Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/qualcomm-to-continue-using-multiple-foundries-for-chip-production.17074/
        )

    [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
)

Qualcomm to continue using multiple foundries for chip production

tonyget

Active member

US chip firm hints at reusing Samsung for GAA
TSMC dominated in 4nm and first batches of 3nm

Qualcomm SVP Don McGuire Image: TheElec

Qualcomm SVP Don McGuire Image: TheElec​
Qualcomm will continue using multiple foundries to manufacture its chips, one of its executives said on Tuesday.

The US chip giant was continuing to collaborate with Samsung Foundry and will use multiple foundries, from the South Korean company, TSMC to GlobalFoundries, depending on their technological maturity going forward, Qualcomm senior vice president Don McGuire said during Qualcomm’s Snapdragon Summit 2022 in Hawaii in a meeting with South Korean press.

Qualcomm has currently given all its workload for 4-nanometer (nm) and 3nm chips to TSMC, the world’s largest foundry.

McGuire’s comments indicate that the US firm could give orders to Samsung Foundry again for follow-up nodes such as gate-all-around (GAA).

At the summit, Qualcomm unveiled its latest application processor the Snapdragon 8 Gen 2.

The chip boasts 4.35 times increased AI performance and 25% faster processing speed compared to its predecessor.

Snapdragon 8 Gen 2 also uses Qualcomm’s new GPU and has a CPU with 40% increase in performance.

The chip will be made using TSMC’s 4nm node. For Snapdragon 8 Gen 1, Samsung Foundry was the initial contract manufacturer but Qualcomm gave the order to TSMC during the second half of last year.

Sources had said Samsung’s low yield rate for 4nm was the cause and the US chip firm will be giving TSMC the order for 3nm chips because of this.

McGuire said Qualcomm’s orders were too large for it to use a single foundry and using multiple foundries is not only advantageous in supply but also price and scale.

The US firm also needed multiple foundries to expand in other business areas besides smartphones, he added.

Meanwhile, Samsung, while it facing difficulty with yield with 4nm, was the first to start production of 3nm GAA chips.

TSMC has also started 3nm chip production but for these uses a FinFET structure rather than GAA. The Taiwanese giant is reportedly planning to apply GAA structure starting with 2nm.
 
Qualcomm has always used multiple foundries, it is their culture and MO. They may get better pricing but if they were exclusive with TSMC QCOM would get better chips, my opinion. From what I hear QCOM is working closely with IFS so Samsung Foundry has some serious competition for the NOT TSMC business. I will be at IEDM next month so we will know more after that.
 
if they were exclusive with TSMC QCOM would get better chips, my opinion.
I think that depends on the progress of 3GAP. If it is a real node before 20A then Samsung might have a product that is competitive for a short timeframe. However right now there is no disputing that TSMC is in a better place; hence why the switched to 4nm for their flagship products. Back when they initially started operating in this way I can't necessarily find fault with Qualcomm's logic. A few years ago Samsung was right on TSMC's heels and occasionally pulling out big wins. Unfortunately they've really fumbled the ball for the past couple of years.
 
I think that depends on the progress of 3GAP. If it is a real node before 20A then Samsung might have a product that is competitive for a short timeframe. However right now there is no disputing that TSMC is in a better place; hence why the switched to 4nm for their flagship products. Back when they initially started operating in this way I can't necessarily find fault with Qualcomm's logic. A few years ago Samsung was right on TSMC's heels and occasionally pulling out big wins. Unfortunately they've really fumbled the ball for the past couple of years.

TSMC N3E has more tape outs than I have seen in years. It will be a record node, absolutely.

GAA is a process and design challenge, like FinFETs, so it will take longer to get to HVM than you may think. Meanwhile TSMC will reap the benefits of the N3 family for years to come, absolutely.
 
TSMC N3E has more tape outs than I have seen in years. It will be a record node, absolutely.

GAA is a process and design challenge, like FinFETs, so it will take longer to get to HVM than you may think. Meanwhile TSMC will reap the benefits of the N3 family for years to come, absolutely.
No doubt. I was more so laying out they hypothetical of Samsung getting lots of yield and pdk learnings from 3GAE. If this happens maybe 3GAP could beat 20A or 18A to market. Given Samsung is a known quantity I could see Qualcomm really digging Samsung for this small period where they would probably have P&P parity with N3E and or i3. Of course this is all hypothetical, because it doesn’t seem like Samsung will really have large scale GAA until the “2nm” generation.
 
Back
Top