Nintendo Switch 2 official specs confirm GPU similar to a mobile RTX 2050

Leaked hardware specifications of the upcoming Nintendo Switch 2 have been circulating for a while, but we now have official confirmation of the new console’s specs. Digital Foundry has claims it has officially confirmed that the Switch 2 will feature a custom Nvidia SoC with an Ampere-based GPU, 12GB of memory, and 256GB of storage.

The CPU portion of the custom SoC sports eight ARM Cortex A78C cores, featuring the ARMv8 64-bit instruction set and cryptography extensions. Feeding the cores are four cache layers: a 64KB L1 instruction cache, a 64KB L1 data cache, 256KB of L2 cache per core, and 4MB of L3 cache that is shared among all eight cores.

Digital Foundry clarified that six of the CPU cores will be available for games, while two will be strictly reserved for the operating system. This isn’t new; it’s a method used in the current Switch as well as in other modern consoles, such as the PS5. The CPU has an official maximum clock speed of 1.7 GHz; however, for the Switch 2, the developers have opted to run the CPU at a significantly lower speed: 1.1 GHz in “portable” mode and 0.998 GHz in “docked” mode. (That is not a typo; the CPU operates at a lower clock speed when docked, for unknown reasons.)

The GPU portion of the SoC is based on Nvidia’s two-generation-old Ampere architecture and sports 1,536 CUDA cores. The physical chip itself is rated for up to 1,400 MHz, but — similar to the CPU — Nintendo has opted to run the chip at significantly lower speeds on the Switch 2. In “portable” mode, the chip will operate at 561 MHz, and in “docked” mode, 1,007 MHz. The Switch 2’s Ampere GPU resembles Nvidia’s RTX 3050 and RTX 2050 mobile GPUs (the latter of which is also Ampere-based). Compared to these GPUs, the Switch 2’s counterpart has 512 fewer CUDA cores (2,048 vs 1,536) and runs at significantly lower clock speeds.

Adding Ampere to the Switch 2 will give the console real-time ray-tracing capabilities and DLSS support. Nintendo has not announced any games with real-time ray-tracing capabilities, but Digital Foundry and other outlets have demonstrated that it is possible on the RTX 2050 mobile. By contrast, several games have been confirmed to be coming with DLSS support, with Cyberpunk 2077 in particular offering variable real-time resolution scaling with DLSS (a feature of DLSS that is exceptionally rare to see, even on PC).

The memory subsystem is powered by two 6GB LPDDR5X memory modules (12GB total) operating on a 128-bit memory interface. The chips themselves operate at 2,133 MHz in “portable” mode and 3,200 MHz in “docked” mode. This translates to 68 GB/s of memory bandwidth in “portable” mode (2,133 MHz) and 102 GB/s of bandwidth in “docked” mode (3,200 MHz). 9GB of memory capacity is available for games to take advantage of, while the remaining 3GB is reserved strictly for the OS.