Intel might outsource 14nm prod to TSMC

[VanayadGaming]

Quote

Intel is encountering tight 14nm process production capacity in-house, and is looking to outsource part of its 14nm chip production to Taiwan Semiconductor Manufacturing Company (TSMC), according to industry sources.

 

Intel intends to give priority to its high-margin products mainly server-use processors and chipsets amid its tight 14nm process capacity, and therefore plans to outsource the production of its entry-level H310 and several other 300 series desktop processors to TSMC, the sources indicated.

https://www.digitimes.com/news/a20180910PD210.html

 

Quote

Intel's latest update is that its 10nm chips will not be ready for commercial production until the fourth quarter of 2019.

Intel had been adopting the so-called "Tick-Tock" model under which every microarchitecture change is followed with a die shrink of the process technology, but the "Tick-Tock" production cycle started slowing down in 2014. Intel's 14nm manufacturing node has become the "longest-lived" process technology in the chip giant's history, the observers noted.

I think the delay of the 10nm really will say its word when TSMC produces intel chips. 

And thinking AMD reaches 7nm before intel's 10...just makes me warm inside.

[Void Master]

I think we will get 14nm++++++++++

then after that 10nm+++++++++++++++++++++++++++++++++++++++++++++++++++++

If this "Tick-Tock" production cycle continues on slowing down.

And then AMD will start beating Intel as is was in the 20's.

 

[TVwazhere]

3 minutes ago, VanayadGaming said:

I think the delay of the 10nm really will say its word when TSMC produces intel chips. 

And thinking AMD reaches 7nm before intel's 10...just makes me warm inside.

[Guest]

Just now, Void Master said:

I think we will get 14nm++++++++++

More like 14nm####

1 minute ago, Void Master said:

And then AMD will start beating Intel as is was in the 20's.

2003 was gulden

[Void Master]

Just now, Warrie said:

2003 was gulden

I think 2019/2020 may be diamond !!!

[Guest]

Just now, Void Master said:

I think 2019/2020 may be diamond !!!

Just a wee bit of time will tell

[Void Master]

2 minutes ago, Warrie said:

Just a wee bit of time will tell

Well I am starting to have more hope in AMD then intel .

I am becoming an AMD fan .

[VanayadGaming]

2 minutes ago, Void Master said:

Well I am starting to have more hope in AMD then intel .

I am becoming an AMD fan .

let's see if we can get a better single thread score first

[Void Master]

@VanayadGaming

I believe if AMD works hard they might get 10% to 15% more single core performance  when the move from 12nm to 7nm, this will make them toe to toe with intel coffeelake .

And this will be great even if intel catches up with is 10nm !

And AMD still beats intel in performance per value which is amazing !

 

[AnalogCanavity]

23 minutes ago, Void Master said:

I think we will get 14nm++++++++++

then after that 10nm+++++++++++++++++++++++++++++++++++++++++++++++++++++

If this "Tick-Tock" production cycle continues on slowing down.

And then AMD will start beating Intel as is was in the 20's.

 

Sorry for the noob question, what do the plus stand for in the 14nm++.

 

Thanks

[Void Master]

Just now, Zic05 said:

Sorry for the noob question, what do the plus stand for in the 14nm++.

 

Thanks

it stands for improved , refined , reused process

[MMKing]

31 minutes ago, Zic05 said:

Sorry for the noob question, what do the plus stand for in the 14nm++.

 

Thanks

Nothing. Process node names are completely arbitrary. The only thing you can be certain of is that Intel 10nm is smaller than Intel 14nm, while TSMC 7nm is not noteworthy smaller than Intel 10nm.

 

What Intel is trying to convey with the ''+'' is improvements in manufacturing. As production ramps up, the foundries gain experience and are able to deliver products with fewer defects. Allowing the average chip to clock higher, run at a lower voltage and/or produce less heat.

 

There exist a standard for determining process name schemes, but due to the process name only containing a single value (transistors are 3D objects) and the realities of researching and producing transistors. The standard can only be loosely followed. This has been further complicated recently with the addition of fins, which makes the shape of transistors more complex, thus a single number value will be even worse at accurately describing the size.

 

 

As to the validity of these rumors... i highly doubt it. TSMC can't just adjust their 12FFC (12nm FinFetCompact) machinery to produce Intel P1272 (Intel 14nm) overnight. It doesn't really make allot of sense. If Intel requires allot of 14nm wafers in the very near future, TSMC can't really do much to help. If Intel needs allot of 14nm Wafers long term, there is little reason not to just increase Intel's own capacity. If the rumors are true, the circumstances would have to be very specific. Intel needs 14nm Wafers for a limited time, so paying a premium for a few hundred thousand Wafers without buying the tools to produce them.

 

I'm not really sure how TSMC would deliver either. TSMC is very busy producing the Apple A12 CPU, and is probably in the process of delivering 7nm to AMD for their upcoming ''Vega 20'' GPU for the professional/enterprise market. Further down the line in 2019, AMD and Nvidia will be needing allot of 7nm Wafers for the mainstream market. Considering global foundries is staying on 14nm, TSMC would be best off just selling a portion of their machinery to Intel and let Intel deal with the hassle of producing Intel P1272 wafers.

[AnonymousGuy]

This is not how silicon fabrication works.  You can't just send your design to another company and have them make it as though you're doing a 3D printing job.

 

Intel does use TSMC already for some stuff that isn't "worth" making in-house, but it's not an easy task to send an existing design.

[PatXioPC]

2 hours ago, Void Master said:

Well I am starting to have more hope in AMD then intel .

I am becoming an AMD fan .

I highly hope by how Intel and AMD are moving forward to building the process nodes for their chips, AMD will, eventually, develop the smallest ( I suppose will begin after the 7nm craze, already have moved their 7nm dies to TSMC) and surpass the gap between Intel clocking their cores so much higher than AMD.

They will kick back with smaller efficient transistors inside their dies, making the die smaller and fitting more and more cores into it and still sustaining high clock speed with a nice IPC. 

[Master Disaster]

Wow, in the near future we might end up in a situation were TSMC are producing wafer for Intel & AMD at the same time.

[Cedimedi]

This will be chipset only, maybe with  an adjusted design. With CPUs it wouldn't work great without a complete redesign

[Brooksie359]

4 hours ago, MMKing said:

Nothing. Process node names are completely arbitrary. The only thing you can be certain of is that Intel 10nm is smaller than Intel 14nm, while TSMC 7nm is not noteworthy smaller than Intel 10nm.

 

What Intel is trying to convey with the ''+'' is improvements in manufacturing. As production ramps up, the foundries gain experience and are able to deliver products with fewer defects. Allowing the average chip to clock higher, run at a lower voltage and/or produce less heat.

 

There exist a standard for determining process name schemes, but due to the process name only containing a single value (transistors are 3D objects) and the realities of researching and producing transistors. The standard can only be loosely followed. This has been further complicated recently with the addition of fins, which makes the shape of transistors more complex, thus a single number value will be even worse at accurately describing the size.

 

 

As to the validity of these rumors... i highly doubt it. TSMC can't just adjust their 12FFC (12nm FinFetCompact) machinery to produce Intel P1272 (Intel 14nm) overnight. It doesn't really make allot of sense. If Intel requires allot of 14nm wafers in the very near future, TSMC can't really do much to help. If Intel needs allot of 14nm Wafers long term, there is little reason not to just increase Intel's own capacity. If the rumors are true, the circumstances would have to be very specific. Intel needs 14nm Wafers for a limited time, so paying a premium for a few hundred thousand Wafers without buying the tools to produce them.

 

I'm not really sure how TSMC would deliver either. TSMC is very busy producing the Apple A12 CPU, and is probably in the process of delivering 7nm to AMD for their upcoming ''Vega 20'' GPU for the professional/enterprise market. Further down the line in 2019, AMD and Nvidia will be needing allot of 7nm Wafers for the mainstream market. Considering global foundries is staying on 14nm, TSMC would be best off just selling a portion of their machinery to Intel and let Intel deal with the hassle of producing Intel P1272 wafers.

Maybe they are making the chipsets not the cpus?

[yian88]

Who ever tought we will see such developments ? :))

 

Its funny because Intel doesnt even need 10nm or 7nm, all they have to do is remove the iGPU on their desktop chips which occupies 1/3 of the silicon wasted, and replace it with more cores for the same price, with their Ring bus architecture and higher OC, AMD will have competition for long time to come.

Intel stuborness to sell you 1/3rd useless silicon with every CPU is going to bury them, really it doesnt make any sense.

[GoldenLag]

2 minutes ago, yian88 said:

Who ever tought we will see such developments ? :))

 

Its funny because Intel doesnt even need 10nm or 7nm, all they have to do is remove the iGPU on their desktop chips which occupies 1/3 of the silicon wasted, and replace it with more cores for the same price, with their Ring bus architecture and higher OC, AMD will have competition for long time to come.

Intel stuborness to sell you 1/3rd useless silicon with every CPU is going to bury them, really it doesnt make any sense.

Ringbus doesnt scale well with extra cores. There is a reason meshbus exist. The more cores the worse latency and IPC. 

 

That iGPU is a huge part pf their sales in office spaces, doubt they will remove it unless they make a seperate "gaming" skue, but that is not where the money is at. Not to mention the iGPU is rather useful for hobby videoeditors and more due to the iGPU being utilized.

 

Also adding cores means adding heat, something that cant go on forever. Not to mention yield will get worse the more cores they add.

 

Intel very much needs a new macroarchitecture and node to keep competetive in the coming years.

[Trixanity]

14 hours ago, Brooksie359 said:

Maybe they are making the chipsets not the cpus?

They are; and primarily the lower end chipsets. It'd be too risky to start taping out high performance silicon at another foundry, not to mention the drop in confidence from investors.

[Arika]

2 hours ago, yian88 said:

all they have to do is remove the iGPU on their desktop chips which occupies 1/3 of the silicon wasted

except a large majority of people dont need a dGPU so cutting it out would actually drive up costs of the devices since they would need need a separate GPU of some description (regular desktop PCs, Laptops, Tablets etc). Intel's primary market is not gamers who need a powerful GPU and CPU