The Nintendo Switch 2 relies on a Custom APU from Nvidia – but what does it look like?
The presentation of the Nintendo Switch 2 was almost a week ago and, in addition to some, sometimes surprising game announcements, also brought a somewhat deeper insight into the technology – with one exception.
Because apart from the information on the official websiteNintendo does not call any concrete special chip that a “adapted processor from Nvidia” is in Switch 2.
The tech experts from Digital Foundry, based on various analyzes (e.g. from the Famiboard Forum) and some information from Linux documentaries for the NVIDIA chip, go into more detail on the processor of the Nintendo Switch 2.
Link to YouTube content
Nintendo Switch 2: The expected Tegra T239 as a drive
In the Nintendo Switch 2, the Nvidia Tegra T239 (Codename: Orin), which has been made several times in Leaks, is a modification of the Tegra T234 known from the predecessor. The 8-NM production of Samsung is used here.
The CPU cores are ARM Cortex-A78C-eight pieces should be in there. The clock rates of the CPU are almost the only detail about which it can still be speculated. Digital Foundry assumes the following values:
- 998 MHz im Docked-Modus
- 1.101 MHz In handheld mode
27:48
Slowed Switch 2: Price, Games, Release Date & First impression!
The SOC continues to focus on a graphics unit from the ampere generation, which with 12 compute units of 128 shader units each comes to a total of 1,536 cuda cores. Here, too, only the clock rates can be speculated:
- ~1.000 MHz im Docked-Modus
- 561 MHz In handheld mode
In principle, the Nintendo Switch 2 should roughly reach the performance of an RTX 3050 with these specifications and to computing up to 3.1 TFlops. Compared to the predecessor, the data sheet of the SoCs is now as follows:
| Nintendo Switch (2019) | Nintendo Switch 2 | |
|---|---|---|
| System on a Chip | NVIDIA Tegra T210 (»Ends«) | Nvidia Tegra T239 (»Odin«) |
| CPU specifications | 4x ARM Cortex-A57 4x ARM Cortex-A53 (deactivated) |
8x ARM Cortex-A78C |
| CPU acts | 1,02 GHz | 998 MHz Docked 1.101 MHz Handheld |
| GPU specifications | Maxwell GM21B 16 Nm TSMC production 2 Compute Units 256 Cuda core |
Ampere 8 nm production 12 Compute Units 1,536 cuda core |
| GPU acts | Mobil: 307 MHz Docked: 768 MHz |
Mobil: 561 MHz Docked: ~1 GHz |
To what extent the theoretically clear upgrade will also be expressed in games is not yet finally determined by independent benchmarks. The digital foundry, however, is positive about the potential of Switch 2, which is also due to the conceivable DLSS capacities.
- Because as a GPU based on Ampere, Switch 2 also has the necessary tensor core to be able to perform the AI-based upcaling. In theory, even DLSS 4 is possible, as not least the desktop GPUs from Ampere prove.
- However, the sticking point is whether the use of DLSS is worthwhile or that the upscaling is not “too expensive” for the Switch-2 GPU with regard to computing power.
