[Digitimes] Intel building 14nm chips for Spreadtrum

[Arachnotronic]

Intel is making 14nm application processors for Spreadtrum Communications with sample shipments starting in October, and the pair is eyeing orders from Samsung Electronics, according to industry sources.

Spreadtrum with its 14nm chips built by Intel is looking to grab orders for Samsung's mid-range smartphone series for 2017, said the sources.

http://www.digitimes.com/news/a20161003PD203.html

 

[ShintaiDK]

Apple, LG etc. You go Intel foundry or you lose out.

 

[oak8292]

Intel has shown some patience and is still developing it's foundry services with new 10nm development kits and working with ARM to provide a baseline of IP, however the value proposition is decreasing. Intel has a process lead without a lot of third party IP support. TSMC has a strong portfolio of IP and is catching up in process technology with the very large volume of wafers from Qualcomm and Apple paying for R&D and equipment.

Here is a link to some opinions from JP Morgan from 2014 on the positions of Intel and TSMC when Altera first moved to Intel foundry and prior to the Altera buyout. The almost two year delay for the Stratix 10 probably does not instill a lot of confidence in third parties. A quote;

• On the design rules, we think Intel needs to improve its flexibility to match Foundries like TSMC. This could result in prospective Intel Foundry customers having concerns about moving to Intel, in our view, since FPGA products (that Altera makes) are usually the simplest products to make in any Foundry process. Mobile SoCs, baseband chips etc are much more complicated and are likely to encounter even more yield problems.
  

http://investordiscussionboard.com/sites/default/files/post-attachments/113857/apple-tmsc.pdf

P.S. If the Spreadtrum SOC has an integrated baseband processor it will be the first from Intel.

 

[Nothingness]

Apple, LG etc. You go Intel foundry or you lose out.

Apple did not go to Intel for foundry and don't seem to go that route for future SoCs (yet?)). They are using Intel modems in some of their iPhone, which is not the same.

Back to the Spreadtrum news, I guess this is for ARM-based SoCs. Another bad news for x86, which makes me happy

 

[raghu78]

Apple, LG etc. You go Intel foundry or you lose out.

Apple is getting the best process out there for mobile chips with good yields and no major supply issues. TSMC 16FF+ in 2015/2016 . TSMC 10nm in 2017 and TSMC 7nm in 2018. Throw in InFo into the mix and its easy to see TSMC has pretty much got Apple all to themselves for the forseeable future. Intel does not stand a chance against the TSMC juggernaut.

 

[ShintaiDK]

Apple did not go to Intel for foundry and don't seem to go that route for future SoCs (yet?)). They are using Intel modems in some of their iPhone, which is not the same.

Back to the Spreadtrum news, I guess this is for ARM-based SoCs. Another bad news for x86, which makes me happy

It points to Apple is going to use 10nm Intel.

 

[ShintaiDK]

Apple is getting the best process out there for mobile chips with good yields and no major supply issues. TSMC 16FF+ in 2015/2016 . TSMC 10nm in 2017 and TSMC 7nm in 2018. Throw in InFo into the mix and its easy to see TSMC has pretty much got Apple all to themselves for the forseeable future. Intel does not stand a chance against the TSMC juggernaut.

TSMC is way behind Intel if you didn't know already. Years behind on the electrical side. Then they can rename the processes as much as they want to make it look good on paper. Why do you think LG picked Intel for 10nm mobile SoCs?

 

[Nothingness]

It points to Apple is going to use 10nm Intel.

It's not a fact contrary to Spreadtrum and LG using ICF.

 

[Arachnotronic]

It's not a fact contrary to Spreadtrum and LG using ICF.

Yep. I suspect Intel is working very hard to at least get into the mix, though. 10nm should be roughly equal to TSMC 7nm in terms of xtor density, and Intel has typically led the industry in terms of xtor drive strength at a given leakage current.

However, there is much more to winning a foundry contract than just the technology (not that I need to tell you this), so even if Intel has the better tech, there's no guarantee that it'll actually win the biz.

If I were Apple, I wouldn't put the fate of my iPhone biz in the hands of an unproven foundry newbie. I would wait to see how those Spreadtrum and LG collaborations go first, at the very least

TSMC is a fierce, paranoid, and very well run company. They aren't going to lose (or split) the Apple contract without a serious fight.

 

[oak8292]

Apple and Qualcomm want early access to advanced nodes with large volumes of wafers. Would Intel put an Apple wafer ahead of an x86 wafer if it wasn't getting the yield it needed to meet contracts?

Intel is working hard to get into the mix of taking second tier vendors that backfill TSMC capacity. Both TSMC and Intel need 'trailing' customers. Intel is now on a five year equipment depreciation schedule.

 

[lefty2]

The only reason Spreadtrum uses Intel foundry is because they were paid to do so. Intel gave them 1.5 billion dollars.

 

[Arachnotronic]

The only reason Spreadtrum uses Intel foundry is because they were paid to do so. Intel gave them 1.5 billion dollars.

This is incorrect. Intel invested $1.5 billion in Tsinghua Unigroup and got a minority ownership stake in the company in return. That is far different than just "giving" them $1.5 billion.

 

[lefty2]

This is incorrect. Intel invested $1.5 billion in Tsinghua Unigroup and got a minority ownership stake in the company in return. That is far different than just "giving" them $1.5 billion.

You can call it "investing" if you want, but you'd have to be an idiot to think that the two things weren't connected.


Personal attacks are not allowed
Markfw900

 

[ShintaiDK]

Apple and Qualcomm want early access to advanced nodes with large volumes of wafers. Would Intel put an Apple wafer ahead of an x86 wafer if it wasn't getting the yield it needed to meet contracts?

Intel is working hard to get into the mix of taking second tier vendors that backfill TSMC capacity. Both TSMC and Intel need 'trailing' customers. Intel is now on a five year equipment depreciation schedule.

The required yield factor for Intels x86 chips and mobile ARM SoCs at the price Apple for example is willing to pay is entirely different.

 

[witeken]

http://www.digitimes.com/news/a20161003PD203.html

Nice, so we will finally see how Intel's process does in mobile chips (power, density)? Although Spreadtrum isn't the most exciting name.

 

[oak8292]

The required yield factor for Intels x86 chips and mobile ARM SoCs at the price Apple for example is willing to pay is entirely different.

I am having a little difficulty parsing the exact meaning of this sentence. However, I agree that Intel expects very high yields to meet their financial metrics. To achieve that yield Intel has about 35 SKU's from 5 die and they do a lot of binning and price optimization to make demand meet supply at different yield points. They have very good margins based on this strategy.

Mobile ARM SoCs have minimal if any binning and they need to meet the demands of time to market which may mean suboptimal yields per wafer. The money is in the end product and time to market is very important.

How much is Spreadtrum paying for 14nm wafers from Intel? Will LG get 10nm wafers to meet LG's production schedule or Intel's?

 

[Arachnotronic]

Nice, so we will finally see how Intel's process does in mobile chips (power, density)? Although Spreadtrum isn't the most exciting name.

One of the largest merchant mobile chip vendors in the market. Spreadtrum may ultimately be used as a proxy by Intel to fight MediaTek, Qualcomm, and so on.