Uh, you're utterly daft if you think that is true. Samsung has some of the best yields in the business. But, hey, plenty of people in this forum seem to live in fantasy worlds where everything is the opposite of reality. Seem you're one of them.
TSMC's issue isn't yields, its "being booked". Theyre already booked 18 months out, Samsung is not (because they dont always do a lot of chip fab for other companies, mostly just for themselves) and Samsung can bump people, if it needs to. If it pisses of a smaller company, they simply dont care. Samsung doesn't rely on outside business in order for its chip fabs to be profitable. So if nVidia comes along and is like "here's a giant bonus, put us at the top of the queue" - Samsung will do it.
TSMC cant get away with that. They are entirely based around producing other people's designs. They piss off a customer and thats lost money.
- - - Updated - - -
More of the reverse reality world.
AMD has NEVER set a date for Ryzen 4. The last set of leaked slides showed it as "Late 2020", and the most recent leaks (from inside TSMC) have suggested Q1 2021 is more realistic. If it launches in "Late 2020" itll be a vapor launch with no stock available. Dont expect "good" availability until spring 2021.
Jay just did a video on how AMD takes literally FOREVER to pivot or recover from a misstep.
They put out the Zen 2 laptop chips WAY too late. They needed to launch in Jan or Feb. Instead, they waited so damn long, that they got a whole 5 weeks of impressive performance vis-a-vis Intel's laptop chips...
And then Tiger Lake launched at basically ate their lunch and their dinner while it was at it. 4 core/8 thread parts beating full 8 core parts by a wide margin. Xe GPU performance twice as good as any of the laptop APUs.
Because they take too damn long (in part, because if their reliance on TSMC, to whom they are just one more customer) to react and pivot.
- - - Updated - - -
Lolwut? 50% of the PC community is going to run out and spend 700$ on a GPU? Are you insane?
The money making GPUs are always the low-midrange SKUs. For every 2080/SUPER and 2080Ti that got sold, 20 1660 variants got sold.
The average gamer doesn't have 700$ to throw down on JUST a GPU.
The "average gamer" is playing on a rig that cost 700-800$ for the ENTIRE RIG.
- - - Updated - - -
It will mostly help with initial loading times and allow them to more quickly stream data for nearby assets much faster, but it wont magically reduce CPU dependency for most things (like AI, issuing draw calls, etc) and it wont even help much other than the initial load for games that aren't streaming entire worlds from storage like, for instance, Doom Eternal. It loads the entire level into memory in one go. This will make that load faster, but it wont really affect gameplay.