Soundwave said:
They don't really even have to use the X2 for extra power. They can merely use it for the die shrink and better battery (which could in turn allow a Switch model to run at full docked mode even when portable though). The X2 IS the 16nm variant of the X1. It just happens to also have a higher power mode to go with it. That doesn't radically make it that much more expensive. In fact at some point likely it's going to cost Nintendo more money to keep using the 20nm Tegra X1 because no one is going to be using the 20nm node at all. I also question exactly what R&D cost Nintendo really spent here ... this is a stock Tegra X1 ... a chip that was designed and finished two years ago by Nvidia. Nintendo didn't have anything to do with it other than slapping it into their box. The R&D costs for the Switch are likely largely invested into those Joycons, the chip is just a standard chip. This probably is from an R&D POV one of the cheapest systems Nintendo has ever made. |
These chips are getting ordered in very high numbers, the more the cheaper. That's one of the reasons why consoles get cheaper over time, even without a shrink the chips get cheaper. To switch (ha!) out the chip that soon would make it pretty expensive, not to mention that there are often multi-year contracts involved.
Using the X2 to save power would in theory work, but in practice it's actually not so sure. Every chip has it's sweet spot in terms of power consumption. Nintendo has clocked the Switch fairly low below the Tegra X1's original clock speed in handheld mode. Pascal on the other hand, is meant to run at higher clock speeds. So clocking so far down to mach the X1's performance could actually be counterproductive and the gains basically nil since if you clock below the sweet spot the gains in consumtion gets too much outstipped by the loss of calculating power, sometimes so much that it actually reverses the process entirely.
Also, matching the power exactly is very difficult to do when such changes are done. That's the reason why the Xbox ONE S is slightly faster than the original XBO, to make sure the architektonical changes don't drag some game's performances down when they are programmed too close to metal to make their fullest out of the new chip. In case of the switch, that could result in the X2 needing a slightly higher clock speed than the X1, making the consumption point mostly moot.
It's mostly a stock Tegra X1, with 2 small changes: In the stock Tegra, the GPU isn't supposed to switch it's speed between such states, so these needed to be implemented (I think a microcode update could do this already). The other thing is that the 4 low power cortex A53 cores got completly deactivated, probably by cutting all their electric contacts to avoid them consuming any power.
R&D for a console doesn't mostly go into the processor, it's actually one of the cheapest things in console R&D, even if it's a (semi-)custom chip. The reason for this is that the console developer just asks the chip maker to build a chip to their needs and rough specifications, and it's up to them to come with a design the console maker accepts. it's also one of the aforementioned reasons why the chips get cheaper over time even without a die shrink, the first batches have the development costs from the chipmaker in them. In the case of the switch, these should be low (and the Tegra ain't really expensive to begin with), so there won't be too much of a drop here
