MSI Afterburner is probably the most useful piece of free software in PC gaming, and also the most misunderstood. Most people install it, open the interface once, get confused by the sliders, and close it. They are missing a lot. This guide walks through setup, the monitoring overlay, safe overclocking methodology, and fan curve tuning: basically everything you need to actually use the tool instead of just having it installed.
installation and setup
Download Afterburner from MSI’s official site and run the installer. During setup, it will offer to install RivaTuner Statistics Server alongside it. Install RivaTuner. You need it for the on-screen display (the overlay that shows FPS, temps, and utilization while you are actually inside a game).
After installation, open Afterburner and go to Settings (the gear icon). Under the General tab, confirm that Unified GPU Usage Monitoring is checked. Without this, Afterburner sometimes reads GPU utilization incorrectly, showing low numbers when the card is actually working hard. Check Start with Windows if you want the monitoring overlay to be available without manually launching the app every session.
Next, open the Monitoring tab. This is where you configure what appears in your overlay. Scroll through the list and tick every metric you actually want to watch: GPU temperature, GPU usage, GPU memory usage, CPU temperature, CPU usage, framerate, and framerate 1% low. Highlight each one and check “Show in On-Screen Display” for each. For framerate and 1% low specifically, also check “Show graph” so you get a rolling visual rather than just a number.
Here’s the thing most people miss: the 1% low framerate reading is more useful than average FPS. Average FPS tells you how fast your system runs in ideal conditions. The 1% low tells you how bad the stutters are when the system struggles. A game averaging 120 FPS with a 1% low of 40 FPS is a bad experience. A game averaging 90 FPS with a 1% low of 75 FPS feels smooth. Monitor both.
To activate the overlay inside games, go to RivaTuner Statistics Server and check that Application Detection Level is set to something other than None. The default hotkey for toggling the overlay is Shift+F2 inside Afterburner’s settings (you can change it).
reading the performance graphs
Once you have the overlay running and you are in a game, what you are actually looking for is behavior over time, not single numbers.
GPU usage should sit between 95 and 100 percent when the GPU is your bottleneck. If it is consistently below 80 percent while framerates are also lower than expected, the CPU is likely the limiting factor rather than the graphics card. This matters because the fix is completely different: you cannot solve a CPU bottleneck by adjusting GPU settings.
GPU temperature under full load should generally stay below 85 degrees Celsius on modern cards. Many cards will happily run at 87 or 88 and stay stable, but if you are seeing consistent 90-plus, the fan curve is not working hard enough or there is an airflow issue in the case. We will cover fan curves below.
VRAM usage is the one people suprisingly undervalue. When a game exceeds your card’s available VRAM, performance does not gracefully degrade: it drops hard as the system starts pulling texture data from system RAM across the PCIe bus. Watch this number at your target resolution and settings. Getting close to the limit is a sign to reduce texture quality before anything else.
safe OC methodology
Honestly, this one is where people either see meaningful gains or destroy an afternoon. The correct approach is methodical: one variable at a time, with stability testing between steps.
Start with GPU core clock. In Afterburner, move the Core Clock slider up by 25 MHz and click Apply. Run a GPU-heavy game or benchmark for ten to fifteen minutes. If the game runs without visual artifacts (pixel flickering, texture corruption, driver crashes), the overclock is stable at that level. If it crashes or you see artifacts, reduce by 10 MHz and test again.
Repeat this process in 25 MHz increments until you find the wall. Most Pascal and Ampere cards can handle somewhere between +50 and +150 MHz on the core depending on the individual chip. Turing cards and RDNA 2 and 3 typically respond better to memory overclocking than core. Test both separately.
Memory clock on the GPU is a separate slider. Basically, what’s happening under the hood is that higher memory bandwidth allows the GPU to feed its compute units faster, which matters most in VRAM-heavy scenarios and high-resolution rendering. Start at +200 MHz on memory and test the same way. Memory overclocks tend to have a harder cliff: stable at +500 MHz and then immediately crash at +550 MHz, with little middle ground.
After finding your stable limits for both, set voltage offset only if your card allows it and you understand what you are doing. For most users, leaving voltage on auto and finding the clock limits without touching voltage is the right call. And look, I know this sounds like a lot, but stay with me: once you have a stable profile saved in Afterburner’s five profile slots, you never have to repeat the process. Apply on startup, done.
fan curve optimization
The default fan curve on most cards is too conservative. It keeps the fan quiet at the expense of temperature, which shortens thermal headroom for sustained workloads and boost clocks. Tuning the curve manually is genuinely one of the more impactful changes you can make for consistent performance.
In Afterburner, click the fan speed percentage display and enable the manual fan control toggle, then click the fan curve graph. You are building a curve: temperature on the X axis, fan speed percentage on the Y axis.
A reasonable starting curve for most setups:
- Below 40°C: 30 percent (quiet at idle, card barely working)
- 60°C: 50 percent (starting to do real work, audible ramp begins)
- 70°C: 65 percent (moderate gaming load)
- 80°C: 80 percent (heavy sustained load)
- 85°C: 100 percent (thermal ceiling, maximum cooling)
The goal is keeping the GPU under 80 degrees Celsius during typical gaming without the fans running at 100 percent constantly. If your card hits 80 degrees quickly and stays there, pull the 70-degree point up to 70 percent. If it runs at 65 degrees but sounds like a jet, flatten the middle section of the curve.
common mistakes
Skipping RivaTuner and then wondering why the overlay does not work. Not enabling Unified GPU Usage Monitoring and misdiagnosing a CPU bottleneck as a GPU issue. Moving both core clock and memory clock at the same time and then not knowing which one caused a crash. Setting the fan curve too aggressive (100 percent from 60 degrees) and wearing out the fans for no real temperature benefit.
The other one worth mentioning: overclocking above stable limits expecting benchmark gains, then being surprised when games crash during thermal stress. Benchmarks run for a few minutes. Games run for hours. A stable overclock is the one that survives two hours of demanding gameplay, not five minutes of 3DMark.
Save your stable profile to one of Afterburner’s slots and check “Apply overclocking at system startup” in settings. From there, the tool runs automaticaly in the background and the overlay is available whenever you want it.
Join the Discussion