I see your point, but i still don't understand how those reports could be true if there is no way to dissipate the heat generated from a 10W+ device.
Well, it depends on what metrics each of those people were using to evaluate the performance of NX. Let's take a hypothetical scenario, which is on the upper end of what I'd consider possible:
CPU: 8-core A53, 1.5GHz in handheld mode, 2GHz in docked mode
GPU: 3 x SM, 500MHz in handheld mode, 800MHz in docked mode
RAM: 6 GB, 60 GB/s, 1GB reserved for OS
We can evaluate the claims one by one:
1) Tegra X1 being noisy in devkits would point to an overclocked chip that simulates the (better) performances of the final one
This one's fairly easy, as to emulate docked performance of our hypothetical chip with the TX1, you'd need to clock it up to 1.2GHz, which would require fairly hefty cooling.
2) Osirisblack saying that the console has little to no issue in handling ports from current get
This is going to depend on the game that Osirisblack's contacts are working on. Let's assume, for the sake of argument, that it's a UE4 game. Now, the SoC I've described above has 614 Gflops of 32-bit floating point performance, but unlike the PS4 and XBO SoCs, Tegra GPUs are also able to operate with twice the throughput on 16-bit floats. You can't just use 16-bit precision for everything, or you'll get rendering errors, but you can get away with it for a lot, and UE4 actually uses it for all pixel shaders when running on mobiles.
If we assume that the game they tested it on has a pixel/vertex shader workload split of 80/20 (numbers I've plucked off the top of my head), and UE4 is treating NX like a mobile platform and using 16FP for pixel shaders, and there are no other bottlenecks at all, then the actual floating point performance they'd be able to squeeze out of the NX would be 204.8 FP32 Gflops + 819.2 FP16 Gflops = 1024 mixed Gflops. Now, given that Nvidia GPUs typically have better performance per Gflop,
and UE4 runs quite a bit better on Nvidia GPUs than AMD equivalents, it's entirely possible that in that particular scenario they'd see performance exceeding XBO and possibly even close to PS4.
Admittedly this is a rather contrived scenario, but it's still a plausible one, and in that situation I could see a developer being very confident about NX being able to handle XBO/PS4 ports.
3) LCGeek's comment about NX CPU>>>>Xbone CPU
Again, this depends on the benchmark used. By
blu's matrix mult benchmark, the A53s at 1.5GHz would actually comfortably beat the Jaguars at 1.7Ghz in XBO, although they might not by other benchmarks.
4) Emily Rogers saying that the NX blows the Wii U out of the water, and that it's close to Xbox One (although that's a bit of a stretch, but if it wasn't somewhat close then i don't even see the point in mentioning the Microsoft console, just say that it's noticeably better than the Wii U but still far from the current gen consoles)
This could be explained if she simply heard the raw Gflops (or the person she talked to did). A 600 Gflop machine is obviously much more capable than Wii U, and if you subscribe to the "Nvidia Gflop is worth X more than AMD Gflop" school of thought (which she or her contact might), then you'd be looking at something which could be described as "close to XBO".