Metro Exodus is close to our usual GPU benchmark treatment, where 36 graphics cards on the bench get a full performance comparison. The game was generally well received, although it has only been available for a few days. We had access to Metro Exodus prior to release, which not only allowed us to build our big GPU benchmark but also to review ray tracing and DLSS performance. Along with the game's release, AMD and Nvidia released updated drivers that were used for this test.
We briefly tested the previous display drivers from both camps, although as far as we could tell we didn't see any real performance improvements. What improved performance was Metro's Day 1 patch, so our results may differ slightly from those released early.
In general, performance fluctuates widely in most games, depending on the levels and areas and therefore what is rendered. Although developers are trying to optimize levels for more consistent performance, we wouldn't say they did it badly with Metro Exodus, but you'll see frame rates with a GTX 1060 at 1080p up to 70-80 fps in some areas and then only 30-40 fps in others.
We set the benchmark for a 60-second pass in the game in a challenging section that is at the beginning of the game. The benchmark takes place very early in the campaign in Moscow and starts at the first checkpoint after the pack of mutated wolves has passed by. For testing, we use the "Ultra" preset, in which Nvidia HairWorks and Advanced PhysX are deactivated by default. However, tessellation is enabled.
The game also supports DX11 and DX12, but performance with a Core i9-9900K was identical with both APIs, so we decided to try DX12.
Our standard GPU test bench contains a Core i9-9900K with 5 GHz and 32 GB DDR4-3400 memory. The latest display drivers have been installed, for AMD the Adrenaline 2019 Edition 19.2.2 and for Nvidia the GeForce 418.91 drivers.
Start at the top and work our way down. As you can see, the RTX 2080 Ti had no problem keeping things above 100 fps, and while 131 fps at 1080p aren't amazing for this extreme high-end GPU, it's better than what we like in titles Assassins Creed Odyssey, Just Cause see 4, Hitman 2 and Kingdom Come Deliverance for example.
If we go through the list, we see that the RTX 2080 delivers very solid performance. Here it was 13% faster than the GTX 1080 Ti, so this title seems to be well optimized for Nvidia's new Turing architecture. The 1080 Ti is still doing well and you can overcome the much newer Radeon VII. Meanwhile, the RTX 2070 and 2060 offer very similar performance and both were faster than not only Vega 64 Liquid, but also the GTX 1080.
For those hoping to get around 60 fps at 1080p, you'll need a fairly high GPU firepower. At least a GeForce GTX 1070 or Vega 56 is required, Vega 56 was the faster of the two, since it roughly corresponded to the GTX 1070 Ti.
Then the Radeon RX 590 or GeForce GTX 1060 works for an average of 50 fps and here we see that the 1060 6 GB is 9% faster than the RX 580 and 14% faster than the 3 GB 1060. Then everything under the RX 570 or R9 390 I would consider it slow to achieve a comfortable gaming experience. This meant that the weakest Nvidia GPU you want to play with is the GeForce GTX 970. As we said before, the game requires serious 1080p hardware if you use the highest quality preset that is not the highest quality preset available, be extreme.
Starting with the RTX 2080 Ti, we see again that the average frame rate remains above 100 fps, while the 1% low dropped to 90 fps. Still, that's pretty impressive since the RTX 2080 couldn't even reach 90 fps on average, but 83 fps. The Radeon VII was 71 fps, 14% slower, 8% faster than the RTX 2070 and only 16% faster than the RTX 2060, not great as it costs twice as much.
The RTX 2060 is actually a really good value here. It matched the GTX 1080 and was not noticeably slower than the 2070. Vega 64 Liquid is fine, but Vega 56 is the standout AMD option, although it was wasted until 2060, but of course this is a title sponsored by Nvidia.
The older Maxwell GPUs from Nvidia are pretty good at it, the 980 Ti fit the GTX 1070, while the 980 fits the 6 GB 1060. The 6 GB 1060 was also 12% faster than the RX 580, although none was great at 1440p and averaged less than 40 fps. Basically everything under the RX 580 or GTX 970 was not playable with this resolution.
Then, at 4K, good luck, you need an RTX 2080 Ti or you have to settle for inferior quality settings. At this point, you might as well bite the ball and just play at 1440p.
The game was playable on a GTX 1080, RTX 2060, Radeon VII, or something better, but as soon as you find that 1% performance drops below 30 fps, the stuttering frame rates affect the experience.
Before we finish, we re-tested 31 GPUs, many of which have not been tested before, to see if you can get away with older hardware at 1080p with the medium quality preset and tessellation disabled. Of course, HairWorks and Advanced PhysX were still activated off.
The medium quality setting enabled the GTX 970 to achieve an average of 92 fps. This is a massive 119% increase in performance over what we saw with the ultra quality preset. Similar increases were also recorded for the RX 570 and the R9 390. The GTX 1050 Ti doubled the frame rate, which meant that it could now deliver highly playable performance, which was good to see.
We also see some vintage cars like the HD 7970 and GTX 780 Ti that offer playable performance, and the 7970 is well aged. The R9 280X is a revised 7970 with a factory overclocking, and we can also see that it outperforms the 780 Ti. The slowest GPUs you can get through here include the R9 380, GTX 780, 950 and 1050.
At the end of the tests we have some preset scaling results. We just saw how the medium quality preset offered a massive performance boost for older GPUs, especially those with 4GB less VRAM. The increases for the 8 GB RX 580 and 6 GB 1060 are not quite as extreme, but they still saw a performance increase of around 50-60% if they were due to the ultra-preset medium quality.
It is interesting here that the RX 580 and the GTX 1060 offer essentially the same performance as the settings for "low" and "medium" quality. With “high”, however, the GTX 1060 pulls forward a few percent and then extends the margin with the ultra preset. It then maintains its 9% lead with the extreme quality settings. The tessellation was enabled for all of these tests, while HairWorks and Advanced PhysX were disabled. We suspect that the level of tessellation increases with the ultra and extreme settings, which gives Nvidia a slight edge.
If you want to enjoy Metro Exodus in all its glory, you need a hell of a gaming rig. To be fair, the game is scaled down to support older or lower hardware well by lowering the quality setting.
If you have less than 4 GB VRAM, we strongly recommend using the medium quality setting. At 1080p with the ultra quality setting, VRAM allocation of around 3.8 GB was frequently achieved for cards with 6-8 GB of memory, slightly more than 4 GB at 1440p and almost 5 GB at 4K.
Of course, we skipped any ray tracing or DLSS tests as we have already covered this. We were quite impressed with the global tracking of lighting beams in Metro, far more than in Battlefield V. The performance hit we tested for this article (39% hit) was more extreme than the one we saw before (29% hit ). With a performance drop of ~ 40%, the RTX 2080 Ti sinks at 1440p from an average of 105 fps to 65 fps, which is not a good thing. While it's a better implementation than what we've seen in Battlefield V, it doesn't say much.
Overall, the game looks great, it's very detailed, and that's why it's so challenging. Both AMD and Nvidia GPUs work well and we were most impressed with how old products like the Radeon HD 7970 work at medium quality.