NVIDIA Has Doubled Down On RTX — It’s Here To Stay

When we previously discussed the merits of NVIDIA’s RTX 20xx series and GTX 16xx series of video cards, we mentioned that we thought it was strange NVIDIA would release a bunch of cards in the RTX 20xx series, marketing real-time ray tracing as the “next big thing”, only to release another series of cards that do not support it at all.  We argued that the high price tag given to RTX cards, which entered the market at an even higher price point than the previous GTX 10xx series of cards, ensured that they would not get the same market share that NVIDIA was used to from previous generations.  We argued that this lack of market share for any cards supporting real-time ray tracing would result in developers not spending the time, effort and money to make proper use of the technology in new games.  After all, if only a small segment of consumers are even capable of seeing the added ray tracing effects, why bother investing in it?

Since then, real-time ray tracing has only been applied to a handful of games, including an RTX re-make of Quake II.  That said, there have been several big developments recently with regards to ray tracing.  First of all, NVIDIA released a driver patch that allows older Pascal cards (The GTX 10xx series) to use their regular graphical computing power to deliver real-time ray tracing.  However, the fact of the matter is, even though these cards now officially support it, ray tracing uses up so much computing power that anything shy of the 1080 Ti is unlikely to provide a smooth playing experience (ie. at least 60 fps on average), even at 1080p.  The 1080 and 1660 Ti may also provide a playable, but sub-60 fps playing experience.

Realistically, you’re going to want an RTX card if you want to take advantage of real-time ray tracing, but even RTX cards suffer a pretty substantial performance hit when ray tracing is enabled – in one test that Tom’s Hardware did, the RTX 2080 saw around a 40% fps loss, while older Pascal cards saw losses up to 60%.  While the RTX 2080 performed better than the competing GTX cards, both on an absolute basis (obviously), and on the basis of percentage of FPS lost when ray tracing is turned on, the fact that the RTX 2080 still saw a loss of 40% suggests to us that dedicated RT cores are not the silver bullet they claim to be.  To us, this begs the question: If dedicated RT cores don’t actually do the brunt of the calculations when ray tracing is enabled, why even have them? We think NVIDIA should either remove the dedicated cores (and in their place add additional general purpose CUDA cores), or alternatively, include enough dedicated RT cores to ensure that the ray tracing workload is explicitly done by the dedicated cores only, instead of spilling over into the rest of the cores, destroying framerates.

But we digress, the point of this article is to demonstrate that NVIDIA isn’t giving up on RTX as we thought they may have when they started releasing the GTX 16xx series of cards, but rather, they have been promoting it while expanding their product offerings.  More substantial than a software update, NVIDIA announced at Computex that they are entering the high-end design sector, one that has traditionally been dominated by Apple, with the introduction of mobile RTX Quadro cards, allowing designers to see in real time how modifications to their three-dimensional scenes effect lighting.

In conclusion, it seems that NVIDIA has decided that real-time ray tracing is here to stay, and while the first set of RTX offerings weren’t met with a huge amount of excitement, we expect that NVIDIA will continue to develop the technology, and hopefully they will find a way to design their RTX cards so that they don’t have a noticeable performance hit with real-time ray tracing enabled.


Be the first to comment

Leave a Reply

Your email address will not be published.