Metro: Last Light Redux (2014, DX11)
Technically, Metro: Last Light was introduced in 2013. But the Redux version we’re testing today came out a year later. Its built-in and highly configurable benchmarking tool makes Metro popular for generating repeatable data. Case in point: tests run two years apart on entirely different platforms fall within 1 FPS of each other thanks to a heavy dependence on graphics performance.
Based on our average frame rate results, the Radeon RX 480’s profile did change over time in Metro, improving 1% from 16.6.2 to 18.7.1.
Those 99th percentile numbers are far more interesting, though. Breaking frame rate down by percentile also shows that the slowest five percent of frames are quite different from one driver build to another. AMD’s newest package isn’t even the best. As we saw in the Battlefield games, last year’s 17.7.2 version offered the best performance. But 17.7.2 and 18.7.1 are both dramatic improvements over 16.6.2.
Need proof? Check out the frame times across our test sequence. All of those red spikes correspond to frames that take tens of milliseconds longer to display than the thicker trendline down below. The resulting animation appears less smooth. Fortunately, AMD addressed the issue through its drivers. The 17.7.2 line, in black, suffers the least amount of variance, while 18.7.1 isn’t quite as tight.
Flip right over to Nvidia’s frame time chart and observe the difference. Three thick trend lines convey fluid animation. A handful of spikes from the 385.28 build stand out, but only because the other two versions demonstrate ideal behavior.
Back to GeForce GTX 1060’s average frame rate, there's a mere 0.7% improvement from 368.95 to 398.36, and no real drama to speak of in the slowest 1% of frames.
Despite minimal improvement over time, Nvidia’s card still ends up 14% faster than the Radeon in Metro: Last Light Redux at 1920x1080 using the Very High preset.
MORE: Best Graphics Cards
MORE: Desktop GPU Performance Hierarchy Table
MORE: All Graphics Content