For AI, not gaming, which again, was my point.
Tegra Thor is an SoC designed for autonomous cars, as per your Wikipedia link. If you look at it's Neoverse CPU cores, they're not even the right ARM architecture for gaming. Why would Microsoft want that for a games console?
I'm sure Nvidia could use a 2560-core Blackwell GPU in a console if they wanted to, but it's the CPU side of the equation where Nvidia are a long way behind. Also, I'm not sure an RTX5050-tier GPU is really next-gen material - it would be slower than a Series X, and arguably only of use for a next-Gen Xbox if they split the tiers again with a very low-end Series S replacement.
Speaking of learning to read which you implied several times was a thing I should do, you may notice, if you check again what was written -
It's not like Nvidia give a shit about gaming any more.
The Switch 2's Nvidia CPU is unimpressive as hell. Qualcomm put out better offerings in midrange phones, at half the power draw, too. Don't get me wrong, the Switch 2 will be incredibly successful but that's despite Nvidia, not because of them.
I'm not versed on why Nintendo chose another Nvidia design, but presumably a big part of the decision was compatibility with the Switch 1's architecture to allow Switch 2 to play Switch 1 games. Without that requirement, just about anything else would be cheaper and better.
Interesting that you're pretending this is by design, and not due to the simple fact it's an ancient chip. Wow, shocker, a six year old architecture on a similarly old process node doesn't perform well against contemporary designs, what a twist.
Please learn to read. My post started with "because Nvidia aren't playing"
Switch 2's Tegra is the best Nvidia have to offer, it's relevant to my point, but it's not my entire point.
Reading comprehension, that's your job at TPU, isn't it?
For AI, not gaming, which again, was my point.
Tegra Thor is an SoC designed for autonomous cars, as per your Wikipedia link. If you look at it's Neoverse CPU cores, they're not even the right ARM architecture for gaming. Why would Microsoft want that for a games console?
I'm sure Nvidia could use a 2560-core Blackwell GPU in a console if they wanted to, but it's the CPU side of the equation where Nvidia are a long way behind. Also, I'm not sure an RTX5050-tier GPU is really next-gen material - it would be slower than a Series X, and arguably only of use for a next-Gen Xbox if they split the tiers again with a very low-end Series S replacement.
You should tell NVIDIA this, they don't seem to differentiate, and it seems to be going fairly well for them, considering their largest competitor with 1/10th their marketshare finally got around to copying them a few years late, as is their habit, to change the much lauded and marketed opensource runs on anything FSR 1-3 to FSR 4, which uses AI and dedicated hardware similar to Tensor cores, and runs on only their latest generation onwards (similar to RTX launch vs GTX, seven years ago), because of course that's the only way to actually compete. Also, you pretending there's a signficant difference between a drive platform and a gaming platform, when the very SoC used in the previous Switch was literally used for both, is misleading, whether intentionally or otherwise.
Looking at the ARM CPU in the Thor SoC it's simply a V9 architecture, with some high bandwidth interconnects, paired with a Blackwell GPU. A slightly tweaked version (such as the X series chips Nintendo has been using) would be excellent as a gaming SoC.
If you look closely, you may notice that the majority of X1 applications were not "gaming", but for phones, dev boards, shield TV upscaling,
drive applications (wow, seems like SoCs designed to be put into cars are also great for gaming, and the other way around, yay CUDA) etc. You pretending that a SoC can only be used for one thing is disingenuous.
"Switch 2's Tegra is the best Nvidia have to offer"
False. We've been over this, and you stating that a Drive chip would be unsuitable purely because "it's designed for autonomous cars" is hilarious, since we both know the Tegra used for the Switch 1 was also used as a Drive platform, and being good at one thing means it's good at the other. Perks of CUDA baby. Again, nothing stopping Nintendo, besides profit margins, to request a semi custom variant of an off the shelf solution, like they've done before. Nope, purely a financial decision to go with ancient architectures on an ancient node on the platform (mobile) that would literally benefit most from more efficient newer tech.
"the Switch 2 will be incredibly successful but that's despite Nvidia, not because of them."
False, there was literally nothing stopping Nintendo from using, modifying, or requesting a newer platform, but, as anyone who can read and analyse history will know, Nintento has a very consistent track record of using massively outdated hardware in their consoles, for cost reasons. Again, this is you trying to conflate Nintendo's
choice of using a very old chip, with NVIDIA's apparent inability to make something better (which they can and have). The Switch 1 could barely maintain 30 FPS. The Switch 2 has DLSS as a crutch but it's still a two gen old GPU architecture, and something like a 5 gen old CPU architecture, on a process node that was obsolete and compared badly against competing nodes on release in 2020. Lets not play the game of pretend, these are standard ARM architecture cores (the same as used in phones, fridges, cars, etc) with some connectivity tacked on, paired with standard CUDA cores, but simply very old versions of them. There would have been zero issues if newer generations were used, the only thing that would change are performance targets developers could aim for, and Nintendo's profit margin. You can see this in the rest of the component choices, small battery based on old chemistry, non OLED screen with ghosting issues, non hall effect/TMR sticks, etc. etc.
"it would be slower than a Series X, and arguably only of use for a next-Gen Xbox if they split the tiers again with a very low-end Series S replacement."
Even if, IF, MS didn't do another custom design like the Steam Deck APU which was reused many times for other devices, one of MS mandates to developers was that every single game released on the Xbox platform had to run on the Series S. This was one of the reasons why the Xbox platform has been unpopular to develop for, but also means that the equivalent PC GPU to the Series S would be something like a 5500XT or 1650 Super, raw horsepower that a theoretical RTX 5050 easily outperforms based on core count, without even taking into account DLSS. Again, something you are likely aware of, I'm sure. Beyond that, the Series X uses an RDNA2 GPU roughly equivalent to a 2070 S or a 6600XT, again, based on core counts, RTX 5050 would be a similar performance level to both of these to even before using AI magic and newer featuresets. But again, this is you talking about Nintendo, then switching to Xbox when it turns out you were wrong about Nintendo only having one option for the SoC. And again, this is assuming MS can't afford or wouldn't pay for a semi custom solution, which both NVIDIA and AMD have a long history of making, and MS has a long history of using in both their computers and consoles.
Its not like any console gamer even halfway cares or even knows there's an AMD chip inside of it now. They press ON and play game.
