If a game starts smoothly and becomes slower after extended play, keep the graphics settings fixed while you investigate. The useful question is what changes over time: the game’s own background work, another app, memory demand or the hardware’s operating conditions.
Unchanged graphics do not guarantee unchanged performance
Vanguard Princess provides a concrete example: its Controller and Performance Update addresses a background controller-detection problem that caused gradual slowdown during longer play. That identifies a game-side fault worth checking for before reducing image quality. It does not establish that controllers cause long-session FPS drops in other games.
Check your game’s current patch notes and installed build first. If a relevant fix is available, record the build change and repeat your usual long session. Avoid promising yourself a particular FPS gain: the useful result is whether the deterioration still occurs on your PC.
Compare the same route early and late
This is a suggested diagnostic method, not an HD OPTI benchmark. A busy late-game battle and a quiet opening area are different workloads. Choose a short repeatable route or built-in benchmark that you can revisit within the same running session.
- Record the game build, graphics driver, resolution, preset, frame cap, upscaling and Frame Generation state. Keep power settings and laptop plugged-in state consistent.
- Let the game complete any shader preparation it requests. Note prior gameplay and warm-up time, then record an early capture after reaching normal gameplay.
- Continue playing until the usual slowdown appears. Repeat the same route with the same capture duration and settings, without restarting the game first.
- Use a frame-time recorder such as PresentMon, or the game’s own permitted performance tools. Keep the tool version and selected metric unchanged; label displayed and presented FPS separately if both are recorded.
- Note elapsed play time and capture average FPS, frame-time traces and the same low-FPS statistic if available. Alongside them, record process CPU activity, memory demand, temperatures and clock speeds using tools that expose those readings.
- Repeat the session before deciding a pattern is reliable. Keep first-run preparation separate, and record the spread of results rather than only the best capture.
Look for longer frames becoming more frequent or a sustained slowdown on the repeated route. A whole-session average can hide when the problem started. Treat unavailable sensor readings as missing data.
High FPS but Game Still Stutters? How to Fix Frame Time Spikes
Match the slowdown to a trend
Monitoring provides clues. A matching trend helps you choose the next test; it does not prove which component caused the slowdown.
| Repeated observation | What to investigate | Useful next check |
|---|---|---|
| FPS falls while temperatures rise and clocks fall | Thermal or power limits | Check available limit indicators and the hardware maker’s cooling guidance |
| One process grows busier as the session continues | Game or app background work | Compare the same session with one optional app closed |
| Process memory keeps rising across repeated routes | Increasing allocation demand | Track the process over time and check system memory headroom |
| Only one game deteriorates, then recovers after relaunch | Session-dependent game behaviour | Repeat on the current build and prepare a bug report |
| The later capture uses a busier area | A changed workload | Revisit the original route before diagnosing deterioration |
Read temperature and frequency together
A high temperature alone is not a diagnosis. Intel processors can adjust frequency and power as part of thermal protection. Compare temperature, clocks and any available thermal-limit indicator at the same workload. A frequency drop can also have other explanations; do not apply one temperature threshold to every CPU or GPU.
Separate memory growth from a confirmed leak
For a closer process-level check, Windows Performance Monitor can track Process > Private Bytes and Process > Virtual Bytes. Private Bytes excludes memory shared with other processes; Virtual Bytes measures virtual address space. Some leaks appear as growth in these counters, but memory pools can take time to settle. A rising graph is a reason to investigate, not a confirmed leak.
In Task Manager > Performance > Memory, record Available and Committed alongside the symptom. Committed demand approaching its limit deserves attention: that limit is supported by RAM and page files, and exhausted commit capacity can prevent allocations. Keep the page-file configuration unchanged during this comparison.
Windows 11 26H2 Memory Changes: What to Check on an 8GB Gaming PC
Isolate one background app or controller condition
Press Ctrl + Shift + Esc to open Task Manager and identify apps using resources during the slowdown. Close one recognised, non-essential app normally, then repeat the long-session test. Reopen it for a further comparison. Keep required game services and security protection running.
If closing that app repeatedly helps, investigate its settings or updates. Closing it changes several resources at once, so the result does not identify CPU or RAM as the sole cause. Avoid changing multiple overlays, graphics options and startup settings together.
If controller detection is a plausible lead, compare separate fresh sessions with your usual controller connected and with it disconnected before launch, using keyboard input where the game supports it. Keep the route and session length comparable, then restore the original setup and repeat. This tests an association with the device setup; it does not reveal the game’s internal polling mechanism.
Record the controller model, connection type, input mode and Steam Input state. Change only one of those conditions in a follow-up test. Do not unplug your required input device during a competitive match to gather evidence.
Use restart behaviour as a clue
Once you have saved the late-session evidence, relaunch the game and revisit the route. Record whether performance recovers and how quickly deterioration returns. A relaunch changes process state and may give hardware time to cool, so recovery alone cannot distinguish a leak, device handling or heat.
For a useful bug report, include the build, hardware, settings, elapsed time, controller setup, early and late captures, and the steps that reproduce the problem. For Vanguard Princess, the update asks for controller model and mode plus display.log from the install folder’s Logs directory. In Steam, use Manage > Browse local files to locate it; check folder paths before sharing logs.
If the pattern remains inconclusive, preserve the baseline and broaden the investigation. The next useful change is one whose effect you can compare against the same long-session symptom.
FAQ
Graphics settings do not control every part of the workload. Game-side background work can change over time, as the Vanguard Princess controller-detection fix illustrates. Background apps, memory demand and thermal behaviour are other leads to test against the same repeatable route.
No. Process memory can take time to settle, and a rising counter alone does not establish a leak. Track the same process over repeated workloads, compare the timing with the slowdown and preserve the evidence for further investigation.
No. Relaunching changes the game’s process state and can also allow cooling time. Save the late-session readings first, then compare the same route after relaunch and record how long any recovery lasts.
Use that as a controlled comparison only when device handling is a plausible lead and the game supports another input method. Start separate sessions with and without the controller and repeat the original setup. A fix in one game does not establish a universal controller problem.
