How to Check Your PC for a Bottleneck

Stop guessing and start measuring; here is exactly how to check bottleneck with free tools. A bottleneck is not a permanent property of your hardware; it is a snapshot of a specific game, at a specific resolution, with specific settings. The question is never whether one exists, because one always does. You can use this exact test method with free tools to see the truth for yourself.

You need MSI Afterburner with RivaTuner Statistics Server. You will also want a frame time logger like PresentMon, which writes a CSV file with milliseconds per frame, one row per frame, so you can see stutter that average FPS hides.

Set up the overlay before you open a game. For each sensor you want to see, click it, then check the box for "Show in On-Screen Display." The critical sensors for a bottleneck check are: GPU usage (%), CPU usage per logical core (%), GPU temperature (°C), GPU clock (MHz), CPU clock (MHz), and frame time (ms). Add the frame time graph, not just the FPS counter, because that distinction matters. A GPU can be 100% busy and still deliver bad frame pacing.

Windows Task Manager is not enough. Use Afterburner for the real data, and use Task Manager only for a quick sanity check when you are not in a game.

Launch the most demanding game you own, preferably one that pushes your system hard. Set the resolution to your native monitor resolution and the graphics preset to the highest setting that stays above 30 FPS in the worst area you know. Play for at least five minutes in a scene that is actually demanding, not a menu and not a cutscene. Open the Afterburner overlay with Ctrl+O and watch the GPU usage number.

The rule is simple: if the GPU usage is at 95% or higher, your graphics card is doing its job and the bottleneck is almost certainly the CPU, the RAM, or the game engine. If the GPU usage is below 95% in a demanding scene, something else is limiting the frame rate. That "something else" is usually the CPU, but it can also be a thermal throttle, a driver bug, or a game-specific optimization problem. The GPU usage number is the single most important diagnostic.

Do not trust the first minute of a game. If the GPU usage climbs above 95% after the initial load, your system is healthy and the game is just loading.

A 16-core CPU showing 40% total usage can still have one core at 100% and that single core is your limit. Games have a primary thread that cannot be parallelized, so the ceiling on that thread is the CPU's single-core speed, not its core count. While you play, watch for any single core pegged at 100%. If one core is at 100% and the others are idle, that is a CPU bottleneck, regardless of what the total CPU usage says. If all cores are below 80% and the GPU is below 95%, the limit is elsewhere, likely RAM bandwidth, a driver bug, or the game engine itself.

This is where the term how to tell if cpu is bottlenecking actually gets its answer: you do not look at the total CPU number, you look at the hottest single core. A game that uses two threads will bottleneck on a CPU with weak single-core performance even if that CPU has 16 cores. More cores do not fix a single-thread limit. The fix for a CPU bottleneck is a faster CPU, not a better GPU, and no amount of GPU upgrade will raise your average FPS if a single core is the wall.

Set the game to a lower resolution, like 720p, and lower the graphics preset to the minimum. Keep the same scene. If the FPS climbs, the GPU was the limit. If the FPS stays the same, you gave the GPU less work and the frame rate did not move because the CPU cannot feed frames any faster. This is the correct way to run a bottleneck test, and it is the only test that isolates the CPU from the GPU without guesswork.

Ignore the 1% low for the pass/fail call; if the 1% low stays low, the CPU is the problem. Run the test twice in each mode to confirm the numbers are stable, because a single run can be thrown off by a background Windows update.

Use this table to match what you see in the overlay to the likely cause. If the GPU usage is above 95% and frame times are flat, the system is healthy. If the GPU usage is below 95% and one core is at 100%, it is a CPU bottleneck. If the GPU usage is below 95% and all cores are below 80%, suspect RAM, driver, or software. If the CPU or GPU clock is dropping under load, check temperatures. In every case, the resolution-drop result confirms the diagnosis.

MSI Afterburner's overlay reports sensors at specific resolutions, and knowing what each number means prevents misreading. The CPU usage sensor reports a percentage per logical core, and the overlay graphs each one in 1% increments. The GPU clock sensor reports in 1 MHz increments, so a 2000 MHz reading is accurate to within 1 MHz. The temperature sensors report in 0.1°C resolution, so you can see thermal throttling start at 83°C or 91°C, depending on the card's factory limit.

The memory usage sensor reports in 1 MB increments, which matters when you are close to the VRAM limit; if you exceed it, the card may be swapping to system RAM, which shows up as a stutter. The fan speed sensor reports in 1% increments, and the power draw reports in 1% of the card's TDP. No calculator can know your background tasks, your cooling solution, your driver version, or the specific game you play. A site that claims your CPU is 20% bottlenecked at 1080p is guessing, because the same CPU-GPU pair can be GPU-bound in one game and CPU-bound in another. The research that would validate a calculator's numbers, a controlled test of 20 games on 100 different CPU-GPU pairings, has never been published by any calculator site.

The only honest answer to "is my CPU bottlenecking my GPU" is "run these three tests and see." The resolution-drop test, the per-core CPU check, and the GPU usage observation together give you a definitive answer for your exact system. Ignore any calculator that gives you a single percentage; that number means nothing. It is not an FPS loss, because no calculator can predict that. It is a synthetic score derived from the site's own weighting of component tiers, and it has no predictive power for any real game. Trust the overlay, not the percentage.

A CPU bottleneck can be caused by software, not just hardware. Background apps like a browser with 50 tabs, a streaming encoder running x264 instead of NVENC, or a Windows update scanning your drive can all steal CPU cycles and make your GPU look underutilized. To test for this, do a clean boot. Press Win+R, type msconfig, go to the Services tab, check "Hide all Microsoft services," then click "Disable all." Open Task Manager, go to the Startup tab, and disable everything. Restart your PC, launch the game, and re-run the resolution-drop test.

If the GPU usage climbs above 95% after a clean boot when it was below 95% before, you had a software bottleneck, not a hardware one. The fix is to re-enable services one by one until the problem returns, then either disable that service or find an alternative. The hardware encoder on a modern GPU has a negligible performance cost, and it frees the CPU for game threads.

Do not skip this step. The clean boot test takes 10 minutes and saves you from a $400 upgrade that does not fix the problem.

Thermal throttling looks exactly like a CPU bottleneck in the overlay, but the fix is completely different. If your CPU runs at 95°C and drops from 5.0 GHz to 3.0 GHz, the GPU usage will fall below 95% because the CPU cannot feed it fast enough. The per-core usage will show a single core at 100%, but that 100% is at a reduced clock speed. The solution is not a new CPU; it is a better cooler, new thermal paste, or better case airflow.

Check the CPU temperature sensor in the Afterburner overlay while you play. If it is above 90°C and the clock speed is lower than the boost spec, you are thermal throttling. The same applies to the GPU: if the GPU temperature is above 85°C and the clock speed is lower than the boost clock, the card is hitting its thermal limit. A GPU that throttles will show GPU usage at 99% but frame rates lower than a card with the same chip that runs cooler.

Thermal paste that is years old can dry out and lose its effectiveness, causing a 5-year-old CPU to run hot. Reapplying thermal paste is a $10 fix that can recover 10-15% performance on a throttling chip. Do not assume a high temperature reading is a hardware failure; check whether the cooler fan is spinning at 100% speed in the overlay. A fan that is stuck at 30% speed when the CPU is at 90°C is a fan problem, not a CPU problem.

A game that runs at 60 FPS average can still stutter if the 1% low is 20 FPS, and that stutter is what you feel. The 1% low is the average frame time of the worst 1% of frames; it captures the absolute worst frame times, and a CPU bottleneck will make it far worse than the average.

To measure this, enable the frame time graph in the Afterburner overlay. A flat line means consistent frame pacing; a spiky line means stutter. If the spikes correlate with a single CPU core at 100%, the CPU is the cause. You can import the CSV into Excel and calculate the 1% low and 0.1% low yourself. How do I know if my stutter is a CPU bottleneck or a game bug? Run the clean boot test. If the GPU usage stays below 95% and one core is at 100% in the Afterburner overlay, it is a CPU bottleneck. A calculator's 20% figure is an artifact of its own weighting, not a prediction of any game's performance.

3. What is the actual FPS loss from a specific CPU-GPU pairing? No one can answer this without testing your exact system, your cooling, your background tasks, and the specific game version. The honest answer is to run the three tests: the resolution-drop test identifies the limiting component, the per-core CPU check determines if it is a single-thread limit, and the clean boot test reveals whether software is the culprit. No calculator can substitute for those measurements.

No test is a substitute for a controlled benchmark.

A system that is GPU-bound in one game can be CPU-bound in another, and a driver update can shift the balance by 5% overnight. The method here gives you a definitive answer for the specific game you test, but it cannot tell you how your PC will perform in a game you have not tried. The only way to know is to test each demanding game you play, and to re-test after major driver or game updates.

If you have followed every step and the GPU usage stays below 95% while all cores are below 80%, the bottleneck is likely software or a game-specific issue. In that case, check the game's forums for known issues, verify the game files, and try a different driver version. A 1% low that is far below the average with no thermal or core issue often points to a game-specific problem that no hardware upgrade will fix.