I think Nvidia have just certified a T214 (Mariko) Jetson board with the Khronos group. If you take a look at Khronos's Vulkan conformant products page
here, you'll see a Nvidia Confidential device at position 353. This isn't that strange to begin with (you'll see quite a few Qualcomm and Intel undisclosed/confidential devices on the page), but the interesting thing is that it's running a new Tegra SoC. You can see it's ARMv8a and runs Linux for Tegra R32. My first thought here is it's an Orin board (the successor to Xavier which Nvidia showed in a roadmap last year), but running over to the Linux for Tegra source code there's an interesting little change from R31 to R32 in
this file: several references to T214. The same T214 we've seen referenced in Switch firmware.
The actual changes in L4T don't tell us much (Nvidia likely scrubbed all T214-related code from the public release, but missed these references in a couple of comments), but there are actually a few things I think we can deduce from this. Firstly, I don't think it's a coincidence that T214 references appeared in L4T in the same release that was used to certify this confidential Nvidia device; I think that device is using T214. Secondly, I think the fact that it's a L4T-based Khronos-certified device means that it's not a Switch Pro/Plus/Mini/etc. or a devkit, as Switch itself uses a different OS, and even if L4T was used in Switch devkits, there would be no reason to have them certified by Khronos, that's only intended for devices for public sale.
If you look at devices which are Khronos-certified and run L4T, they all fall in a single category: Jetson boards. Releasing a Jetson board for a new Tegra chip isn't exactly surprising, but what is interesting is that it means
T214/Mariko isn't a Nintendo-exclusive chip. It's being used by Nvidia in other devices, and possibly being sold to other manufacturers as well (as a large part of the reason Jetson boards exist is for hardware designers to use for prototyping/development of Tegra-powered hardware). This isn't really a change from the TX1 used in the original Switch, but it could well impact what kind of SoC the new Switch device(s) get. A Nintendo-exclusive chip would be designed 100% by and for Nintendo, they'd be paying for R&D and making every design decision. A new Tegra SoC which Nvidia will be shopping around to other manufacturers, though, could see Nvidia making more of the design decisions (and presumably paying for more or all of the R&D).
I still feel that, as Nintendo's going to be by far the largest customer of T214, it's been designed primarily around their needs, but Nvidia may make decisions Nintendo wouldn't have, in order to make it more appealing to other customers. A 7nm process becomes a little more likely (still unlikely, but I feel Nvidia would push for 7nm, where Nintendo would push for 12nm), as does a CPU setup capable of hitting higher peak performance levels (ie A76 cores instead of A73/A75). I also think Turing (and even Tensor cores) become more likely as Nvidia would want to sell their latest GPU tech. I don't think it massively changes the performance levels I'd expect from a new Switch, but it does perhaps give a bit of insight into the design process which led to the new SoC powering it.
