Ever since Nintendo announced the Switch, there have been questions about the platform’s launch lineup, especially after we discovered all of the footage displayed in the launch trailer was inserted in post-production. A new set of rumors about the platform suggests it’ll launch with a new Super Mario Bros, possibly a new Splatoon (it’s not clear yet if this will be a brand-new game, a port of the Wii U version, or something in-between), and maybe Skyrim, but Bethesda isn’t saying one way or the other yet. Earlier information suggested Nintendo would push The Legend of Zelda: Breath of the Wild out for the Switch’s launch, but follow-up reports have suggested the game needs more localization and bug-fixing work and may not arrive until summer. Nintendo has no experience with this kind of open-world title, so we aren’t surprised to see the company may need some extra time to finish it.
Geek.com is reporting that a new Mario title with an open 3D hub world will be available at launch with cooperative multiplayer and a tether mechanic to keep players in the same area of the game world. ArcadeGirl64 also reports that Nintendo is planning to port Mario Kart 8 to the Switch — the new version of the game will contain all of the Wii U content and DLC, but will add new tracks, new playable characters, and a new battle mode. It’s reasonable to think Splatoon might be a similar half step between a brand-new game and a straight port.
How much performance can the Switch offer?
I’ve been having some fun trying to model what kind of performance we might expect from the Switch relative to the Wii U. There’s a fascinating set of advantages and disadvantages to consider, and it’s a fun puzzle to try and figure out where the platform will land.
The CPU inside the Wii U was an absolutely archaic PowerPC 750CL core with very limited support for SIMD (Single Instruction Multiple Data) a relatively low clock speed (1.2GHz) and an odd three-core arrangement with an asymmetrical L2 cache (one core had 2MB, the other two cores had 512KB). I’m going to assume that the Switch will target 4-8 CPU cores in a big.Little configuration, with the CPU itself most likely based on an ARM Cortex-A57 or A-72. Nvidia’s Project Denver is also a possibility, but we don’t know enough about how that chip may have evolved to estimate anything about its performance. The Switch’s CPU should be far more powerful than the Wii U’s, even after allowing for a mobile platform and much smaller power envelope. Even if we assume a 2GHz sustained clock rate, the Switch would have a 1.67x clock advantage over the Wii U, at least 33% more CPU cores, a much improved SIMD implementation, and a much more modern 64-bit instruction set compared with the old PowerPC chips.
The gap between the Wii U’s GPU and Switch’s Pascal-based hardware isn’t as large, but should still be substantial. Nintendo was also cagey about what, exactly, the Wii U’s GPU could do. But the best available analysis suggested the chip was based on AMD’s old HD 4000 architecture, which debuted in 2008. That’s better than the 750CL (1999 wants its architecture back), but Pascal will still be far more efficient and supports GPU compute in ways not contemplated by hardware designers of eight years ago. Again, if this were a traditional living room console, we’d be predicting big visual and performance improvements within the same TDP (the Wii U drew ~35W at the wall).
Just for kicks, I mapped out the degree of improvements TSMC claims to have delivered with each successive foundry generation. The Wii U was built on 40/45nm technology, while the Switch is presumably a 16nm FinFET design at TSMC or a 14nm chip at Samsung. If you start with the 35W power draw measured for the Wii U and start cutting it by the degree of improvement TSMC claims to deliver at each successive node, a 16nm Wii U would draw about 13W of power. Tablets typically target between 7-15W of power consumption, so we’ve got a good starting point.
Here’s where things start getting tricky. The Wii U’s power measurements at the wall wouldn’t include a display, while the Switch has to provide one, cutting into the device’s power budget. We can assume that the Switch’s Pascal + Project Denver or ARM architecture design is significantly more efficient than the Wii U, so let’s figure that the new architecture is 2.5x more power efficient (that’s over and above the efficiency gains the platform picks up from its 40/45nm – 14/16nm die shrink). Now we’re providing a Wii U-equivalent experience in just 5.2W. Toss in another 1.5W for the display, and our hypothetical Switch is pulling just under 8W for a Wii-analogous platform. So far, so good, but everything we’ve theorycrafted to date just gets us to the Wii U. How much additional improvement can the Switch offer?
The answer is almost certainly “less than some people would like.” Nintendo is going to have to balance weight and battery life. The 720p 6.6-inch screen (estimated) is a good sign that the company wants to hit targets that appeal to the mass market rather than pushing high-resolution content, but there’s still a lot we don’t know about the platform. Since the Switch is meant to be docked when attached to a big screen, it’s possible that Nintendo will include custom upscaling hardware in the dock itself, thereby saving battery life when the system is in handheld mode. Wi-Fi radio performance and power management are also significant components of keeping tablet power consumption low, and we don’t know yet who provided the Switch’s radio(s).
Add up all the various factors, and the Switch should be able to offer a modest improvement over the Wii U, but will still lag the original Xbox One and PS4 — there’s no way to take consoles that drew more than 100W in 2013 when gaming and shrink that into a 10W form factor just four years later. The fact that we’re seeing Skyrim shown off as a potential launch title, even if Bethesda hasn’t confirmed it, would support this. Skyrim is a great game, but it was built for older console hardware. On the bright side, porting Skyrim to the Switch would mean all those comparisons between console and mobile graphics would finally start being accurate… provided you were talking about the PlayStation 3 or Xbox 360.
So here’s my best guess: I think Nintendo will focus on a relatively conservative lineup of launch titles with a few tentpoles and some Wii U ports that showcase the improvements of the new platform with additional content and features. The Wii U sold so poorly, plenty of these titles will be new to Switch owners if the console takes off, and porting older games may be an easier lift in some respects than creating new ones from scratch. Battery life on the new platform isn’t rumored to be all that great, and that’s not surprising — every watt Nintendo devotes to outpacing the Wii U’s graphics is one less watt for longer gaming sessions.
We still don’t know if the dock will contain any additional hardware that would boost the Switch’s performance (Nintendo filed patents alluding to this possibility, but has said nothing about it to date). Assuming it doesn’t, the Switch will probably be a play for customers who aren’t looking to take the Sony or Microsoft plunge into 4K. This strategy worked extremely well for Nintendo when it launched the Wii, so again, it makes sense to revisit it now with the Switch.