Considering it will be extremely rare that a game, if any game at all, will need part of the 3.5GB as VRAM (and still, it's 336GB/s, faster than the 4K Xbox One, it's not like it's DDR4 or something. It's still usable as VRAM for lower priority uses), the XSX memory solution is much better than the PS5's but at the same time, it's more expensive. Not only that MS had to add another 64-bit worth of controllers on the APU, making it bigger, they also have to buy 10 chips instead of Sony's 8 chips. even though it's the same amount of RAM, MS is paying a bit more because they are buying more chips.
IMO that's the PS5's weak point, not the TF count. the PS5's GPU is more powerful than the 5700 XT and the 5700 XT has the same bandwidth, 448GB/s. At the same time, the PS5's 448GB/s will have to deal with 4K (while the 5700 XT is labeled by AMD as a 1440p card), RT (which is a big bandwidth hog and the 5700 XT doesn't have that either) and share it with a CPU. It probably won't be a big problem, but IMO this is the PS5's biggest weakness, not the TF count. In the end, it will probably just manifest itself in developers having a harder time hitting 4K, which I don't think is a big deal.
Like I've said before during the jolly Github days, I won't be surprised if the only difference between the PS5 and the XSX in 3rd party games will be 1800p VS 2160p and in reconstructed games 1800p reconstructed VS 2160p reconstructed. TBH, I can't tell the difference between 1800p and 2160p on my 4K 65" LG.
I only have one request
Transistor, make mine the worst. I mean really bad, like worse than
chris 1515.
I would argue that every PS5/XSX CPU core is more powerful than the whole of PS4's Jaguar CPU. They are similar in power, but a single core will outperform a multithreaded 8 core CPU with the same power because it has less overhead.