PC Optimization Guide: 15 Tweaks That Actually Raise FPS
Most optimization guides are a landfill of registry hacks copied from a 2011 forum post. This one is different: every tweak below was tested on our mid-range bench — a Ryzen 7 7700, 32 GB of DDR5-6000 and an 8 GB mid-tier GPU — across five games, three repeatable runs each. If a change didn't move average FPS or 1% lows beyond run-to-run noise, it didn't make the list. Where the gain is small, we say so.
Windows-level tweaks
1. Turn on Game Mode. Settings → Gaming → Game Mode. It suppresses background tasks and Windows Update activity while you play. On our bench it bought nothing in averages but steadied 1% lows in CPU-heavy titles like Baldur's Gate 3 — about 4% better lows. Free stability; leave it on.
2. Enable Hardware-Accelerated GPU Scheduling (HAGS). Settings → Display → Graphics → Change default graphics settings. On modern NVIDIA and AMD drivers this moves scheduling work off the CPU. We measured a consistent 2–5% uplift in GPU-bound scenarios at 1440p. Requires a reboot, costs nothing.
3. Set the power plan to High Performance while gaming. The Balanced plan parks cores and throttles clock ramp-up. On AMD systems in particular, High Performance (or the Ryzen chipset plan) recovered 3–7% in CPU-limited titles and cut frame-time spikes after alt-tabbing. Switch back when you stop playing if idle power draw matters — but don't game on Balanced.
4. Fullscreen optimizations — mostly a myth. "Disable fullscreen optimizations" on every .exe was useful on early Windows 10 builds. In 2026 the flip-flop presentation model is equal or better, and disabling it breaks borderless alt-tab speed in several engines. Our testing: zero gain in four of five games, a small loss in the fifth. Stop doing this.
Driver-level tweaks
5. Do a clean driver install when you upgrade. Not every time — just when you jump GPU generations or chase a specific game fix. Use the installer's "clean install" checkbox; DDU is only for when something is actually broken. We've chased more than one "bad driver" bug that was really three stacked years of leftover profiles.
6. Size your shader cache properly. If you rotate through several big DX12 titles, raising NVIDIA's shader cache (or leaving AMD on Driver Default, which manages it sensibly) eliminates recompilation stutter in games like Cyberpunk 2077 and Hogwarts Legacy. A smoothness tweak, not an FPS tweak — and it matters more than most realize.
7. Low-latency modes. NVIDIA Reflex (in-game, where supported) is a genuine improvement: enable it. The driver-level Ultra Low Latency toggle is the legacy fallback and does nothing in DX12 games that manage their own queue; AMD Anti-Lag follows the same logic. Use the in-game option unless the game predates Reflex.
In-game settings, ranked by cost per pixel
Settings are where the real gains live. Ranked by how much FPS they cost relative to what they add visually:
8. Shadows: drop one tier, not to Low. Ultra-to-High shadows bought us 12–18% across the suite with a difference you have to freeze-frame to see. High-to-Low gains almost nothing more but looks obviously worse. One tier down is the sweet spot.
9. Volumetric fog and clouds: the silent FPS eater. In Cyberpunk, RDR2 and most UE5 titles, Ultra volumetrics cost 15–20% for atmosphere you'll stop noticing after ten minutes. Medium volumetrics is our single most-recommended change.
10. Ray tracing: be honest about your card. Full RT costs 30–50% on anything below the high end; path tracing is enthusiast-only. On 8–12 GB of VRAM, RT High with upscaling is the ceiling. Our take: reflections are the one RT effect worth paying for; RT shadows are not.
11. Upscaling: use it, in this order. DLSS Quality, then FSR Quality, then XeSS. At 1440p, DLSS Quality recovered 25–35% with image quality close enough to native that we stopped A/B testing it. Avoid Performance mode at 1080p — text and foliage smear.
Overlay and background hygiene
12. Kill the overlays you don't use. Every running overlay (Discord, Steam, GeForce, Xbox Game Bar, EA App) hooks the render pipeline. One is harmless; four stacked cost us 3–6% and introduced random hitching. Keep the one you actually use.
13. Close the browser, or at least the tab hoard. Chrome with 40 tabs held 6 GB of RAM and periodic CPU bursts on our bench. On a 16 GB system running a game that wants 12, that's the difference between smooth and stutter.
14. Update your chipset drivers and BIOS. Boring, but AMD's chipset drivers control core parking and boost behavior. We've measured 5%+ swings in CPU-bound games from a stale BIOS. Do it once a year.
15. Cap your frame rate below your monitor's refresh. If you can't hold a locked 144, cap at 120. A stable capped rate with consistent frame times feels faster than an uncapped rate oscillating between 90 and 144 — and 1% lows improve measurably.
Summary table
| Tweak | Expected gain | Effort |
|---|---|---|
| Hardware-Accelerated GPU Scheduling | 2–5% avg FPS | 1 minute + reboot |
| High Performance power plan | 3–7% (CPU-bound) | 1 minute |
| Shadows down one tier | 12–18% | 30 seconds per game |
| Volumetrics to Medium | 15–20% (heavy titles) | 30 seconds per game |
| DLSS/FSR Quality upscaling | 25–35% at 1440p | 30 seconds per game |
| Ray tracing off / reflections only | 30–50% (mid-range cards) | 30 seconds per game |
| Trim overlays to one | 3–6% + less hitching | 5 minutes |
| Chipset drivers + BIOS update | Up to 5% (CPU-bound) | 30 minutes, yearly |
| Frame-rate cap below refresh | Better 1% lows, consistency | 2 minutes |
Five popular tweaks that do nothing
Registry "FPS optimizations." There is no magic key. SystemResponsiveness, GPU priority entries, TCP ack tweaks — every one we tested landed within run-to-run noise.
Third-party "booster" and "game mode" apps. They close background apps you could close yourself, then sit resident consuming resources. Two triggered anti-cheat warnings in our testing. Hard pass.
Disabling the pagefile. This doesn't free performance — it removes a safety net. Several modern games commit more memory than physical RAM and will crash or stutter without a pagefile, especially on 16 GB systems. Leave it system-managed.
Setting process priority to "High" or "Realtime." Priority changes don't survive reboots, and Realtime can starve input and audio drivers, causing the stutter you were trying to fix. Windows' scheduler is better at this than you are.
Debloating Windows to the bone. Removing Defender and half the system services via scripts gained us 0–1% and broke Windows Update on one bench install. The idle CPU those services use doesn't show up in a GPU-bound game.
The pattern: real gains come from GPU workload (settings, upscaling), scheduling (power plan, shader cache, background hygiene), and consistency (frame caps). Anything promising 40% from a registry file is lying to you.