The Nintendo Switch 2 has finally been formally announced, alongside a number of unexpected surprises despite the extensive leaks over the last few months. With a number of launch titles confirmed (including the exorbitantly priced $90 physical edition of Mario Kart World), a launch price of $450, and an all-new dock with 4K output support, it's the end of the road for a lot of speculation about the successor to the wildly successful Nintendo Switch... or is it?

The thing is, there's not a lot we know about the hardware of the Nintendo Switch 2, aside from the fact that it's a "custom processor made by Nvidia". The original switch used a "NVIDIA customized Tegra processor", which was the Tegra X1, but this time around, the processor itself is said to be custom​​​. As it turns out, leaks regarding the upcoming Nintendo Switch 2 spell a significant performance improvement, and these leaks have been further corroborated as time has passed. The Nintendo Switch 2's biggest upgrade over its predecessor is almost certainly its processing power.

Contextualizing the Nintendo Switch 2's hardware against its predecessor

The Nintendo Switch was amazingly underpowered

Before getting into the details of the Nintendo Switch 2's hardware, we'll contextualize it against the original Switch so that you can have a better understanding of how big these upgrades are. The original Nintendo Switch packed a Tegra X1 manufactured on a 20nm process before being switched to a 16nm process in later revisions, with a quad-core configuration made up of Cortex-A57 cores and a maximum clock speed of 1.8GHz when docked. Four Cortex-A53 cores are also present on the Tegra X1, but these are disabled on the Nintendo Switch. It also has a Maxwell-based Nvidia GPU, the GM20B, with 256 cores and a maximum clock speed of 768MHz. It has 4GB of LPDDR4 RAM with a maximum frequency of 1600MHz, and finally, it has 32GB of eMMC storage. The OLED model has 64GB of eMMC storage.

This is an incredibly underpowered system in so many ways, and we'll start by looking at the CPU. The clock speed doesn't matter anywhere near as much as the actual cores that were used. The Cortex-A57 was considered a high-end CPU core at the time, forming the main cluster on the Qualcomm Snapdragon 810 that powered smartphones like the HTC One M9, OnePlus 2, and the Microsoft Lumia 950 XL. With that said, being considered a "high-end" CPU core in 2015 doesn't necessarily mean much in 2025, especially in the mobile space, where there have been yearly advancements that have seen smartphones become capable of so much more than they were back then.

Things break down even further when we look at the other components, which have all aged considerably. The GM20B was considered powerful when the Tegra X1 launched and has arguably been the component that has made the Nintendo Switch capable of running "impossible ports" like DOOM (2016). However, it's based on Nvidia's Maxwell architecture, which debuted with the Nvidia GeForce 700 series in early 2014. The RAM, clocked at 1600 MHz by default, is also a significantly limiting factor, and even mild overclocks have seen considerable performance boosts in games like The Legend of Zelda: Tears of the Kingdom. Modern Vintage Gamer states that he believes memory bandwidth has been one of the most limiting factors when it comes to the Nintendo Switch's performance. Finally, the eMMC storage has a maximum theoretical read speed of 250 MB/s.

The smaller RAM size (which also serves as VRAM) and reduced memory bandwidth of 25.6GB/s contribute to a vicious cycle of poor performance, as streaming assets from the cartridge to RAM for rendering will be slower, particularly as those assets are first fetched by the CPU and decompressed before being sent to RAM. The hardware here was not particularly impressive at the time, and nearly a decade later, it's amazing to look back at the incredible library of games that the Switch managed to cultivate. We eventually did see signs towards the end of a genuine struggle, with many of the Resident Evil games and Marvel's Guardians of the Galaxy being cloud-based on the Switch. The Switch 2, though, looks likely to pack a massive generational uplift.

The Nintendo Switch 2's custom Nvidia processor is likely to be the Tegra T239

Reputable leakers have all been reporting the same thing, and it's looking increasingly likely

Nintendo's Switch 2 supports 4K 60 FPS when docked and supports 1080p 120 FPS in handheld mode, which is already a sign that the company is confident that there will be games that can actually make use of that hardware. Despite the fact that those features are for "compatible" games, the company has already shown Metroid Prime 4 running in both 4K 60 FPS with HDR enabled in "Quality Mode" and 1080p 120 FPS in "Performance Mode" in a docked Switch 2. There are a few things that we can take away from this, particularly when it comes to assumptions about the hardware. However, we'll look to what the leaks have said first.

Back in November 2023, Digital Foundry first reported that Nintendo appeared to be testing the Nvidia T239 processor and that it wouldn't really make sense for the chip to be developed for anything other than a game console. It appears to be a cut-down version of Nvidia's T234, a chip designed for the automotive and robotic market. Nvidia has its own Linux distribution, which details information about processors, and the differences between the T234 and the T239 that were found in this Linux distribution provide an apparent "smoking gun" that really hammers home the point that this is the chip being used.

First of all, it was confirmed that Nvidia is testing a new graphics API dubbed NVN2, where NVN is the name of the graphics API used exclusively for the Nintendo Switch. What's more, NVN2's internal testing shows that Nvidia has been using it with an emulated T239, making it a very strong possibility that this is the SoC powering the next Nintendo Switch. Finally, it introduces a File Decompression Engine, or FDE. Like the PlayStation 5, it can facilitate ultra-fast loading by reading assets directly into the GPU's addressable space and decompressing them on the fly.

With that evidence out of the way, the Nvidia T239 looks to be a monster of a chip compared to the Tegra X1. It packs eight Cortex-A78C Arm cores, an Ampere-based GPU with some Ada Lovelace features backported and 1,536 CUDA cores, and a significantly larger memory bandwidth than the original Switch. According to a user on Famiboards (the origin of many of these leaks), the CPU is supposedly clocked at 1100.8 MHz when in handheld mode and 998.4 MHz when docked, and the GPU is supposedly clocked at 561 MHz when in handheld mode and 1007.25 MHz when docked. Then, we get on to two massive supposed upgrades: the Switch 2's memory (said to be 12GB of LPDDR5X RAM) will apparently be clocked at 4266 MHz in handheld mode and 6400 MHz when docked, boasting 68.256 GB/s and 102.4 GB/s bandwidth speeds, respectively, and UFS 3.1 storage instead of eMMC storage, with a theoretical read speed of up to 2,100 MB/s.

Some have suggested that those CPU speeds might be swapped, though the leaker has doubled down on their assertion that the CPU speed decreases when docked. Digital Foundry theorized that perhaps some of those cores are disabled in handheld mode, but the numbers overall seem plausible. These are very powerful cores, especially when compared to what the original Nintendo Switch packed, so it's not out of the question that Nintendo would disable some of these in order to eke out better battery life when in handheld mode. What lends further credibility to these leaks is that many of these leaks, taken from the Nintendo Switch SDK and tracking shipment orders to factories, all line up with Digital Foundry's original assertion that the chip inside is the Nvidia T239.

Bringing it back around to the Metroid Prime 4 demo that was shown during Nintendo's Direct, Nintendo may be facilitating Nvidia's DLSS technology to achieve 4K output. The T239 is certainly capable of it, thanks to its Ampere architecture, and Nintendo mentioned that 4K output can be "upscaled" to match a compatible TV. To be clear, this isn't to say that games are being output as 4K with upscaling, and this could refer to the actual output and system UI. Nintendo has been extremely vague when it comes to this particular point of Switch 2, and the company has very little information on its own specifications page.

There is some evidence that we may see upscaling in use, though. Digital Foundry tested the RTX 2050 with 4GB of vRAM in order to achieve the "closest approximation" of the T239 to get a "ballpark idea" of what it's capable of. What's especially interesting is that it was capable of running Cyberpunk 2077 at a 720p resolution at 30 FPS, with DLSS performance mode netting some slight improvements. While there were some hitches, it was certainly playable, along with a number of other titles, too. Given that Cyberpunk 2077 was announced for the Switch 2, it's possible that this title may utilize DLSS in order to improve performance from the get-go, though games can also benefit from being built for a specific console where no other hardware needs to be taken into account, such as in the case of DOOM (2016).

Finally, Nintendo updated a patent referencing AI-powered upscaling a few weeks ago. Given that other patents that were also recently updated, like turning the Joy-Cons into a mouse, turned out to be accurate, this is another sign that we may see upscaling in use on the Nintendo Switch to achieve 4K output in some games.

The Nintendo Switch 2 is, comparatively, a performance beast

The original Switch has been well and truly outclassed

These performance improvements are absolutely massive, and it looks incredibly likely that this particular chipset is the one at the center of the Nintendo Switch 2. There's a massive amount of corroborating evidence, and the same leaker who predicted other hardware capabilities of the Switch 2 also predicted the 120Hz VRR HDR display. Specifically, this person leaked a docked TFLOPS figure before the GPU clock speeds leaked, and those GPU clock speeds roughly match the TFLOPS figure given. There's a lot of corroborating going on between leakers, and nobody appears to be contradicting each other when it comes to performance capabilities and hardware.

Bobby Kotick, former CEO of Activision Blizzard, placed the Nintendo Switch 2 roughly in line with PlayStation 4 level performance in the Microsoft FTC court case. However, it may be the case that he was referring to that level of performance on a specification level. Nintendo is said to have shown a demo to developers with The Matrix Awakens running on the Switch 2 at Gamescom in 2023, and that game was used as a technical demo to demonstrate how powerful the PlayStation 5 and the Xbox Series X were back in 2021. On the Switch 2 it was said to have used Nvidia's DLSS technology, along with ray tracing for realistic lighting.

There's a lot to be excited about when it comes to the Nintendo Switch 2, and I wouldn't be surprised to see a massive performance improvement across the board when it comes to the games that we see released for it. With buy-in from companies like CD Projekt Red with Cyberpunk 2077, it's clear that developers are a lot more confident in putting massive games on the successor to Nintendo's original Switch, which lends credence to the assertion that big improvements are on the way.