Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/samsung-3nm-gaa-hows-the-gamble-paying-off.15657/
        )

    [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
)

Samsung 3nm GAA - how's the gamble paying off?

R

rgrindley

New member
GAA transition will be hard. INTC targeting 2024 on 20A and TSMC on 2nm on a similar timeline, I believe. Samsung decided to gamble and start early, to get a jump on the learning curve. I assume the bet is that while Samsung 3nm may be a write-off, they will work out the kinks and be able to make GAA really work at 2nm, while INTC and TSMC are still early on the learning curve and have to play catch-up. Any early indications that Samsung's gamble might pay off?
 
I don't have any inside information. But I guess Samsung need more 3nm GAA orders to practice and learn. Not sure Qualcomm will give them lots of orders to practice next few year or not while MTK will be using TSMC 3nm. Qualcomm will be risking it's market share betting big on Samsung.
 
Last edited:
rgrindley said:
GAA transition will be hard. INTC targeting 2024 on 20A and TSMC on 2nm on a similar timeline, I believe. Samsung decided to gamble and start early, to get a jump on the learning curve. I assume the bet is that while Samsung 3nm may be a write-off, they will work out the kinks and be able to make GAA really work at 2nm, while INTC and TSMC are still early on the learning curve and have to play catch-up. Any early indications that Samsung's gamble might pay off?
Click to expand...
The problem for Samsung is that their 3nm GAA process is more expensive, higher power and lower density and yield than TSMC 3nm FinFET -- and nobody buys on whether it's a new sexy technology, they buy on PPA and cost.

It's easy to say that they should "write off" 3nm GAA but this leaves a big hole in their income for at least a couple of years, while TSMC are raking it in. Samsung also always seem to have inferior density/yield, maybe because their raw technology came from the IBM consortium who were always more concerned with claiming the latest greatest technology in conference papers than getting one that actually delivered what the customers really wanted in mass production -- again maybe due to the IBM mainframe/PowerPC mindset that all that matters is performance, cost and yield don't matter when each CPU sells for tens of thousands of dollars.

Unfortunately for Samsung TSMC *are* generally very good at getting yielding processes into mass production with a regular cadence, even if they're not always the first to market with the latest new technology (e.g. GAA)...
 
Yesterday analyst asked Samsung about 5 nm bad yield. Samsung said it was seeing a stable increase in its yield rate. That means Samsung admits it's 5 nm yield is bad.
 
Last edited:
IanD said:
The problem for Samsung is that their 3nm GAA process is more expensive, higher power and lower density and yield than TSMC 3nm FinFET -- and nobody buys on whether it's a new sexy technology, they buy on PPA and cost.

It's easy to say that they should "write off" 3nm GAA but this leaves a big hole in their income for at least a couple of years, while TSMC are raking it in. Samsung also always seem to have inferior density/yield, maybe because their raw technology came from the IBM consortium who were always more concerned with claiming the latest greatest technology in conference papers than getting one that actually delivered what the customers really wanted in mass production -- again maybe due to the IBM mainframe/PowerPC mindset that all that matters is performance, cost and yield don't matter when each CPU sells for tens of thousands of dollars.

Unfortunately for Samsung TSMC *are* generally very good at getting yielding processes into mass production with a regular cadence, even if they're not always the first to market with the latest new technology (e.g. GAA)...
Click to expand...

I whole heartedly agree. Tangentially, as I stated on another thread:

Given Samsung is technologically ahead of Intel, and given Samsung has admitted to falsifying yields on their 4nm process (which very likely means they have lied about all their latest nodes), Intel probably won’t catch Samsung for a few years, and will likely never catch TSMC, despite the yarn Pat Gelsinger so loosely spins.
 
You must log in or register to reply here.
Back
Top