Intel becomes a foundry, offers up its 22nm process | Hacker News Reader

Intel becomes a foundry, offers up its 22nm process

extremetech.com

93 pointsFlemlord18 comments

Achronix has been in partnership with Intel for over a year:

http://online.wsj.com/article/SB1000142405274870447790457558...

http://www.eetimes.com/electronics-news/4210263/Intel-to-fab...

Gotta think that Intel and Apple must be having discussions about moving successors to the A5 chips over to Intel's process.

Getting Apple's ARM processors moved to 22mm would be a spectacular coup for both Apple and Intel. You'd think that the volume and funds Apple brings would paper over any hard feelings about not using x86.

It would be fantastic for apple, but not so sure about intel.

Remember, intel are right now trying to bring atom-- itself not even on the 22nm process yet-- to exactly the space that the A5 occupies. It would seem suicidal to give away their process advantage against ARM at exactly the time they are pushing hard to get Atom caught up with their lead process (going 22nm next year and 14nm the year after; the same time as their full power processors).

It will be interesting to see if there are any high end ARM chips made at intel, or if they just stick to areas they deem to not be competing with them (which can't be a lot that needs 22nm?)

Given that Apple processors will only be available in Apple devices, there's still plenty of room for Intel to sell a 22nm Atom while allowing Apple to take advantage of their 22nm process.

It would not be suicidal, why? Because they have the 22nm lead and will most likely keep the lead for years. It means that Apple will not be able to go back to another provider with the same performance. Which means they can lock Apple and force Apple hand. This would be brilliant.

It would be suicidal for their aim of getting atom into tablets and smartphones, I meant. (Which would almost certainly hold more profit than making someone else's processor.)

I figure it would take a while for that to happen, even if Intel wanted to do that. Plus, I think ARM signed a partnership with IBM to manufacture its chips at 14 nm, I believe in 2014, which would be very competitive with Intel timeline wise.

It could just be that Intel wants to increase their stock price and sees low hanging fruit for expanding revenue and net income in the foundry business which does not damage any of their other competitive positions.

They have ramped up capacity recently and it is beneficial to them to fill that capacity rather than shut it down if they can add some revenue.

Later if they need that capacity for higher margin products like some x86 phone processor that none of us know about Intel can end their agreements with the fabless semiconductor companies and produce their own.

It still competes with their CPU business if it takes fab capacity they could otherwise use for (much higher margin!) CPUs. I suspect, as the linked article posits, this is a hedge. They want to maintain the capability to do on-contract fab work, so partnering with a handful of small volume firms (who can be purchased if it turns out they are successful) makes sense.

Intel won't be hooking up with a big ARM SoC designer any time soon.

Yes, if there is a market for their higher end CPU business then that would make sense, but presumably they would be using those fabs for that business if they thought there was a market for those units at this point.

I'm saying that Intel may think that the market size for computer processors will continue to grow and this would be space they could fill if necessarily. Realistically though, they would probably just build more fabs.

They are using them for that business. Intel's fabs are at capacity, and they're building a new giant one in Oregon right now.

Not just D1X in Oregon. Fab 42 is being built in Chandler, Arizona as well.

They are also building one in China. A fab produces an enormous number of chips, they are building more capacity than is currently required by the market.

In a world when a unique soc design using the new processes is becoming more and more rare(because it costs a lot), most of those designs have large programmable components , to enable many uses and many chips sold.

Now, if you make the best programmable components, you'll sell most of the chips that use new processes, and thus financially strangle other chip manufacturers.

Another option might be that they're looking to integrate FPGAs on Haswell dies.

Seems pretty simple to me: they have spare capacity.

... and this how FabCloud was born ;)

"The other option is that Intel might be looking to boost the performance of its upcoming smartphone- and tablet-oriented Medfield platform."

I'd go with that one (the first reason was pretty implausible). Get the third party suppliers used to doing test chips on their fab, then pick and choose silicon-proven blocks like the SoC builders on ARM &co are used to.

Possibly, it would be a good idea to have more than one buyer (Intel) in that market ...

Reply on news.ycombinator.com