Arguments for TSMC 4NM:
- According to deep analysis from members of this community, the leaked specs points towards a chip that would be inefficient for an 8nm chip. An smaller chip would have been more cost effective.
- Some argue that an 8nm portable chip would not be capable of running the Matrix demo with "comparable" graphics to PS5.
- 8nm would be very old for a 2024 launch.
Am I missing something?
I would discard the second argument for 4NM: We do not know how the Matrix was running on the system, other than vague statements.
Respectfully, this gets the situation backwards. Lemme over-explain, as I so often do
If you're playing Amateur Chip Detective, figuring out the process node is critical, because it dictates the
size of the chip, and the
rough electrical efficiency. If you are willing to throw more money at a chip - basically willing to make it fucking huge - and willing to burn down a rain forest to power it, you can make any node do just about anything you want.
You could make an 8nm chip that ran the Matrix demo no problem. Nvidia made millions of them, they're called the RTX 30 series, and they are desktop GPUs that consume more power by themselves than the Switch CPU+GPU+Screen+Controllers consume, combined.
So you try to figure out the process node, because you can make a good guess on how much the company will charge, and from there, you can guess the performance. But we know the performance. We know the number of cores. So we know the thing that we usually are trying to figure out from process node.
If Nvidia and Nintendo have somehow managed to make a cost-effective, Switch-sized product with that performance on 8nm they will have performed a technical miracle that is frankly
way cooler than building on 8nm, It would be the fucking TARDIS of chip designs. An old blue box might not be as slick as a Constitution class starship, but I'll take a TARDIS over the Enterprise any day of the week.