Well, pretty much every big-name handheld PC is running 16GB LPDDR5, but Switch 2 is looking to be 12GB of LPDDR5X. 16GB would be nice, but likely not needed in the long run. Having that extra bandwidth, however, is more important, and a good trade off.
In any case, I honestly believe the T239 will be TSMC 4N, even if the Samsung 8N is cheaper, because Nintendo thinks about the long term. With Switch, they were already in the process of making a weaker chip before getting convinced by Nvidia to use the Tegra X1. When they made the "switch", they were more or less locked into the 20nm process for the time being. When the 16nm Tegra X1+ became available, they immediately "switched" over to it, dropping the older process node. Here with the T239, they have the option from the start because they are in charge of the specs for Nvidia to design. What point would they have to go with an older process node to end up needing to use a newer node down the road (which costs money?) when they could use the newer node now? We all thought Nintendo would use LPDDR5, but they didn't. They 1-uped on it. To utilize what LPDDR5X offers over LPDDR5 means having hardware that pushes more, generating more heat, consuming more power. Using Samsung 8N would be counterintuitive, and simple a slap in the face.
YES YES YES
Nintendo isn't stupid. I'm sure they're scared out of their boots about repeating a Wii U situation. They were in the worst situation they've ever been in with that console and immediately hit the golden goose with the Switch. They're now in literally the wealthiest situation of their entire company history, but memories of the Wii U are still fresh. They've had the past nine years to think about where they want to go next.
The Switch was put together from whatever Nvidia had off-the-shelf because they were desperate. It's a night-and-day situation with the Switch 2. The SoC is custom. Everything has undoubtedly been analyzed from head-to-toe to keep this momentum going. They aren't going to suddenly be reckless and suddenly starting selling consoles at a loss or re-join the console wars that's tearing Playstation and Xbox apart. All they need to do is be as financially savvy as they've been over the past few years and they'll be totally fine.
As you say: when the Switch 1 first launched, they were on the 20nm node. But they had to quickly transition to something better. So while they do want to be cost-effective with their spec choices, there's a point where it's not financially prudent to cheapen out. Not if it's going to hurt the console and they're going to end up having to shift gears anyway. At the time though, it was a completely different story and Nintendo just needed to get something out there to keep themselves afloat.
File Decompression and super fast I/O like the other 9th gen consoles for super fast load times, the first console with built-in tensor cores for the best AI upscaling and Ray Tracing on the market, a powerful modern GPU frankensteined from both Ampere and Ada Lovelace, and now just confirmed, 12gigs of ram using the
newer LPDDR5X...
Nothing about the Switch 2 is "cheap" so far. Cost-effective and not selling at a loss so families can buy it for Xmas so their kids can play Mario Kart? yes, but not cheap for the sake of being cheap. If anything, it hits a lot of sweet spots to be a competitive and worthy choice for the 9th gen of gaming, especially for a handheld device. There has clearly been a lot of careful thought put into this. The only part to me that's cheap is the LCD screen which I can completely understand. But the OLED model and Lite are proof that there will be several SKU's to keep the train going. Nintendo is playing the long-game and not putting all their eggs into one premium expensive package that they have to sell at a loss. They know what they're doing...
Which brings me to the point: why would Nintendo purposefully put themselves through the same situation with the node? why invest so much hard-earned revenue to create a totally bespoke SoC for a new Switch 2 hybrid console just so they can get an older inefficient node that they'll have to replace anyway? Yes, you can argue that Nintendo often makes miracles with "withered technology", but everything we've seen thus far from the Switch 2 is hardly withered at all. If anything, what they're doing is practically the best that a cost-effective early/mid 2020's handheld
could be. I find it hard to believe that they'd purposefully cheap out on an essential like the node when it'll cost them more in the long-run anyway as opposed to a smaller 5nm node like the TMSC 4N that'll last longer, get them more battery life/performance, and it's not like they couldn't afford. With how wildly successful the Switch was, a company would be out of their mind to not cut Nintendo a deal on a higher-quality node so that everyone reaps the benefits.
A lot of rambling, but TLDR: it doesn't make sense to me that Nintendo, Nvidia, and all their partners would put so much time, thought, and money behind this new console only to cheap out on a component they'd almost certainly have to replace in the long-run anyway.