Nvidia GeForce RTX 2080 Super: ray tracing performance

Just how 'super' are the new cards at the next big thing in gaming graphics?

Review by Richard Leadbetter Technology Editor, Digital Foundry

Updated on 26 Sep 2019

Hardware accelerated ray tracing has proven to be a controversial addition to Nvidia's line-up of GPU features, but the fact that the latest PC titles including Cyberpunk 2077, Watch Dogs Legion and the new Call of Duty Modern Warfare support the technology shows us that the technology is only gaining traction, while support in the next-gen consoles effectively guarantees that the future will be brighter - or shinier and lit more correctly, that is.

In the here and now though, accurately benchmarking games supporting ray tracing is difficult to say the least, because the rendering isn't purely comprised of RT. Rather, we see a combination of techniques - hybrid rendering, where conventional 3D techniques are used alongside ray tracing in creating any given scene. So in effect, what this means is that plenty of game environments may only use RT lightly, meaning that the RTX hardware isn't being pushed to its fullest extent consistently throughout the benchmark. So, in order to get an idea of pure RT power, we do need to rely on synthetic benchmarks - with 3DMark's Port Royal doing the job nicely.

Additionally, we've also devised a path-traced benchmark for Quake 2 RTX, which uses the hardware acceleration properties of the new Nvidia cards extremely heavily. This benchmark is geared towards testing the RT core - so in addition to running through the level, we're also moving the sun position constantly. This stops the game from using its internal temporal component to draw info from the last frame. In 'real world' gaming, Quake 2 RTX runs at approximately twice the speed - but in this stress test, every single rendered frame is forced to calculate data from new path traced calculations. In this manufactured scenario, not even the mighty RTX 2080 Ti can sustain 60fps.

As neither of these tests are proper gaming experiences, we've included a couple of DXR tests too. A benchmark like Shadow of the Tomb Raider's is less indicative of potential RT performance, but more representative of this game's 'lighter' use of the RTX technology - the load on the dedicated RT tech adjusts according to shadow coverage. That means the performance numbers are closer, but plunge into a shadowy pit like the one at the very beginning of the game and the performance between each card radically adjusts (and unfortunately, this scenario is not represented in the bench). Meanwhile, the Provence stage we test in Battlefield 5 looks beautiful and is representative of general gameplay - but screen-filling reflections will see performance drop dramatically.

A lot of readers may be wondering how the RTX 2080 Super compares to the RTX 2080 Ti, especially when overclocking, and how that may affect RT performance. With that in mind, we've included overclocked results. For stability, I could only add 100MHz to the core but astonishingly, the GDDR6 memory overclocked comfortably from 15.5Gbps to 18Gbps - a 16 per cent improvement. When comparing the Supers to the first gen RTX cards, remember we're using Founders Edition boards. The initial RTX offers all had a +90MHz factory OC in place.

Overclocking generally places the RTX 2080 Super at a mid-point between stock performance and 2080 Ti FE throughput - which is the kind of result many people were hoping for out of the box. The combination of extra speed, cores and memory bandwidth produces an OC result that's pretty impressive, but the RTX 2080 Ti FE races ahead regardless. If you're looking for more OC results, check out the 1440p and 4K results on page six.