Metro: Last Light (DX11)
Despite its age, Metro: Last Light is a solid pillar of our benchmark collection, as the game can still really sweat even the most modern graphics cards. We use the "Very High" preset with 16x AF as well as Motion Blur and Tesselation on "Normal".
At 2560 x 1440 pixels, all cards starting with a Radeon R9 Fury X have quite easy play with Metro. The Titan X (Pascal) and GeForce GTX 1080 Ti deliver comparable performance, outperforming the normal GTX 1080 by around 19 percent.
The 4K readings are marginally more interesting, as the flagships of the last generation cannot provide a completely fluid representation in the top detail settings in Metro, whereas the fastest Pascal-based maps can do so. Not even a GeForce GTX 1070 is enough.
As we saw in GTA V (also a DX11 title), the GTX 1080 Ti has to suffer a small defeat at 2560 x 1440 pixels, and then turn it into a small victory at 3840 x 2160 pixels. The resulting average frame rate is 26 percent above Nvidia's normal GTX 1080 and 60 percent higher than the aging GTX 980 Ti.
Rise of the Tomb Raider (DX12)
We've also tried to make Rise of the Tomb Raider a bit more challenging compared to older tests. For this purpose, we have enabled 2x SSAA instead of SMAA, as the latter post-processing effect leaves many surfaces untreated.
The performance slump by improving the internal rendering solution is understandably huge. Where the Titan X delivered 110+ FPS with SMAA in August last year, the card is now below 80 FPS. However, this is still a good result for ultra-high-end GPUs.
With an average value of 79.5 FPS, the GTX 1080 Ti is about 34 percent faster than a GTX 1080 and 70 percent faster than a GTX 980 Ti.
Although this is a DX12 title, AMD's Radeon R9 Fury is struggling. Several large frame-time rashes are noticeable as problematic sequences, which our Unevenness index classifies as unplayable. For the lower half of the test field, things are not looking very good just before the switch to 4K.
Even the GP102-based graphics cards suffer from super sampling. According to our Unevenness index, the Titan X (Pascal) and the GeForce GTX 1080 Ti are still in the playable range, but the other cards are jerking comfortably in front of them. However, given the pixel density of a 27-square-mile 4K display, it would be a good idea to simply dispense with anti-aliasing and enjoy the additional performance.
Tom Clancy's Ghost Recon Wildlands (DX11)
We got our keys for Ghost Recon Wildlands when we were just finished measuring. Nevertheless, we decided to include this title – the key word novelty factor. The game's "ultra" detail settings have an incredibly powerful impact on performance, especially given a graphics quality that we think is good, but not groundbreaking. So we scaled it back to "Very High," which has the positive side effect of shutting down Nvidia's "Turf Effects" feature. Until we can quantify the impact of this feature on AMD's hardware, that seems only fair to us.
Ghost Recon Wildlands uses the AnvilNext 2.0 engine, but launches as a DX11 title. So we weren't surprised that Nvidia's cards perform significantly better than their AMD competitors in this TWIMTBP game.
The better comparison is probably the one between GTX 1080 Ti and Titan X (the Ti is a bit faster), GTX 1080 Ti and GTX 1080 (23 percent advantage for the Ti) and GTX 1080 Ti vs. GTX 980 Ti (Pascal trumps Maxwell by almost 56 percent).
The "Very High" detail setting is a heavy load under 4K even for the fastest single GPU graphics cards. Interestingly, the Titan X grabs the crown here, while the GTX 1080 Ti is "only" 18 percent faster than the normal GTX 1080.
Our Unevenness index doesn't benefit from a single card here – all are far from delivering perfect results. Ergo, we might need to take a closer look at Ghost Recon if both AMD and Nvidia could do a bit of driver optimization.
Kommentieren