This could potentially impact performance and visuals more on some of the other levels, but given we measured a 1–3% drop in performance, you can probably leave it at ultra. Set this to low, and the puddles don't have any approximated reflections. Post-process quality didn't impact performance much at all, but looking at the screenshots, this appears to include effects like screen space reflections. Dropping it to low boosted performance by about 10%. However, it's not a massive visual difference, and outside of ray tracing, this was the single most demanding setting. Flipping between low and ultra, there's less detail and deformation in the terrain on the low setting. Terrain quality seems to alter both the resolution of the terrain textures as well as the amount of terrain geometry. Part of that might be the CPU becoming a bottleneck with the faster RX 6700 XT, however. ultra improved performance by just 5% on the AMD card and 11% on the Nvidia GPU. Shadows on objects farther away may disappear entirely, like on the semi-truck dumpster in the left of the image. At the low setting, the resolution of the shadow maps drops and so shadows-like of the nearby ranger robot-become less distinct. Lighting quality generally deals with the shadows you'll see in the game. As far as performance goes, it's again those same settings that have the greatest potential to improve framerates-or hurt them in the case of RTAO. Oh, and of course, ray-traced ambient occlusion, but that really does need a potent GPU and/or running at lower resolutions (perhaps with DLSS) to be viable. You can flip through the above images to make your own evaluation, but the short summary of the biggest impact settings is that only lighting quality, terrain quality, and post-process quality change the visuals much. Even so, the margin of error is going to be slightly higher than the normal 1–2% range we'd get from a good built-in benchmark. We also ran the game in solo/co-op mode and played against bots because we didn't want other humans wondering why one player kept running in circles around one area of the map.
That means that besides trying to avoid heavy clouds in the sky (which drops performance), or if a tornado shows up, we exit and restart. We've attempted to mitigate that variability by running each test multiple times and also tried to get a "clear" day for the battle on the Discarded map. Weather effects can also impact performance, and just the general mayhem of a 64-player or even 128-player match means there's far more variability between runs. Of course, there isn't a built-in benchmark, which means we had to run every test manually-good for getting a true look at performance, bad for repeatability.
CONFIG PC BATTLEFIELD 4 HOW TO
Benchmarking Battlefield 2042 is, frankly, horrible-this is pretty much the opposite of most of the things we like to see, as we discussed in how to make a good game benchmark. Our test equipment consists of a Core i9-9900K CPU with 32GB of DDR4-3600 CL16 memory, a 2TB SSD, and of course, the various GPUs we've tested.