By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Bofferbrauer2 said:
sc94597 said:

On a 5nm manufacturing node @ 15W, which makes the most sense for reasons mentioned here, one should expect something on the order of performance of a GTX 1650 desktop, assuming (for good reason) Nintendo is sticking with Nvidia.

A GTX 1650 is decent bit (about 60%) more powerful than the PS4's GPU and more like the GPU in the PS4 Pro (of course without being bottlenecked by a crap CPU.)

In hand-held mode, roughly PS4 level makes sense though. 

The Switch 2 is also expected to have a lot more ram and a much, much better CPU than the PS4. 

This is the chip I was referencing too - but the performance results I found are quite a bit lower, only about half a 1050 or a third of a 1650's performance (You can find both of them when scrolling the chart). Keep in mind that this is the 8GB version, not the 16GB one, and it might have been running on a lower TDP, but that's still a good ways below a 1650 even when accounting for. I don't think that going from 8nm Samsung to 5nm TSMC alone has such a big impact to triple the performance.

Going from 8nm to 5nm improves performance by about 30% per watt. Even the "Drake" GPU that was rumored to be in the Switch 2 based on Ampere is between a GTX 1050ti (which is slightly above PS4 level) and a GTX 1650 in terms of performance. 

I don't know where you're getting "half a 1050" from. 

Edit: The Orin NX 8Gb you linked is not the same chip as the T239/"Drake chip". It has far fewer Compute Units (50% fewer CU's of the T239 rumor), is running at a 765 Mhz clock, and has fewer CPU cores (6 vs. 8.) It is designed for SBC's to do machine learning, not gaming. These SBC's don't have active cooling. The Switch 2 almost certainly will.

https://www.notebookcheck.net/NVIDIA-Tegra-T239-New-leak-offers-clues-about-rumours-next-generation-Nintendo-Switch-chipset.654330.0.html

Last edited by sc94597 - on 11 September 2023