Switch 2 discussion around clocks is reassuringly similar to with Switch one, just with some different names expressing disbelief and disappointment.
Not using dynamic frequencies would be another mistake. Just give developers a power budget and let them optimize based on that.
I'm trying to think of what this might mean for price conscious handheld, and why Nintendo might not want it. Hmm.
If you give a power budget and let clocks fluctuate wildly, I expect this would mean you're far more often
at the power budget limit, and that would mean much worse battery life. So either a poor user experience or a more expensive battery. By giving clocks instead, you're normally well under the power budget, and probably only get close in a few areas of a game. So that would probably be a cost thing that Nintendo might want to avoid.
On top of this, while the max power output of the chip might be the same, the thermal output of the hotspots on the chip e.g. the CPU cores (like the ALUs and stuff) will increase in a none linear fashion. This is going to require a more expensive cooling system, as you have to dissipate more heat from a smaller proportion of the surface area of the chip (thought perhaps not so much for un-docked mode). So again that would probably be a cost thing that Nintendo would want to avoid.
Finally, implementing and testing a deterministic clock control system - the same amongst all units like AMD did for Sony - might require more development work on both Nvidia and Nintendo's side. And again that's cost.
Things would just be a lot simpler to set conservative predetermined clocks (or combinations of clocks), know you can handle them if you need to, and expect to normally be comfortably under them.
Finally (part 2), Nintendo are very conservative with hardware, and innovate on the software side. They know fixed clocks, and they've never not worked for them. Ruthlessly managing BOM and keeping R&D costs relatively low (so they can survive a WiiU like moment) are things they probably value more than squeezing max performance out of every mm^2 of silicon.
The most disappointing part for me is that whatever performances target they had in mind, they could have used a better fabrication process. And I can't see how TSMC 5nm could be much more expensive in 2025 (vs SEC8N). This would have given the console better battery life and make it run cooler. And no need for customers to buy a refresh console in two to three years.
Samsung 8nm is probably very cheap now. Nvidia got a good deal from Samsung way back when, and are perhaps the ones negotiating with Samsung for the chips as supplier to Nintendo. This may be part of an ongoing supply agreement with Samsung.
Nvidia will already have the design libraries they need to design what Nintendo want, and Nintendo aren't going to pay potentially 100s of millions of dollars to move the design over and validate it. The R&D costs for Switch 2 will be being kept as low as possible, risk of using extremely mature tech on an extremely mature process is low, and the cost of the chips will be relatively low.
If Nintendo are indeed using 8nm part made at Samsung, then I'm pretty sure that
nothing is going to be a cheaper overall solution for the product they think they need to make!
