Nintendo Switch 2 motherboard leak reveals RAM capacity and ageing Nvidia SoC

[Redaktion]

A new leak shows off the Nintendo Switch 2's motherboard. On it, we can see an Nvidia chip manufactured on what is almost certainly Samsung's 8 nm node and some LPDDR5 RAM from SK Hynix.

https://www.notebookcheck.net/Nintendo-Switch-2-motherboard-leak-reveals-RAM-capacity-and-ageing-Nvidia-SoC.939624.0.html

[Mega64]

I am concerned about that person's fingers. They look weak and fragile. Please forward this message to the leaker and tell them they should take better care of their fingers. Thank you!

[GeorgeS]

While on the one hand the fanboi's and the press have been claiming outlandish performance improvements (almost non-stop) the article then has this line:

"Then again, Nintendo has clearly bowed out of the high-end console rat race and repeatedly stated the Switch 2 will be more of a side grade than a real upgrade."

So like WHEN EXACTLY has Nintendo "repeatedly stated" this? Huh? Inquiring minds wish to know.

I've read each and every Switch2 article posted on this site and this is the FIRST one which mentions that the Switch2 will be a 'side grade' vs upgrade.

While Nintendo staying out of the "performance rat race" was to be expected by anyone with even remotely realistic expectations I see that this site has degraded further to now claiming a Switch2 "release" rather than a mere product announcement (as was expected for some time).

As always take whatever you read on this site with a LARGE GRAIN of salt.

[Discostew]

There's no actual evidence that suggests it's made on Samsung 8nm. That's just an assumption because of the markings possibly pointing to Samsung while continuing to play off the original claim from the guy that got one thing correct and everything else wrong about the T239. That said, it may still be made by Samsung, just not on 8nm (as the die shown is still too small, even by 8nm standards). It could very well be made on Samsung 5nm that isn't too far off from TSMC 4N (Nvidia-optimized 5nm) at lower clocks. It's going to have additional tech in it like Lovelace's clock-gating and an FDE block, which we don't know how big that will be.

Honestly, having LPDDR5X capable of 120GB/s (from shipping data that points to websites also saying LPDDR5 but list speeds only possible with 5x) is still the biggest factor that supports it is not on 8nm, because it makes no sense otherwise. You don't throw that kind of RAM into a system that can't make use of it, especially when it's more expensive than using plain LPDDR5 modules.

[Ramsey]

This is in line with the recent NVidia release of the Jetson Orin Nano Super.

It's the same hardware. No BS of custom made 4nm SOC for Nintendo. It a repurposed SOC, although, thanks god, with extra RAM.

And that's fine. It no more than what Nintendo needs.

[Calan]

You mentioned the ram that it has but not the GPU specs. then tried to claim that the switch 2 will be able to play 4K games I'm sorry but system ram didn't how you determine how good a system can play a game the GPU is the main factor. The switch will more than like do 1080 at 60fps and checkerboard up to 1440 maybe. Ain't no way it's doing any 4k resolution you're out your mind

[Jameson]

You are such a pedantic loser.

[DemoWhat]

As someone mentioned, it's directly in line with the nano orin super dev platform that is set at a price point consistent with having economy of scale. Similarly, the article points out Ampere (RTX 3000 series) GPU architecture, a leap of three full generations from the Maxwell Tegra SOC used in the Switch. This means that it will have all the hardware bells and whistles for DLSS on top of the compounded performance improvements per GPU core between. Make no mistake, it'll be a massive jump in performance even if some of it is found in DLSS scaling wizardry. Possibly so much improvement that it'll compete well against the first generation Steam Deck in a smaller form factor.

Of course, it's worth taking that with a grain of salt because it's very likely that Nintendo will once again limit the chip in favor of a longer battery life and less heat production. This also depends on Nintendo actually giving game devs access to the additional features of Ampere GPU architecture.

[Doughboy]

The chip size isn't consistent with an 8nm Tegra Orin, which would be too big for the slot. It has to be the 5nm.

[opckieran]

Going from Maxwell based graphics to Ampere is a "side grade"? Please explain how. Especially when even the Switch OLED is estimated to be less than 1/6th the performance of T234, which itself is said to be modified for the Switch 2.

https://gadgetversus.com/graphics-card/nvidia-tegra-x1-plus-maxwell-gpu-specs/

[Kahn]

Extremely important with regard to the structure width of 8 nm for microchips! Many people criticize the fact that the Switch 2 "only" uses a chip with a structure width of 8 nm. What is disregarded: The specifications for the structure width of 8 nm or the more modern 4 nm, previously often used as the gate length for transistors, are currently only advertising statements regarding the transistors in the chip. The specifications such as 8nm or 4nm are intended to show that a chip "only" simulates the performance of 8nm or 4nm chips, i.e. that it delivers similar performance. According to chip specialists, the structure width of microchips has actually remained at around 20 nm. Chip manufacturers are currently unable to achieve much more than this size for reliably functioning microchips. At sizes smaller than the 20nm structure width, the transistors are said to become too fragile and, due to the small distances, electrical charges are said to jump between the transistors or their parts more often and more easily. Microchips with a structure width of 4nm are actually supposed to be packed more "densely" in terms of transistors, so these microchips with more densely packed transistors are also hotter than chips with a specified structure width of 8nm, where transistors are spaced further apart.

Conversely, this means that an 8nm chip runs cooler than a 4nm chip, if both have the same performance, because the transistors are further apart on an 8nm chip, or: because the transistors are built closer together on a 4nm chip! This also means that a chip with an alleged 8nm structure width is more robust than a chip with a 4nm structure width! - That's why I think it's a good idea for Nintendo to use a chip with an 8nm structure width in its handheld!