Intel to Fab Altera FPGAs at 14nm

[Blandge]

http://www.bloomberg.com/news/2013-0...rocessors.html

Intel will build so-called field programmable gate arrays, or FPGAs, for San Jose, California-based Altera when the two chipmakers move to 14-nanometer production technology, the companies said today in a statement.

In a separate statement, Altera said Taiwan Semiconductor will remain its “primary foundry,” and the two companies will work together on future production technology.

[dave_the_nerd]

14nm? Want.

[NTMBK]

Intel still struggling to get its fab utilisation up, then?

[dave_the_nerd]

Originally Posted by NTMBK

Intel still struggling to get its fab utilisation up, then?

Do they have any 14nm products of their own?

Seems to me getting the kinks worked out with other peoples' products before the post-Haswell 'Tick' isn't a bad idea.

And they can sell as much 14nm NAND as they can make.

[jones377]

Intel is the biggest or one of the biggest customers for FPGA chips, so I am sure they will buy most or all of them. Intel needs this if they are to manage future R&D while being so much ahead in process tech compared to the rest of the industry.

[ShintaiDK]

Originally Posted by NTMBK

Intel still struggling to get its fab utilisation up, then?

I assume its more a prospect of buying the company.
http://newsroom.intel.com/community/...-manufacturing
http://www.eetasia.com/ART_880066185...T_ccb1dfe7.HTM

All the same category.

[Blandge]

Originally Posted by NTMBK

Intel still struggling to get its fab utilisation up, then?

Intel's custom foundry business is completely separate from its internal manufacturing. Desipite what some may initially believe, Intel did not suddenly think, "Hey! We have excess manufacturing tools. Lets put them to good use." Instead, they've planned for this and built out additional capacity to accomodate for their foundry business.

[BrightCandle]

Altera in order to get access to this isn't allowed to manufacturer anywhere else at 14nm, only at larger nodes. This means Intel's process will have the exclusive on the top end and cheapest chips.

[Charles Kozierok]

IIRC Intel and Altera already had a joint project recently involving embedded Atom chips combined with FPGAs.

[Blandge]

Originally Posted by Charles Kozierok

IIRC Intel and Altera already had a joint project recently involving embedded Atom chips combined with FPGAs.

Yes that would be the Stellarton, AKA Intel Atom E600C processor family.

http://ark.intel.com/products/codename/42360/Stellarton

http://www.eejournal.com/archives/ar...23-stellarton/

Intel used to call it by the code name Stellarton, but it’s now officially known as the Atom E600C processor family. It puts two separate silicon chips in the same package...

[Phynaz]

Originally Posted by NTMBK

Intel still struggling to get its fab utilisation up, then?

I would think it's the other way around. You have to go to Intel (and pay their rates) if you want to produce chips at 14nm anytime soon.

[Blandge]

Nvm

[videogames101]

Originally Posted by Phynaz

I would think it's the other way around. You have to go to Intel (and pay their rates) if you want to produce chips at 14nm anytime soon.

when you put it that way

[Pilum]

Originally Posted by BrightCandle

Altera in order to get access to this isn't allowed to manufacturer anywhere else at 14nm, only at larger nodes. This means Intel's process will have the exclusive on the top end and cheapest chips.

No, Altera will also work with TSMC for 20nm production (Link in German). Intel production will be for the top end, a deal which should work out for both sides; Alteras top FPGAs cost up to $30,000, there should be enough profit in there for Intel.

[NTMBK]

Originally Posted by Phynaz

I would think it's the other way around. You have to go to Intel (and pay their rates) if you want to produce chips at 14nm anytime soon.

See, I see it the other way round. Intel's process is their ace in their hole, their single massive edge over anyone else, and they normally keep it very tightly locked down and under control. But when they are using less than 50% of their fabs, they have to get other customers in to make their economics work.

The classic Intel model has been to use the high volume of consumer PCs to subsidize fab development, and let them keep their high-value parts (e.g. Xeons) ahead of the competition. But between the current drop in PC sales due to the economic downturn, and Intel's ever decreasing die sizes (more chips per wafer = less wafers used = lower utilization) they're struggling to keep that equation balanced. Hence testing the water on outside customers.

Of course, if Intel wanted to get utilization up, they could start selling 6 core mainstream desktop parts.

[NTMBK]

This article has a few more interesting quotes from Intel: http://www.theregister.co.uk/2013/02...era_chip_deal/

If and when we are called upon to serve large mobile customers who can drive a lot more volume, we could serve them today in terms of capability.

[Pilum]

Originally Posted by NTMBK

See, I see it the other way round. Intel's process is their ace in their hole, their single massive edge over anyone else, and they normally keep it very tightly locked down and under control. But when they are using less than 50% of their fabs, they have to get other customers in to make their economics work.

Not really; they wouldn't work with Altera in this case. Altera has quarterly revenues of ~$500M. As pointed out above, the cooperation is only about the top-end FPGAs, Altera mass production is scheduled for TSMC 20nm. So let's say $100M quarterly revenues generated with Intel-produced chips. This amounts to maybe 1% of Intels fab capacity, which is pretty much insignficant. This is probably more about extremely lucrative margins on some high-revenue silicon.

[Blandge]

Originally Posted by NTMBK

See, I see it the other way round. Intel's process is their ace in their hole, their single massive edge over anyone else, and they normally keep it very tightly locked down and under control. But when they are using less than 50% of their fabs, they have to get other customers in to make their economics work.

The classic Intel model has been to use the high volume of consumer PCs to subsidize fab development, and let them keep their high-value parts (e.g. Xeons) ahead of the competition. But between the current drop in PC sales due to the economic downturn, and Intel's ever decreasing die sizes (more chips per wafer = less wafers used = lower utilization) they're struggling to keep that equation balanced. Hence testing the water on outside customers.

Of course, if Intel wanted to get utilization up, they could start selling 6 core mainstream desktop parts.

You are on the right track, but you're looking at this too narrow.

Is the PC segment declining? Yes
Is Intel looking for ways to expand it's portfolio to increase revenue? Yes
Is growing it's foundry business one way it is expanding it's portfolio? Yes
Is Intel fabbing chips for Altera because it has excess capacity in its 14nm process? No

Intel is fabbing chips for Altera to grow their foundry business and establish a source of revenue that isn't dependent on the PC industry. This isn't a reaction to excess capacity, this is a proactive measure to ensure the growth of their business.

Why do you think TSMC is going to remain Altera's primary manufacturer for bulk product? The answer is that they will be paying Intel so much for 14nm wafers that they wouldn't earn any money on lower ASP SKUs. I think you can safely bet that Intel is going to be making good money in this deal.

However, if Intel picks up an ultra high volume mobile customer like is mentioned in your previous quote, then your statement about economies of scale subsidizing Intel's high margin business will undoubtedly be true.

[jpiniero]

When I saw this, I was like "Hey, Intel could make a killer Bitcoin processor!"

If and when we are called upon to serve large mobile customers who can drive a lot more volume, we could serve them today in terms of capability.

I think you would only see this if they were willing to admit defeat to ARM.

[podspi]

They're probably talking about semi-custom Atom chips...

[Phynaz]

Originally Posted by Blandge

Intel is fabbing chips for Altera to grow their foundry business and establish a source of revenue that isn't dependent on the PC industry. This isn't a reaction to excess capacity, this is a proactive measure to ensure the growth of their business.

To expand upon just this a bit, I believe the US Govt. had something to do with Intel getting the Altera business. I think it was because their chips are used by the military.