Speculation and rumors run rampant when it comes to Nintendo's mysterious Switch. The hybrid console allows for portable and home console gaming, and lots of analysts, spectators, and drive-by tech gurus have been offering their take on just how powerful the Nintendo Switch might actually be.
According to a lengthy report from Venture Beat, developers have been given early development kits of the Nintendo Switch, and the device is supposedly running on a custom made Tegra X1 chip based on the Maxwell 20 nanometer design as opposed to the smaller, more efficient, more powerful 16 nanometer Pascal.
According to the people they spoke with under anonymity, the Nintendo Switch is estimated to have around 1 teraflop worth of processing power and is behind the base model PS4 in raw output, which is estimated at around 1.8 teraflops.
They mention that the overall memory bandwidth is lower than the PS4 given it has 4GB of RAM (but they don't specify what kind of RAM) and the display is likely going to be 720p undocked, and up to 4K docked thanks to getting a feed of extra power from the station.
Now here's where things get interesting: They're basing the specs and raw performance on the Maxwell designs used for Nvidia's Shield, which came out back in 2014. So for those of you thinking that the Nintendo Switch is sounding extremely underpowered, keep in mind that even though the basic development kits were based on the Tegra X1, those weren't the actual specs of the Nintendo Switch.
According to Nvidia they've made a custom version of the X1 based on the Maxwell chip, specifically for the Nintendo Switch, which is why the console is not backward compatible with the Wii U, since it's a complete overhaul of the architecture. While we know it won't be equivalent to the Pascal line, which is linked to the GTX 10xx series, the Maxwell is still powerful enough to get the job done, as evidenced by the longstanding success of the GTX 980.
And here's where things get even more interesting. Early speculation states a Pascal in something as small as the Nintendo Switch would burn the whole thing up from the inside out due to heat generation. However, Venture Beat indicates that the Nintendo Switch's custom chip could be designed to clock higher as an APU when docked, as opposed to when it's not docked. This would mean that when undocked, the APU clocks lower to conserve power and extend the battery life. When docked the Nintendo Switch gets powered up by the direct feed and the unlocked clocks would give the single CPU and GPU chip scaled performance. How much does this performance actually scale? We don't know yet.
Essentially, when docking the Nintendo Switch it would work equivalently to overclocking the system. When undocked it would run in a low power state that's still powerful enough to run games at 720p HD at around 30fps.
Venture Beat speculated that the reason Nintendo passed on the Pascal was because it would have delayed the release of the Nintendo Switch and it would have required more R&D to get the chip's heat consumption to shrink to an appropriate size that would have proper heat regulation and still maintain performance.
Of course, right now we can only go by anonymous sources working on older dev kits based on even older Nvidia portable hardware. So while we have a good idea of what the specs of the Nintendo Switch might be, we won't know for sure until Nintendo reveals more about the system on January 12th next month, or until tech gurus get their hands on the actual hardware and run some thorough benchmarking tests.