How Will be Switch 2 Performance Wise?

[sc94597]

Pemalite said:

sc94597 said:

The original Switch was much lower clocked than the Nvidia shield, but that was because its chipset wasn't customly designed for it and Nintendo purchased it as a sort of "hand me down." 

This is false.

The original Switch was much lower clocked due to battery life and thermals.

Running games in a closed console environment has the expectation that the hardware would be 100% utilized more often than not as software is designed to strictly target the hardware.

In an open platform like a Shield or PC, there is an expectation that the CPU or GPU utilization may be lower than 100% as the software is targeting the OS/API's rather than the hardware.

Nintendo also needed a cheap quick-to-market chip and the Tegra X1 fit that bill perfectly, Nintendo *could* have offered 50% more performance by using the Pascal variant, but I guess financials got in the way.

Nintendo didn't buy the Tegra X1 in the Switch as "hand me downs" - The Tegra X1 in the Switch HAD to be manufactured from scratch, nVidia didn't have a warehouse of these chips sitting around doing nothing.

So you basically just agreed with what I said (they use the same chip, and that is why you can compare clock rates), and then said "this is false." The chipset in the original Switch was not designed specifically for the original Switch, unlike the chipset in the Switch 2. It was designed for a wide range of systems (including the Nvidia Shield) and then repurposed for/inherited by the Switch with very minor modifications meant to fit its form-factor (reduced clocks being one.) This is characteristically different from the T239, which was built with the Switch 2 in mind from the very start and with its form factor under consideration throughout the whole development process. There is no other hardware with the T239, unlike the T210.

If the hardware for the Switch were designed specifically for it, there wouldn't be a discussion of "under-clocking" because there is nothing else to compare it against to say "it is under-clocked." It is appropriately clocked for its target hardware that it was specifically designed for. We see this in that with the T239 there really isn't an analogous chip to compare against. The closest thing to it is the T234, which is very different hardware with a different targeted form factor and purpose (edge compute for sensors vs. gaming.) 

The one part where you disagreed is the bolded. The bolded ignores the fact that the major cost for Nvidia (as the designer) isn't the cost of producing the chips but the R&D of designing the chips. Nvidia needs to recoup those R&D costs. So yes, it is a "hand me down." They were selling essentially more of a product they needed to recoup the R&D cost on that they already did the bulk of the design work for (that is money already spent) with other devices being considered in that work, and therefore they gave Nintendo a deal, rather than fully designing hardware from a less developed stage (and accruing more R&D costs.) If they were specifically designing the hardware for Nintendo they wouldn't likely provide them with the T210, but something more custom, even if similar. Nowhere did I suggest there was manufactured hardware already produced. You pulled that out of the aether. Nvidia's (being fabless) role in this process is as chip-designer not chip-manufacturer. It's the mature extant design that is handed down, not extant physical chips. 

Last edited by sc94597 - on 03 May 2025