By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Soundwave on 11 February 2024
Edit Thread
haxxiy said:
Soundwave said:

Problem remains this is a massive waste of SMs. You could get the same performance from 8SMs and just clock a bit higher and have a cheaper chip. Why pay for 12SMs, it's a way larger chip, doesn't really make a whole lot of sense. Like this doesn't even align with things Nintendo has done in the past if that's the argument. 

210 MHz, lol, the Wii's 2001-era GPU has a higher clock than that. These are insanely low clocks for a more expensive and much larger chip for no reason. This is like going out of your way to buy a jumbo popcorn at the theater and paying the $6 premium for it and then eating 10% of it, you could have just bought a freaking regular popcorn and not have paid the extra money. And if the argument for doing so is "well I did that because I'm cheap" ... it's like what? lol. How does that make any sense. 

Because you'd save power. Most hardware runs at just a fraction of how efficient they can be because they're placed far, far above the optimal point in the voltage vs. frequency curve. Besides, again, the older node would be cheaper even with a larger chip, and the frequencies would be higher than that.

What, you think Nintendo would take in the cost of the die shrink just so it can run at a higher frequency? Just so they could have better graphics?

Reminder this is the same company that released the Wii U in 2012 with a ~ 1997 architecture CPU, and underclocked a 15W Tegra by 65% to be 20 times slower than a GTX 1060 with the undocked mode Switch.

All of that being said... I do think the console can be 5/4 nm as I said before, and I hope it is. It's just that a lot of people here are fuming and screaming at the mere thought of it and it definitely there's a universe it could.

I mean you could still clock ridiculously low and get the same performance from 8SMs. I'm not even talking about any other node, I'm talking about 8nm strictly.

At 8nm you could clock an 8SM chip 50% higher and the clock speeds would basically still be very low, 210 MHz increased by 50% clock speed is still a very low 315 MHz, that's still lower than the current undocked spec for the current Switch, lol (current Switch runs at 384 MHz undocked). 

So why pay for 12SMs. It's a massive chip with a huge number of extra graphics cores for no reason. If the argument is "because Nintendo" that doesn't even make sense, because if the logic is "they're cheap" ... so they're cheap ... but they're paying for a more expensive chip with more graphics cores ... because ... ?

I'm not even sure what kind of flexibility Nintendo would have with an 8nm Samsung chip because from what I've heard, 8nm process is a dead end that isn't compatible with their 4nm/5nm processes. Nintendo wouldn't be able to die shrink for future models like how the Switch got the Mariko model + Lite model. Nvidia has not booked any Samsung business outside of their 8nm node either and it doesn't look like they will (they have booked TSMC 3nm, which would die shrink nicely from TSMC 5nm/4N). So I guess Nintendo is just good with losing 20-25 million hardware sales right there?

There's a lot about it that just doesn't make sense.