News - Intel to manufacture world's first 64-bit quad-core ARM chips

[HEXUS]

Altera Stratix 10 SX chips will utilise Intel 14 nm 3D Tri-Gate transistor technology.

Read more.

[3dcandy]

I have always wondered why Intel didn't do this before, it's an obvious way to make huge sums for themselves

[krazy_olie]

headscratch

[Kanoe]

I have always wondered why Intel didn't do this before, it's an obvious way to make huge sums for themselves

I had always expected that they were pretty much running to full capacity just manufacturing their own chips to sell. If they now have extra capacity to sell off does this mean they are scaling back the number of chips they produce for themselves?

It could be very bad for AMD if nVidia were able to get such a competitive advantage by producing chips on 14nm but i suspect Intel will be very picky about who they manufacture for, ie not big rivals in the same markets as them.

[CAT-THE-FIFTH]

The chip is for embedded usage and hence does not compete against them in the mobile market.

[3dcandy]

Oh yes, just to be clear, they don't need to fab chips that compete directly with their own stuff at all, but they must have spare capacity, they have more fabs than the rest put together...

* probably not true, it just appears that way *

[CAT-THE-FIFTH]

Oh yes, just to be clear, they don't need to fab chips that compete directly with their own stuff at all, but they must have spare capacity, they have more fabs than the rest put together...

* probably not true, it just appears that way *

It also shows that costs for 14NM are probably very high and that Intel is not making enough chips to cover those costs satisfactorily,so now they are trying to compete with independent fabs. AFAIK,they spend more on R and D than the next two largest fabs and ship less volume overall.

The issue is that their cheaper CPUs like Bay Trail appear to be sold at very low costs to penetrate the market,so Intel needs as much volume as possible. I also suspect,we will start to find the larger core based desktop and laptop CPUs getting smaller and/or getting more expensive,as Intel will need these to shoulder more and more of the cost of the process node development costs. However,the problem is that these markets are getting smaller and smaller,and Intel appears to be trying to target the server and supercomputer markets more,as these have helped negate the decline in traditional markets.The growth markets in the consumer area will be lower price chips like Bay Trail,not CPUs like a Core i5 4670K which probably have much larger margins.

So,expect Intel to get more and more companies to try and use their fabs. Its all a race to the bottom.

The problem is that SOCs like the Apple A7 still do very well against Bay Trail:

http://www.anandtech.com/show/7460/a...pad-air-review

That is a 28NM process node without finfets and it does very well. You then need to consider what happens next year when the TSMC 20NM process is up and running,which means further power consumption decreases for the ARM based SOCs.

None of these competing processes use finfets yet,which are supposed to help especially in low voltage scenarios. It makes me think what will happen in 2015 and 2016 when other companies start using them.

[MilesPebody]

Interesting. I think Intel have utilised ARM at least once before. E.g. they developed (and later sold to Marvell) the X-Scale series using ARMv5 architecture. Quite ancient history though.

[mud_z]

Please somebody slap me. What did I just read /(0_0)\

[Luke7]

Will this have anything to do with Project Denver and Maxwell? (Maxwell is an arm co-processor design, yes?)

Or am I getting completely the wrong end of the stick?

[3dcandy]

It also shows that costs for 14NM are probably very high and that Intel is not making enough chips to cover those costs satisfactorily,so now they are trying to compete with independent fabs. AFAIK,they spend more on R and D than the next two largest fabs and ship less volume overall.

The issue is that their cheaper CPUs like Bay Trail appear to be sold at very low costs to penetrate the market,so Intel needs as much volume as possible. I also suspect,we will start to find the larger core based desktop and laptop CPUs getting smaller and/or getting more expensive,as Intel will need these to shoulder more and more of the cost of the process node development costs. However,the problem is that these markets are getting smaller and smaller,and Intel appears to be trying to target the server and supercomputer markets more,as these have helped negate the decline in traditional markets.The growth markets in the consumer area will be lower price chips like Bay Trail,not CPUs like a Core i5 4670K which probably have much larger margins.

So,expect Intel to get more and more companies to try and use their fabs. Its all a race to the bottom.

The problem is that SOCs like the Apple A7 still do very well against Bay Trail:

http://www.anandtech.com/show/7460/a...pad-air-review

That is a 28NM process node without finfets and it does very well. You then need to consider what happens next year when the TSMC 20NM process is up and running,which means further power consumption decreases for the ARM based SOCs.

None of these competing processes use finfets yet,which are supposed to help especially in low voltage scenarios. It makes me think what will happen in 2015 and 2016 when other companies start using them.

I thought Samsung was going to 14nm and finfets? Can't find anything concrete yet though

[DanceswithUnix]

Will this have anything to do with Project Denver and Maxwell? (Maxwell is an arm co-processor design, yes?)

Or am I getting completely the wrong end of the stick?

Wrong end, or possibly the wrong stick

They are Nvidia designs, and I can't imagine any scenario where Intel will give Nvidia fab space. Actually I can imagine, it would take a change of CEO at Nvidia.

[Wozza63]

This + Windows 8.1 + mobile nvidia GPU = my next tablet

[keithwalton]

Keep your friends close, but keep your enemies closer.

What better way for intel to learn the strengths / weaknesses of armv8 designs than to fab some themselves.

It's not as if they're making them for arm, or qualcomm or nvidia but the obscure altera, who btw signed a manufacturing deal with intel back in February ...

Intel have so many fabs running various process nodes which they constantly update. From the fact the atom has moved to the latest node suggests they have no shortage of capacity.

[kingpotnoodle]

I have always wondered why Intel didn't do this before, it's an obvious way to make huge sums for themselves

Probably because there are better margins in retail sales of your own chips than wholesale manufacture and sale at wholesale price of someone elses chips.

[loyal986]

Even if intel would make all the Cars, train and every other tech, I would gladly use a self made bike and use a calculator(thanks GOD for other alternatives). They have good tech but I hate them, greedy bastards. I have a laptop with intel 4500MHD graphic card that is direct x 10 capable but hey intel needs money and why make windows 8 drivers for this card. I wish intel to vanish...