In this last days there was a lot of talking about floating point precision (fp16 vs fp32), floating point operations per second (TFLOPS here and there), etc... all because our subconsciousness hopes that the NS could be as powerful as a PS4, but with the added value of portability. I think we have to shift the point of view for a moment and aim at something more realistic and lower our expectations, and to back up this I'm going to justify through real world examples (and not entirely through abstruse hardware numbers) why we don't really need a portable device as powerful as a PS4.
As a starting point, I'm going to assume the worst case scenario considering a NS running a 1-year-old Tegra X1. For the next examples I'm also going to consider the already selling consoles PS3 (or Xbox 360) and PS4.
First, let's consider these videos 1 - 2
As you can see, at first, there seems to be not a big difference between the PS3 and the PS4, but with more attention to the detail you will notice that the textures in the PS4 are just a little bit more crisp and more detailed. Also, there is an obvious difference in resolution, because the PS3 supports 720p while the PS4 supports 1080p.
What about the Tegra X1?
Here is a gameplay comparison involving Resident Evil 5, Android Tv (Tegra X1) vs Xbox 360
As you can see they are graphically almost equal. Before you tell me, yes, it's obvious there is a difference in performance, but I'll explain the reasons later.
Now, why Law of the Diminishing Returns?
Here is a good image that summarizes this concept.
Economically speaking, in terms of energy economy and computing economy, an order of magnitude increase in graphical precision nowadays is barely noticeable while requiring a lot more power to support it.
Now we need to talk about sets. The NS is both a portable system and home system, but the home system capability leverages on the portable hardware. So, portability is the real set we need to consider because it's the worst case. On the other end, the PS4 is home gaming only.
All these things said, we can't compare the NS with a current gen home gaming console because:
- The NS screen is 720p so at least its GPU supports 720p gameplay, while the PS4 supports 1080p;
- In terms of power consumption, achieving current gen graphics requires more power with little graphical quality gain over the previous gen graphics; power consumption is more important on mobile;
- Most importantly: the NS is portable while the PS4 is not;
We can't compare a portable device with a PS4. A more realistic expectation would be comparing the NS with the previous gen PS3 and Xbox 360, because the mobile technology is now able to give us chips with smaller transistors and better architectures giving more power efficiency both computationally and energetically (manufacturing process from the PS3 60nm to the Tegra X1 20nm).
What about the performance differences between current implementations of the Tegra X1 (Shield Android Tv) vs previous gen consoles? As always economy kicks in. While the hardware of the Tegra X1 is a lot more powerful than that of an Xbox 360, we should also take in consideration the software side. If the software is not optimized for the platform it runs on, then it would not use the hardware at its full capabilities. Software optimization costs, so videogame companies are not going to spend money and human resources on optimizations for a niche device, the Android Tv, that would only yield small marginal profits, considering its sell rates vs home gaming consoles. On the other hand, the NS is not going to be a niche device, but a fully fledged console made by one of the biggest players of the videogame industry aiming at selling 2 million devices in the first month.
Talking about software, now things get interesting. Everyone in this sub loves filling his mouth with hardware jargon, but no one talks about software, the real money maker for Nintendo (usually console producers sell consoles at a loss, making profit on the videogames) and the supporting videogame companies. Ok, I said we should compare the NS with the PS3 hardware-side in the worst case scenario, but what about software-side? The Tegra X1 supports both Directx 12 and Vulkan (while previous gen consoles don't), which both introduce low level management of GPU resources increasing the overrall software performance. Directx 12 and Vulkan are fairly new and pretty difficult to use, so maybe not everyone is going to use them. However, the Tegra X1 also supports OpenGL 4.5, which give the devs a lot of optimization tools like tessellation which wasn't possible with the PS3.
So, we have an efficient SoC capable of better than PS3 graphics and most importantly it's mobile, so better energy efficiency. All of this is already possible with 1-year-old technology.
In conclusion, we shouldn't be worried about the NS being able to compete against Sony and Microsoft in the home gaming battle, because Nintendo is already the winner of the handheld war and having an handheld capable of PS3.5 graphics usable both at home and away from home is a good compromise.
Here is a good read about mobile CPUs vs Dektop CPUs explaining why they shouldn't be compared solely considering their operational frequency:
For the interested, here is a collection of boring technical material that supports my post:
An interesting read about fp16 vs fp32 (I think we should love fp16 though, but that needs its own thread):
