PC Performance and Graphics Settings

Match the published PC target before tuning

The minimum target is 1080p at 30 Hz on Low with performance super sampling. The recommended target is 1080p at 60 Hz on High with quality super sampling. Use the complete matching tier, not one isolated GPU label, and do not diagnose a machine against an unstated resolution, refresh target, or preset.

The minimum record includes Windows 10 64-bit, an Intel Core i5-6600 or Ryzen 3 1200, 12 GB RAM, one listed minimum GPU, DirectX 12, broadband internet, and 35 GB storage. The recommended record raises the CPU, memory, and GPU fields. Open the PC requirements guide for every exact model before deciding which published baseline applies.

Build one repeatable local baseline

Record the operating system, CPU, installed memory, GPU, driver version, display resolution, graphics preset, super-sampling mode, and the exact scene used for comparison. Close unrelated heavy workloads through their normal controls, restart the game when the driver or client changed, and return to the same route or practice condition. This is a local test setup, not a benchmark for other hardware.

Describe the symptom precisely: consistently low frame behavior, short hitching, delayed input, visual corruption, a process crash, or a network message. Only the first two belong directly to a performance comparison. Send a process termination to the crash guide and a connection message to the network guides so one visible slowdown is not made to explain every technical symptom.

Change one graphics family at a time

Start from the chosen published preset and change one family, such as the overall preset or super-sampling choice, before repeating the same scene. Keep resolution fixed during that comparison. If the observed frame behavior improves without a new visual or input problem, preserve the exact change; if it does not, restore the prior value before testing another boundary.

The official optimization material discusses demanding gore, weapon-impact, VFX, shadow, lighting, model-detail, and distance workloads. That list identifies areas the developer worked on; it does not rank every menu option or state which single setting controls a current slowdown. Use the labels actually visible in the installed build, and do not invent a hidden toggle or numeric cost.

Treat the Nvidia driver note as dated evidence

The launch-window issue post said some Nvidia hitching appeared to respond to a graphics-driver update. Confirm the GPU vendor and current driver, use the vendor's normal update path, restart the PC, and test the original scene without changing other graphics settings. Preserve both driver versions and the observed result.

If the hitching changes, the comparison supports the driver boundary for that machine and scene. If it does not, the result does not prove the official note wrong or justify cycling through unrelated unofficial packages. A driver observation also does not establish the cause of a crash, network interruption, or server-side delay.

Separate rendering symptoms from service conditions

Frame pacing and local image quality can be observed with a controlled scene. A queue delay, party failure, or lost-connection message involves different systems and should be recorded with time, region, and squad scope. Lowering graphics during a wider service incident cannot demonstrate that the network condition improved, even if a later session happens to connect.

Check the newest official feed when the problem began after a patch. Compare the current game version and publication date before treating an older optimization note as current. This page has no live hardware telemetry beyond the values the reader records and no server telemetry at all.

Keep measured results separate from unsupported promises

For every accepted change, record the same scene before and after and label the observed result. State whether the symptom improved, remained, or changed form. A result on one map, resolution, driver, and PC does not promise a frame rate on different hardware, and the published targets do not guarantee every combat moment will equal the target.

Stop when the controlled changes no longer isolate a variable. Preserve the baseline, the smallest changed setting, and the repeatable scene for escalation. Do not delete unrelated settings, claim an optimal universal preset, invent performance percentages, or present a local lab observation as current field data for all players.