Nvidia Misleading DLSS/AI and Tensor Core
I made a promise to myself to never talk about Nvidia tricky marketing anymore
but I think GTX users and players in general deserve the true behind Tensor Core and DLSS/AI
Let me be clear on one thing that I don’t appreciate about Nvidia, it is the NV telemetry which are extremely invasive when it come to privacy, they basically use it to gather as much metadata as they wish and add on top of the that the fact they don’t care about users loyalty and they make it clear to force people to push to buy new technology on top of technology.
In the risk of repeating myself, there is no game out there exploiting even half of the power of a GTX in general, count also all the AMD and RTX in the market.
Game are capable to run on every platform, it is all a matter of developers to take in consideration how versatile their Engine are and do they take in consideration each GPU features to run well.
Power is nothing without proper control, a bad software will run bad on a strong power but a good software will run good on poor hardware depending of the software adaptation of the hardware, you can as for example IOS and Apple hardware, when when is all about optimization.
Now let’s go back to what Nvidia is hiding behind Tensor core, AI and DLSS, same goes to raytracing, which if you looked at my magazine G2K and previous work, I make it clear that Raytracing can run on every GPU but you have to work on a specific render which are most core are capable to execute, there os different way to provide raytracing but the way Nvidia do it is pretty lame.
The reality behind Tensor core is that they are just dedicated cuda core same as the one on GTX, they are just cuda core being use differently for certain execution, is like dedicating PhysX to being rendered by CPU or GPU, basically the driver make it clear where to look at to execute the algorithm.
Tensor Core is similar, there is a thing that most developers linked to nvidia don’t want you to know or technology media doesn’t know about, on top of the Windows system for example there is a sub layer use by Nvidia at low level which you have to worry or aware of but it is the most important part of the process that bring those AI, DLSS and so on possibility.
Called NGXCore, this is pretending to be the AI, while is just an algorithm engine that triggers function depending on the architecture you are using, let’s say is like a Denuvo DRM, while the scandal made around Gsync was similar, a licensed key under the hood unique to each platform to make it work.
This is what Nv telemetry has been doing, learning and gathering how users use their computer, after that they introduced NGXCore that deal with verification, validation, creating, terminated process linked to DLSS, Tensor core = Cuda Core but with NGXCore make it only possible to RTX architecture, it was the same for Raytracing.
To be clear DLSS is definitely not an AI or Tensor core needed function, and Nvidia has been pretty misleading everyone about there supposed AI software and Tensor core which are only part of a full rendering engine under NXGCore
DLSS is what console has been doing for decade which was available on everything possible platform, that called dynamic scaling, let’s set with put the texture in 4K but we rendered in lower resolution, it is basically the same thing and that is mostly software or engine related, ID Tech have been doing that for far so long, it is all a matter of developers introducing and the way how Nvidia is locking this function only on RTX and pretending Tensor core while it is only Cuda Core + NGXCore engine, that pretty lame for Nvidia.
They have been offering this on Nintendo Switch (GPU Pascal based architecture) they basically can offer that to everyone but they won’t to force you buy new hardware.
AI,Tensor Core, DLSS, ResizeBar are all part of a engine sub layer execution, which is why they are boycotting Linux and MacOS platform because those platform doesn’t give them full control to take care of your computing usage.
While AMD have been giving all possible tools so users can enjoy, Nvidia has been the worse at involving their way of dealing with users need, to be clear is not the users who should adapt to the market but the market to adapt to users need and not release each year a new GPU and lock out previous one, this greedy way of pushing people to buy more and more, only make the video gaming industry worst by rushing game and serve players with half optimized game.
To be clear and this come only from my past experience as a dev, AI, DLSS, Tensor core function can be fully available on previous architecture, it would be nice that all youtubers, media web or Nvidia stop false claim that DLSS for example is only possible on RTX, this is was a full of lies on their part and a way to push forced marketing acceptance.
By the way, NGXCore was introduced way before RTX was announced, nearly at the same time than Nv Telemetry and it was this sub layer engine that was dealing with unique function proper to Nvidia such as Gsync, PhysX, etc this one also deal and trigger execution with DXGK, which are part of the Kernel Windows, all this are made at the low level of Windows
Last thing, It is funny to go back talking about that, when I was announcing that Gsync, Fastsync and so on, was part of this and no one previously believe this matter, but after Gsync was introduced to most monitor and so on, people recognize their mistakes, same as RayTracing, but raytracing is even more lame, Nvidia purposely slowing down the performance of the cuda core when it come to previous generation by redirecting and forcing a wrong way of rending under NGXCore.
I’m going to repeat but most media, youtubers and so on playing Nvida game should be ashamed considering how much users have paid for their GPU and what is expected.
Let’s not talk about Sony Ps5 and the PS4 Pro, same goes for Xbox One X and Series X. ( PS4 Pro could run as good as the PS5, PS5 need double amount of transistors to execute the same exact operation as the PS4 Pro, the Series X itself is unique as it double the one X but previous architecture is actually better sometimes than the new one when it come to same rendering, which is why those platform was actually not needed right now)
Anyway that the end of this article, thanks for taking to read all the way to the end, like I said, sharing is caring and I think users/players that invest hard work money and so on technology should deserve the true behind and have the possibility to use it at the full potential.
Peace, take care, stay safe