Again, z0m3le, what is technically possible provides an upper limit on what some hardware could actually be, it doesn't provide the answer to what it will be and never has.
I mean, you actually say 600 GF, which is already significantly less than anything one could calculate by looking at technical possibilities, so you seem to agree with that. And really, the difference between 600 GPU GF and the 512 GPU GF in the rumored dev kit is minute.
Personally, I don't think it's unlikely for the hardware to reach that, or even 750 GFlops. What I'm far more conservative about is what clock frequencies Nintendo will actually allow in portable mode.
That png? I certainly hope everyone does, it screams fake out of every pixel.
Again, I'm calculating Pixel C's power consumption at 8 watts and Nvidia's statement about 40% improvement over Maxwell with Pascal, Pixel C is again passively cooled, so docked, the device's performance exceeds Pixel C by a large enough amount that this device that is twice as thick needs active cooling.
Pixel C has a large bright screen, it is both bigger and likely brighter than NS, thus the NS likely uses less power on the go. At 3 Hours max, with a battery the same size of N3DS XL, you'd be looking at about ~7 watts on the go, which is Vita's level. That puts this device around 600gflops.
Thanks to Laura's source, who I'm going to trust until she stops batting 100% on NS, confirmed that the clocks of the device increase when docked, Pascal as you know sits very comfortably at 1.5ghz, it wouldn't surprise me at all if this device is around 750gflops docked, and hey look, that is what the benchmark a couple pages back says about Pascal Tegra, so yes those numbers are not wild guesses, but educated ones.
BTW the rumored dev kits were overclocked, thus they were >1ghz, saying they were running hot and loud with Maxwell, which leads me to believe that ~600gflops is pretty much where it would have to be, since X1 has been out for a year and a half and runs fine at 1ghz.