Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/samsung-all-in-on-3nm-liklihood-of-overtaking-tsmc.13275/
        )

    [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 all-in on 3nm: liklihood of overtaking TSMC?

TxAggie1995

New member
I work in Serdes verification with a fabless company; close enough to know about VLSI design and the foundry process, but far enough away to not be plugged in to all the news. Been reading the news on semiconductor processes just out of personal interest and that's how I found this board.

So TSMC's obviously destroying the market, but it sounds like Samsung is going all in and sinking billions into 3nm GAAFets to try and catch up. If they can succeed in 2 years, they're probably right back in it with TSMC. If they don't, TSMC is alone on top of the mountain, and that has just huge ramifications for the planet effectively being single-sourced for new CPUs/SoCs. So there's a lot riding on it; so far, how does Samsung look to be doing?
 
I will write more about this later but today Samsung's biggest foundry market is that they are NOT TSMC. Similar to what AMD was to Intel back in the day. Nobody wants a monopoly especially in semiconductor manufacturing so cost sensitive companies like QCOM and Nvidia will straddle TSMC and Samsung. ASIC companies are also good Samsung customers due to margin constraints (lower wafer costs). The one ingredient that Samsung is missing in the foundry recipe for big leading edge customers is trust.

Samsung Intensifies Chip Wars With Bet It Can Catch TSMC by 2022

Samsung Takes Another Step in $116 Billion Plan to Take on TSMC

Bottom line: Can Samsung Foundry Really Compete with TSMC?
Sorry, not today, not at 3nm. The TSMC 3nm PDK is already in use at the top semiconductor companies around the world and have the full support of the ecosystem. The Samsung 3nm PDK on the other hand is still evolving as are the tools and IP that will support it. Just my observation, experience, and opinion of course.
 
Given that Apple's M1 chip has been universally praised by the media, I'm wondering if there will be a industry trend of embedding the DRAM into the SoC? Are the costs of adding DRAM now justified or is this an example of something only Apple could do? More broadly, could this be the start of TSMC dipping it's toes in the water and ultimately taking some of Samsung's memory business away?

I readily acknowledge I have NO technical knowledge with semiconductors...only a strong curiosity.
 
Given that Apple's M1 chip has been universally praised by the media, I'm wondering if there will be a industry trend of embedding the DRAM into the SoC? Are the costs of adding DRAM now justified or is this an example of something only Apple could do? More broadly, could this be the start of TSMC dipping it's toes in the water and ultimately taking some of Samsung's memory business away?

I readily acknowledge I have NO technical knowledge with semiconductors...only a strong curiosity.
TSMC has done eDRAM before as had Intel. Afterwards, no hot pursuit. One reason: consideration of MRAM, although that may not pan out either as SRAM extension. Regular standalone DRAM from the usual vendors in the same package may still be a better deal.
 
Given that Apple's M1 chip has been universally praised by the media, I'm wondering if there will be a industry trend of embedding the DRAM into the SoC? Are the costs of adding DRAM now justified or is this an example of something only Apple could do? More broadly, could this be the start of TSMC dipping it's toes in the water and ultimately taking some of Samsung's memory business away?

I readily acknowledge I have NO technical knowledge with semiconductors...only a strong curiosity.
The M1 has two standard Samsung LPDDR4 dies alongside the 5nm CPU die in the same package. This is not eDRAM, it's co-packaged DRAM -- like HBM but bigger, cheaper and slower.

The plus side is (potentially) higher speed and lower power because it's right next to the CPU instead of in a separate DRAM module, but this needs customized DRAM (with a different interface) which Apple are not currently using.
 
As Daniel said, the industry wants Samsung to succeed as a viable second source to TSMC. That said, while I expect Samsung to be competitive, I don't think they will surpass TSMC. TSMC is unique in it's culture, innovation, and position in the ecosystem. Samsung is a company that is very good at copying the successes of other companies, but I've never seen any real innovation from them. The type of company culture that creates a TSMC is different than what creates a Samsung, and that can't be bought with any amount of money. It's the legacy of Morris Chang.
 
Back
Top