You may have noticed that Nvidia has been incredibly busy in recent years as it essentially moves away from its gaming business. Not only has the company gone all-in on artificial intelligence, but it also sells graphics cards (GPUs) to various data centers, causing prices to skyrocket as demand outstrips availability. As such, it can be difficult to find many true Nvidia fans.
Despite the negativity surrounding the company, there are still three things to praise. The work he put into various things, like image scaling and generation, or ray tracing, still can’t be denied. It would have been great to include Nvidia Broadcast, but almost every thread on Reddit seems to talk about how it’s a resource monopoly in version 2.x updates. There are also a few things we haven’t covered here that you didn’t know your GPU could do.
The multi-billion dollar company is a dominant force in the IT space. So it’s no wonder that some of the things its graphics cards can do have a few supporters. Just be sure to avoid these Nvidia GPUs if you decide to switch.
DLSS
Deep Learning Super Sampling (DLSS) is Nvidia’s frame scaling and generation technology. The benefits of both in improving game performance have been widely praised by users. It is divided into different aspects, with each version of DLSS either bringing upgrades to the methods implemented by the algorithms that power it, or specific features such as “ray reconstruction” to facilitate robust ray tracing methods. Most recently, DLSS 5 was released, which incorporates generative artificial intelligence models to introduce attempts at hyper-realistic graphics for video games. This particular version was not well received.
However, DLSS is primarily the scaling technology, introduced on the RTX 20 series, using the Tensor cores to power it. Essentially, DLSS lowers the game’s resolution, then blasts it with algorithms to achieve the desired resolution, all while taking weight off the graphics card. This is particularly handy for more demanding games on lower-end hardware, but some have seen it as a bit of a crutch in recent years. That said, it is regularly touted as the best upscaling technology available, offering better visual fidelity and performance than its Intel and AMD counterparts.
Nvidia’s frame generation technology has also become popular among Internet users. Despite similar criticisms regarding upscaling, image generation can elevate good performance to excellent, and Nvidia’s special sauce behind the scenes with DLSS can still produce a better image.
Ray tracing
Nvidia still holds the crown when it comes to ray tracing, thanks to several years of improving its technology on the graphics cards themselves. Accelerated by RT Cores, Nvidia even rebranded its GTX line to RTX with the 20 series, as it brought real-time ray tracing to a few select games. Things have improved greatly over the years, mostly due to work in other areas, but online, Nvidia remains the gold standard in ray tracing.
Ray tracing is not new, having been around for decades. Even the Commodore Amiga from 1986 was able to feature a very first ray tracing demo in “The Juggler”. However, Nvidia has been working on the hardware side for so long that it has a natural advantage over AMD or Intel. In benchmarks, AMD’s RX 9070 XT is regularly beaten by Nvidia’s equivalents in terms of ray tracing.
Ray tracing has started to evolve into path tracing, which aims to illuminate games or 3D environments for more realism. Games like “Resident Evil: Requiem,” the capabilities of Nvidia cards and what the game can produce have been highlights of the game’s coverage at launch.
CUDA
Nvidia’s secret weapon in its arsenal, CUDA (Compute Unified Device Architecture), is the software layer developers use to interact with the GPU. This is done via code, CUDA can be used in a variety of software, but is primarily done with C++. With it, programs like DaVinci Resolve can make more use of the hardware to push playback and speed up the rendering of effects on screen. More recently, it has become the primary way for artificial intelligence developers to advance their work on model training.
CUDA has become the backbone of many non-gaming initiatives that Nvidia has worked on. This is part of the basics of robot training, for example teaching the robot to pour into a kettle. From hundreds of attempts, the software, with the support of CUDA, can siphon successful attempts from failures. With AI development at an all-time high and Nvidia continuing to reap huge profits from this boom, it’s no wonder that CUDA constantly makes people wonder whether they should learn it online or make a career in it.
