Next week, Nvidia is holding a major event at Gamescom to tease the launch of a new GTX — or apparently, RTX — 2080 high-end GPU based on its just-announced Turing GPU architecture. As part of its Turing architectural unveil last night, NV published a teaser video for the next rollout. The video contains a number of clues to the naming convention, including a user named “Not_11” saying “Eating, give me 20.” Add up the screenshots seen on Discord, and the brand name — RTX 2080 — comes into focus, even if Nvidia hasn’t formally, officially declared it yet.
So far, Nvidia has focused on talking up its RTX cores as a major force to be reckoned with and a sea change in where gaming and professional graphics could be headed next. As we alluded to last night, it’s not the first time we’ve seen a company take a major bet on a new rendering technology and hardware to accelerate it. Nvidia’s strength and market share in graphics make it possible that the company will push a new wave of capabilities into the market in a way as fundamentally different to graphics as the advent of the G80 back in 2006 (Nvidia itself has explicitly made this comparison). But if I seem dubious, it’s because, well, the history of graphics and computing, in general, doesn’t favor the fast sea change.
The debut of Vulkan and DirectX 12 was supposed to revolutionize gaming, yet Vulkan and DX12 are used today by a bare handful of titles. Performance gains over DX11 have not generally materialized as hoped for. There are a variety of reasons for this, but one of the most important is that hardware refresh cycles are slow these days, it takes time to update both hardware and software to support new APIs, and taking advantage of the capabilities of those APIs can be more difficult in some cases than the previous, less-optimized solution. Even so, if you date the appearance of low-level APIs in PC gaming to Mantle back in 2013 — and we should — it’s now been 5.5 years since AMD first introduced a major new type of API. And games using those APIs are few and far between.
This is far from the only example of this kind of trend. VR remains bottlenecked and available to a scant handful of consumers even after Nvidia put a huge push behind it in 2015 – 2016. Both Nvidia and AMD put a huge push behind 3D gaming and multi-monitor gaming, respectively, back in 2012 – 2013. In both cases, just a handful of gamers bought into these capabilities and features.
My point here is not to slam Nvidia’s work on ray tracing or to diminish the concept. Ray tracing has been a topic of interest to me since before Intel killed Larrabee, and I think it’d be fascinating to see the technology deployed widely in game engines. But while we haven’t yet seen what Nvidia can do with its new RTX GPUs in modern game engines, there’s one thing I can absolutely tell you — ray tracing is a fundamentally different approach to graphics than conventional rasterization. Even if we assume that Nvidia has discovered the Next Big Thing and that the entire graphics industry will race to get aboard, it’s going to take some years for these changes to proliferate through the product lines. It’s all well and good to launch a new line of highly capable RTRT accelerators, but the fact that it’s just those cards that can hit such performance figures will limit the rollout of new technology and features. Developers, after all, have to maintain high performance on entire families of GPUs, and plenty of people are still using older cards today.
It’ll be damn interesting to see what Nvidia has in store and why it thinks ray tracing is the future today. But that future will take a few years to arrive, no matter what the RTX 2080 offers in terms of specialized ray tracing hardware.