will just cancel it and roll with Haswell Refresh until Skylake is ready.

With 14nm, and the cost savings and performance benefits that come with it

That's the problem - the benefit of going to a new node is less and less because the costs have skyrocketed. The bar that Intel has to reach in order for 14 nm to make sense economically may simply be too high at this point.

 

[OCGuy]

If Broadwell really does get delayed into 2015 then you almost have to wonder if they will just cancel it and roll with Haswell Refresh until Skylake is ready.



That's the problem - the benefit of going to a new node is less and less because the costs have skyrocketed. The bar that Intel has to reach in order for 14 nm to make sense economically may simply be too high at this point.

Broadwell is in production. This is purely a marketing strategy.

 

[Homeles]

Broadwell is in production. This is purely a marketing strategy.

I'm not so certain that it's in production yet. It's supposed to go into production this quarter... and there's a fair bit of time before Q1 ends. It almost seems like all deadlines and launches happen at the end of a quarter.

 

[mrmt]

Just like Semiaccurate?

Semiaccurate doesn't seem to have any relationship with TSMC. When talking about foundry, they bring mostly Globalfoundries stuff, probably because of their relationship with AMD. Regarding foundries I find their cover fairly bland.