Make one repeatable comparison
Choose a short route and one fight you can repeat. Record the resolution, graphics preset, upscaling method, frame generation state and observed frame rate. Compare the same scene after each change; an empty safehouse and a busy co-op encounter are different workloads.
This page provides a measurement routine and official settings paths. It does not claim a fixed FPS gain for a GPU we have not benchmarked.
1. Finish shaders before judging a preset
Graphics changes can trigger shader recompilation. The official guide asks players to allow at least 15 minutes of play after changing settings before assessing the result. If every unchanged launch repeats an unusually long shader pass, address the shader-cache problem first.
2. Turn on an appropriate upscaler
Open Options → Visuals → Advanced → Resolution Scaling. Choose a method supported by your setup: DLSS, FSR, TSR or XeSS. 2K recommends trying Performance quality and Frame Generation On as a performance-oriented starting point.
Compare that to a higher-quality setting in the same scene. Keep the setting that balances legibility and responsiveness on your display. A generated-frame counter is not the same measurement as the game’s underlying rendering rate, so record whether frame generation was enabled in both tests. For missing options, use the DLSS diagnostic.
3. Lower the preset, then isolate the expensive option
Start with a lower built-in preset at your normal resolution. If it helps, raise one option at a time while repeating the same route. If it barely changes the result, test a lower render resolution too; this helps distinguish a resolution-sensitive workload from a limit elsewhere.
Keep a small comparison table:
| Test | Resolution / upscale | Preset | Frame generation | Same-scene FPS / visible hitch |
|---|---|---|---|---|
| Baseline | Your current setting | Current | Record On/Off | Record |
| Lower preset | Unchanged | Lower | Unchanged | Record |
| Lower render load | Lower resolution or stronger upscale | Same as prior | Unchanged | Record |
Do not copy a streaming-pool number from another game into an assumed Engine.ini path. A setting that is not documented for the installed BL4 build is not a reliable baseline.
4. Remove background interference for one test
Close third-party overlays, capture software and other demanding applications. Keep the game on an SSD that meets its requirements, with enough free space for updates. Update the GPU driver through its vendor and repeat the route after the new shader pass finishes.
If the result improves only with an overlay closed, re-enable tools one at a time. If it worsens only in a particular mission or after a specific action, include that reproduction in a bug report instead of treating it as a universal graphics setting issue.
5. Compare solo and co-op separately
A delayed network response can feel like stuttering. Compare the same activity solo, then with the same co-op host. If only the online session has the problem, record host/platform/party size and use the connection checklist. Avoid attributing every hitch to shader compilation.
FAQ
Should I change Hardware-Accelerated GPU Scheduling? Follow the official guide for your Windows/GPU setup. It includes HAGS as part of its supported frame-generation troubleshooting; this site does not claim a measured percentage gain.
Will a cap always fix stutter? No. A frame cap cannot repair a crash or a failed connection. If you test a cap, keep other settings unchanged and compare frame pacing as well as average FPS.
Can one preset guarantee 60 FPS? No. Hardware, resolution, scene and update version matter. Keep your comparison results for a useful support report.
Source and reporting a remaining problem
This checklist follows the official PC troubleshooting guide, checked October 18, 2026. It is a diagnostic sequence, not a promise that one step fixes every PC. Record your CPU, GPU, driver, resolution, game version, exact error and the result of each test. Send those details with DxDiag and MSinfo to Borderlands Support if the problem remains.