Samsung Discloses Backside Power Delivery For Next-Gen Chips

[del42sa]

https://wccftech.com/samsung-disclo...y-next-gen-chips-joining-the-race-with-intel/

The new BSPDN method hasn't been adopted by foundries yet, and Samsung is the first one to disclose the results of the innovation method. According to the Korean giant, they reduced the area by 14.8%, compared to the traditional method. Reduction in area results in the company having more room to add more "goodies" into a die, such as a transistor, leading to an overall increment in performance.

 

[pcchen]

It looks like Intel will be the first to market with 20A in 2024, followed by TSMC with its N2P process in 2026, if Samsung does not get something working before that.

 

[Seanspeed]

It looks like Intel will be the first to market with 20A in 2024

Yea we'll see. Intel is still pushing this timeline, but I still find it a tad difficult to believe Intel is going to jump from Intel 7 in 2022, to Intel 4 in 2023, to 20A GAA w/BSPD in 2024. These are all major node leaps with the latter one incorporating some serious manufacturing shifts.

Dont get me wrong, I'm not an Intel hater, and I've long been telling people to not write them off when many others were, but I dont think they deserve a lot of benefit of the doubt in terms of their ability to execute in such an incredibly timely manner. I'm half expecting Arrow Lake to get delayed to 2025, while we get some sort of Meteor Lake-S part next year instead. Either that, or Intel makes a panic switch to TSMC 3nm for Arrow Lake if they insist on releasing it next year.

 

[arandomguy]

Intel 7 itself though wasn't a major node leap. It's an iteration of Intel 10nm combined with a rebranding shift. Intel 4 is the node that has experienced the acknowledged delay (I think 2 quarters?) which is why it looks more compressed between it and Intel 20A. I also think Intel 20A's timeline is "late" 2024.

Otherwise if you frame it another way it's 3 major nodes from 2018-2020 (depends on what you consider the start of Intel 10nm as) through to 2024 which no longer seems as aggressive.

 

[Seanspeed]

Intel 7 itself though wasn't a major node leap. It's an iteration of Intel 10nm combined with a rebranding shift. Intel 4 is the node that has experienced the acknowledged delay (I think 2 quarters?) which is why it looks more compressed between it and Intel 20A. I also think Intel 20A's timeline is "late" 2024.

Otherwise if you frame it another way it's 3 major nodes from 2018-2020 (depends on what you consider the start of Intel 10nm as) through to 2024 which no longer seems as aggressive.

I didn't say Intel 7 was a major leap from 10nm. Obviously they're the same thing. I'm saying the leap from Intel 7 to Intel 4 is a major leap, and then a year later they're supposed to do another major leap from Intel 4 to 20A with all-new cutting edge lithography features? That's pretty insane, even accounting for a 6 month Intel 4 delay. Coming from Intel, it just seems a little too good to be true, ya know? I expected them to get their act together in time, not quantum leap back to the front in a few short years.

Another option is that 20A will be a somewhat rushed node, not delivering great characteristics in its initial generation. Samsung similarly has that problem, where they've managed to hang on in terms of density competitiveness, but in their desperation to keep up, have struggled to offer great performance and efficiency gains. I'm rooting for both companies, as we very much need a more competitive fab industry, but I really want to see real world results before I get too excited. Samsung in particular are very known for exaggerating the state of their situation.