Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/intel-has-announced-its-first-major-customer-for-its-new-intel-foundry-services-business-qualcomm.14455/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Intel has announced its first major customer for its new Intel Foundry Services business: Qualcomm

tonyget

Active member

Intel has announced its first major customer for its new Intel Foundry Services business: Qualcomm. Best known for designing the Snapdragon chips that power most major Android phones, Qualcomm will start to have its chips manufactured by Intel in the coming years using Intel’s upcoming 20A process.

No timeframe has been announced for when the first Intel-made Qualcomm chips will arrive or which of Qualcomm’s products Intel will produce.

Additionally, Amazon’s AWS will be working with Intel Foundry Services, relying on Intel’s packaging solutions (although Intel won’t be directly making chips for Amazon).

Screen_Shot_2021_07_26_at_5.33.38_PM.png

Qualcomm will rely on Intel’s newly announced Intel 20A technology node, scheduled for release in 2024. Intel 20A will introduce a new transistor architecture, RibbonFET, Intel’s first since 2011.

Intel previously announced its new foundry business as part of new CEO Pat Gelsinger’s “IDM 2.0” strategy shortly after he took the reins of the company. Intel Foundry Services was a key part of that plan, one that would see Intel expand beyond making its own chips to handle production for third-party companies. Qualcomm and Amazon mark the first — and the highest-profile — partners that Intel has announced so far for IFS, although Gelsinger has previously mentioned that Intel is in talks with over 100 companies for foundry work.
 

Intel has announced its first major customer for its new Intel Foundry Services business: Qualcomm. Best known for designing the Snapdragon chips that power most major Android phones, Qualcomm will start to have its chips manufactured by Intel in the coming years using Intel’s upcoming 20A process.

No timeframe has been announced for when the first Intel-made Qualcomm chips will arrive or which of Qualcomm’s products Intel will produce.

Additionally, Amazon’s AWS will be working with Intel Foundry Services, relying on Intel’s packaging solutions (although Intel won’t be directly making chips for Amazon).

Screen_Shot_2021_07_26_at_5.33.38_PM.png

Qualcomm will rely on Intel’s newly announced Intel 20A technology node, scheduled for release in 2024. Intel 20A will introduce a new transistor architecture, RibbonFET, Intel’s first since 2011.

Intel previously announced its new foundry business as part of new CEO Pat Gelsinger’s “IDM 2.0” strategy shortly after he took the reins of the company. Intel Foundry Services was a key part of that plan, one that would see Intel expand beyond making its own chips to handle production for third-party companies. Qualcomm and Amazon mark the first — and the highest-profile — partners that Intel has announced so far for IFS, although Gelsinger has previously mentioned that Intel is in talks with over 100 companies for foundry work.

Unless of course Intel 20A is late. QCOM plays the foundry field quite well and likes to tweak TSMC and Samsung whenever possible. This is clearly a tweak.
 
Warning: Truth bomb!

This really is a nothing burger. QCOM has made many of these announcements that have resulted in nothing. QCOM has been a foundry whore since the beginning. In fact, thanks to QCOM, TSMC's IP was spread amongst the other foundries for 2nd, 3rd, and 4th source manufacturing. QCOM would design to TSMC first then send GDS to UMC, SMIC, and Chartered (GF) for replication. This spawned many generations of "T like" processes. This stopped with FinFETs and QCOM's foundry panties have been bunched up ever since.
 
why can't you spread TSMC's IP with FinFETs?

FinFETs are not created equal like CMOS transistors. Intel 14nm, Samsung 14nm, TSMC 16nm, UMC, 14nm, and SMIC 14nm are all very different processes. design IP must be requalified/characterized on each node and design GDS is not even close to being compatible.

Some chip companies will design different products to different foundries but none to my knowledge have manufactured the identical product with different foundries successfully. Apple tried to do this at 14nm Samsung and 16nm TSMC (iPhone 6s) and the infamous Chipgate resulted. Same design but different die size and different performance and power.

Does this answer your question?
 
FinFETs are not created equal like CMOS transistors. Intel 14nm, Samsung 14nm, TSMC 16nm, UMC, 14nm, and SMIC 14nm are all very different processes. design IP must be requalified/characterized on each node and design GDS is not even close to being compatible.

Some chip companies will design different products to different foundries but none to my knowledge have manufactured the identical product with different foundries successfully. Apple tried to do this at 14nm Samsung and 16nm TSMC (iPhone 6s) and the infamous Chipgate resulted. Same design but different die size and different performance and power.

Does this answer your question?
yes. thank.
 
Stuff not said is as important as stuff said, so does that mean QCOM is not using Intel 7, 5 & 4?
My guess is no. None of these nodes will be leading edge on release, and will probably be more expensive than foundry to boot, and Qualcomm is probably already in deep with Samsung/TSMC for this generation.
 
GF 14 is Samsung 14. GF 12 is Samsung 12.
Not exactly the case either, at least based on what was published at VLSI in IIRC 2018? 1) Samsung calls their half-node "11" and 2) the library architectures for GF12 uHD and S11LPP uHD are different. AMD's 12nm parts are not using this library anyhow, so maybe the individual module improvements over the common 14nm are consistent for both Foundries, but I would doubt that. I've yet to see a true dual-sourced design on S14LPP and GF14, but if the DRs are consistent presumably one can migrate a design easily.

SMIC 14 OTOH is a copy of TSMC N16FFC, this is pretty clear from the Kirin710A teardowns from Tech Insights. So again not likely to be true dual-sourced, but porting designs between the two nodes should be relatively straightforward. I don't recall seeing the uHD library from N12FFC in that design, just 9T/7.5T, so TSMC should still have an advantage there (plus the RF capability, eNVM, etc etc)
 
Back
Top