Quick Answer
Install MSI Afterburner with RivaTuner Statistics Server, enable GPU usage, per-core CPU usage, FPS, and frametime in the on-screen overlay, then play a demanding scene for 15 minutes. GPU usage at 95 to 100 percent with CPU headroom means a GPU limit. CPU cores pinned near 100 percent with the GPU idle means a CPU limit.
Introduction
You have probably asked yourself a version of this question: is my CPU holding back my graphics card, or the other way around? Online calculators can estimate an answer, but the only way to know what is actually happening inside your PC is to measure it while a real game runs. Running a proper msi afterburner bottleneck test is how you get that real answer, because Afterburner is free, works with graphics cards from every brand, and shows the exact numbers on screen while you play.
This guide walks you through the whole method, from a clean install to a confident diagnosis in about 15 minutes. You will set up the Afterburner overlay step by step, learn how to read the usage numbers and the frametime graph, and learn the misreads that fool most beginners. By the end, you will be able to look at your own numbers and say with confidence which part of your PC is the limit, and what to do about it.
If you want the bigger picture first, our broader walkthrough covers eight different ways to find a bottleneck, from quick software checks to hardware swaps. In short, it frames bottleneck hunting as a process of elimination across the whole system, while this article goes deep on the single most reliable method: measuring with Afterburner. You can read that overview here: How to Find a PC Bottleneck: 8 Diagnostic Tests That Actually Work.
 Why MSI Afterburner Is the Right Tool for Bottleneck Testing
A bottleneck test is only as good as the numbers behind it. Guessing from FPS alone tells you that performance is low, but not why. MSI Afterburner solves that by recording what each component is doing, frame by frame, while the game runs. The program was built by MSI for graphics card overclocking and monitoring, it is completely free, and according to MSI it works with graphics cards from any brand, not only MSI cards. That last point matters, because plenty of monitoring advice assumes you own specific hardware. Afterburner does not care whose logo is on your GPU.
The key companion is RivaTuner Statistics Server, usually called RTSS. This is the small program that draws the on-screen display, the overlay of numbers you see in the corner of your screen while gaming. The Afterburner installer offers to install RTSS at the same time, and for bottleneck testing you want to accept that offer. Without RTSS, you can only watch the graphs on a second monitor or after the fact. With it, the data is right in front of you during real gameplay, which is exactly when it matters.
One safety note before you download anything. MSI explicitly warns that there are phishing sites offering modified Afterburner installers that can steal your data. MSI says the genuine program exists only on msi.com and Guru3D. Download it from MSI’s official Afterburner page and nowhere else. This matters for bottleneck testing too, because a tampered monitoring tool could feed you wrong numbers.
Install Afterburner and RivaTuner Statistics Server Correctly
Run the installer you downloaded from MSI. When the component selection screen appears, make sure both MSI Afterburner and RivaTuner Statistics Server are checked. Some people skip RTSS thinking it is only for overclockers. For our purpose, RTSS is the more important of the two, because it renders the overlay.
After installation, launch Afterburner once and let it detect your hardware. You will see the familiar main window with sliders for core clock, memory clock, power limit, and temperature limit, plus a hardware monitor pane. Here is the part beginners often misunderstand: you do not need to touch any overclocking slider for bottleneck testing. Leave every slider at its default. We are using Afterburner purely as a measuring instrument, like a thermometer. Touching the clocks would change the results of the test.
Practical takeaway: install both Afterburner and RTSS from MSI’s official page, launch once to confirm your GPU is detected, and do not change any overclocking setting.
Choose the Exact Metrics to Monitor
Open Afterburner and click the gear icon to open Settings. Go to the Monitoring tab. This is the control room for the whole test. You will see a long list titled “Active hardware monitoring graphs.” Each item can be shown as a graph in Afterburner’s monitor window, and more importantly, each can be sent to the on-screen display by checking the box “Show in On-Screen Display” below the list.
For a bottleneck test, enable exactly these metrics. Fewer and you will miss the diagnosis. More and the overlay becomes unreadable clutter.
| Metric | What it reveals | How to enable |
|---|---|---|
| GPU usage | Whether the graphics card is the limiting factor; 95 to 100 percent under load is the classic GPU-bound signature | Tick it, then check “Show in On-Screen Display” |
| CPU usage (per core) | Which individual cores are maxed out; one pinned core can bottleneck a game even when total CPU usage looks low | Tick each CPU1, CPU2, CPU3 entry up to your core count |
| Framerate | Your average FPS; the baseline number everything else is judged against | Tick it, then check “Show in On-Screen Display” |
| Frametime | How long each frame takes in milliseconds; reveals stutter that average FPS hides | Tick it, then check “Show in On-Screen Display” |
| GPU temperature | Whether the card is thermal throttling, which looks like a bottleneck but is not one | Tick it, then check “Show in On-Screen Display” |
| GPU power | Whether the card is hitting its power limit instead of a true hardware limit | Tick it, then check “Show in On-Screen Display” |
| CPU temperature | Whether the CPU is throttling its clocks under load | Tick it, then check “Show in On-Screen Display” |
| CPU clock | Confirms the CPU is actually boosting to its rated speed during the test | Tick it, then check “Show in On-Screen Display” |
| RAM usage | Catches the rare case where running out of system memory is the real problem | Tick it, then check “Show in On-Screen Display” |
The per-core CPU entries are the ones most people skip, and they are the most important detail in this entire setup. A game that uses four threads will max out four cores on a 16-core CPU while overall CPU usage sits near 25 percent. If you only watch the total, you will wrongly conclude the CPU has plenty of room. Tick every individual core entry your CPU has.
Practical takeaway: in the Monitoring tab, enable GPU usage, every per-core CPU entry, framerate, frametime, both temperatures, GPU power, CPU clock, and RAM usage, and send each one to the on-screen display.
How to Set Up the Overlay for Bottleneck Testing
With the metrics chosen, the overlay itself needs configuration. In the Monitoring tab, for each metric you ticked, check the “Show in On-Screen Display” box. Select the metric, check the box, move to the next one. When you are done, click Apply, then OK.
Now launch RTSS. It usually starts automatically with Afterburner, and you will find its icon in the system tray. Open it and check a few things. First, confirm that “Show On-Screen Display” is set to On. Second, look at the “On-Screen Display zoom” and “On-Screen Display palette” options if the text is too small or hard to read against bright games. Third, and this catches people often, make sure “Application detection level” is set to a value that actually hooks into your game. The default works for most titles, but if the overlay never appears, raising the detection level one step is the standard fix.
Two more settings make the test far easier. In Afterburner’s Monitoring tab, the “Hardware polling period” controls how often the numbers update. The default of 1000 milliseconds means the overlay refreshes once per second, which is readable and perfectly fine for a bottleneck test. Leave it there. Then, still in the Monitoring tab, look at the “Hardware monitoring history logging properties” section at the bottom and check “Log history to file.” Afterburner will write a continuous log of every metric to a file on your drive while you play. This gives you a permanent record you can scroll through after the test, which is much better than trying to memorize numbers mid-firefight.
Practical takeaway: check “Show in On-Screen Display” for every chosen metric, confirm RTSS is running with the overlay enabled, keep the default one-second polling period, and turn on logging to file so you have a record to review afterward.
How to Read CPU-Bound and GPU-Bound Signatures
This is the payoff for all the setup. The overlay is live, you are in a game, and numbers are moving. Here is how to read them.
Think of the CPU and GPU as a two-person assembly line. The CPU prepares each frame’s instructions, and the GPU draws it. Whichever worker is fully occupied while the other waits around is your bottleneck. Afterburner shows you exactly who is waiting.
| What you see | What it means | Typical numbers |
|---|---|---|
| GPU usage steady at 95 to 100 percent, CPU usage moderate, FPS stable | GPU-bound, the healthy and normal state for most gaming PCs | GPU 97 percent, CPU 40 to 60 percent, per-core usage even |
| One or more CPU cores pinned at 90 to 100 percent, GPU usage well below 90 percent, FPS lower than expected | CPU bottleneck; the processor cannot feed the graphics card fast enough | Two cores at 98 percent, GPU at 55 percent, overall CPU only 35 percent |
| CPU cores pinned and GPU usage very low across every game | Severe CPU bottleneck; a much stronger CPU or a settings change is needed | CPU cores at 100 percent, GPU at 40 percent, the classic forum question |
| GPU usage at 100 percent, CPU cores also high, FPS exactly at your monitor’s refresh rate | No bottleneck to worry about; the system is simply fast enough to max out the display | GPU 99 percent, CPU 70 percent, FPS locked at 144 on a 144 Hz monitor |
| Both GPU and CPU usage oddly low while FPS is poor | Something else is wrong; suspect a frame cap, thermal throttling, or a driver issue | GPU 45 percent, CPU 30 percent, FPS 40; see the misreads section below |
The row that confuses people most is the second one. A 12-core CPU can show 35 percent total usage while two cores sit at 98 percent, and that is a genuine CPU bottleneck in any game that leans on a few main threads. This is why the per-core view from Step 2 is non-negotiable. Total CPU percentage is an average across all cores, and averages hide the one core that is drowning.
A real example from a Quora question illustrates the pattern perfectly. A user reported their CPU usage at nearly 100 percent in every game while their GPU sat around 40 percent, asking whether it was a bottleneck with a Ryzen 3 3200G and a GTX 1650. The answer was yes, a clear CPU bottleneck. That 4-core, 4-thread processor simply could not prepare frames fast enough for the graphics card in CPU-heavy titles, so the GPU spent much of its time waiting. When you see that same shape in your own overlay, you are looking at the same situation.
Practical takeaway: GPU pinned near 100 percent with CPU headroom means the GPU is the limit, and that is fine. CPU cores pinned near 100 percent with the GPU waiting means a CPU bottleneck. Always judge the CPU by per-core usage, never by the total.
How to Read the Frametime Graph
Average FPS tells you how many frames you got per second. It does not tell you whether they arrived evenly. Frametime is the time each individual frame took, measured in milliseconds, and it is the metric that separates a smooth 60 FPS from a stuttery 60 FPS.
In the overlay, the frametime entry shows the current frame’s time in milliseconds. In Afterburner’s monitor window, and in RTSS with the graph enabled, you can watch it as a rolling line. A healthy system shows a fairly flat line, each frame taking roughly the same time. A system with stutter shows spikes, sudden tall peaks where one frame took far longer than its neighbors.
One more thing the frametime graph reveals is shader compilation stutter. The first time a modern game loads a new area, the CPU may pause to compile shaders, producing tall isolated spikes that vanish on later runs. This is not a hardware bottleneck at all, and it is worth knowing about because Microsoft is actively attacking this exact problem. In October 2026, Microsoft began expanding its Advanced Shader Delivery feature to all NVIDIA RTX GPUs on Windows 11 version 24H2 and newer, delivering precompiled shaders with supported games like Gears of War: E-Day and cutting first-launch shader compilation time by roughly 95 percent according to Microsoft’s own figures. If your frametime graph is full of spikes during the first ten minutes of a new game but smooth afterward, you are likely seeing shader compilation, not a CPU bottleneck. Test again after the shaders are cached before drawing conclusions.
Practical takeaway: watch the frametime line for spikes, not just the FPS number. Repeating spikes under load point at the CPU. Spikes only during a game’s first run point at shader compilation, which is temporary.
The 15-Minute Bottleneck Test Routine
Setup is done, and you know how to read the results. Now run a proper test. Fifteen minutes sounds short, but it is enough if the minutes are structured. Randomly wandering through a game for an hour gives you noisier data than a focused quarter hour.
Follow these steps in order:
1. Close background programs. Browsers with video tabs, Discord streaming, and recording software all consume CPU and will pollute the test. Close them or at least note that they were open. 2. Launch Afterburner and RTSS, and confirm the overlay appears. If you use “Log history to file,” clear the old history by right-clicking the monitor window and choosing to clear it, so the new log starts fresh. 3. Launch your game and load into actual gameplay, not a menu. Menus barely load the hardware and will give you fake low readings. This is one of the most common testing mistakes. 4. Play a demanding scene for five minutes. Busy combat, a crowded city, or a large multiplayer battle. Note the GPU usage, the highest per-core CPU usage, the FPS, and whether the frametime line is flat or spiky. 5. Change to a second area or scene type and play another five minutes. An indoor area versus an outdoor area, or a quiet section versus combat. Bottlenecks can shift with the scene, and one scene is not a full test. 6. Run the confirmation test. Drop the game’s resolution one step, for example from 1440p to 1080p, without changing anything else, and play two more minutes. If FPS jumps up noticeably, the GPU was the limit. If FPS barely moves, the CPU was the limit, because lightening the GPU’s workload changed nothing about what the processor can produce. 7. Alt-tab out and review the logged graphs. Scroll through the history and look at the shapes, not single moments. A reading that lasts three seconds is noise. A pattern that holds for minutes is a diagnosis.
If you want a second opinion without extra software, the resolution test in step 6 is also the classic low-tech check. It is worth repeating because it is so decisive: same settings, lower resolution, watch the FPS. Big jump means GPU-bound. No movement means CPU-bound.
Practical takeaway: close background apps, test real gameplay in two different scenes for ten minutes, then run the resolution-drop confirmation test for the final verdict.
7 Misreads That Fool Beginners
Most “bottleneck” panics on forums turn out to be one of these seven misreads. Check each one before you spend money on an upgrade.
1. A frame cap is not a bottleneck. If V-Sync is on, or the game has an FPS limit, or you set a frame limit in RTSS itself, the GPU will sit well below 100 percent because the system does not need more frames. This is the speed-limit effect: once you hit 144 FPS on a 144 Hz monitor, the hardware stops pushing. Low GPU usage with FPS exactly at your cap is a healthy system, not a bottleneck. Diagnose only with caps turned off.
2. Testing in menus or loading screens. Hardware barely works in a menu, so usage numbers collapse. Always measure during real gameplay. Several Tom’s Hardware forum threads show users confused by low readings taken in menus, and the answer is always the same: test under real load.
3. First-run shader compilation stutter. Tall frametime spikes during a game’s first minutes look like a CPU problem. As covered above, they usually vanish once shaders are cached, and Microsoft’s Advanced Shader Delivery rollout is specifically designed to reduce this. Retest after the first run before blaming your CPU.
4. Thermal throttling disguised as a bottleneck. If the GPU temperature climbs past its throttle point, the card lowers its clocks and usage drops, which looks like the CPU is holding it back. This is why GPU temperature is in your metric list. Our explainer on the topic puts it simply: thermal throttling is the card protecting itself by slowing down, it shows up as falling clocks alongside high temperatures, and the fix is cooling, not a new CPU. The full breakdown is here: GPU Thermal Throttling Explained.
5. Background apps eating the CPU. An antivirus scan, a browser with dozens of tabs, or a stream running in the background can pin CPU cores that the game needs. Check the per-core graph: if non-game cores are busy while game cores wait, the bottleneck is your multitasking, not your hardware.
6. Overall CPU percentage hiding pinned cores. Worth repeating because it fools almost everyone once. A 16-core CPU at 30 percent total can still be CPU-bottlenecked if the game’s main threads have two cores at 100 percent. Always read the per-core entries, never the average.
7. Testing only one game. Different games stress different parts. An open-world RPG may be CPU-heavy while a cinematic shooter at 4K is GPU-heavy on the same PC. Test at least two or three games from different genres before declaring a verdict. Our guide to the difference puts it this way: CPU-intensive games lean on simulation, AI, and draw calls, GPU-intensive games lean on resolution and visual effects, and knowing which type you are testing explains why your numbers change between titles. Read it here: CPU-Intensive vs GPU-Intensive Games.
Practical takeaway: before blaming hardware, rule out frame caps, menus, shader stutter, heat, background apps, misleading averages, and single-game results.
What Trending News Tells You About Your Readings
Bottleneck testing does not happen in a vacuum. Three current stories from October 2026 change how you should interpret what Afterburner shows you.
Monitoring matters more than ever on hot-running flagships. A Reddit modder known as DallasGrave recently soldered a custom power distribution board onto an RTX 5090, adding two conventional 8-pin PCIe inputs alongside the card’s 16-pin connector. Running the card at a sustained 550W for 48 hours, the modder reported wire temperatures never exceeding 40 degrees Celsius, with roughly 3 amps per conductor. The story is a perfect real-world advertisement for monitoring: GPU power and GPU temperature are in your metric list precisely so you can see whether your own card is running within comfortable limits or flirting with trouble. Coverage of the mod is on Guru3D. You do not need to copy the mod, which voids warranties and requires soldering skill. The lesson is simpler: watch your power and temperature graphs, because they tell you things usage percentages cannot.
Frame generation moves the bottleneck to the CPU. NVIDIA’s DLSS 4 Multi Frame Generation can generate up to three additional AI frames for every traditionally rendered frame on RTX 50 series cards, multiplying displayed frame rates several times over. Here is what that means for your test: frame generation multiplies the base frame rate your CPU can deliver. If your CPU caps the base game at 45 FPS, generated frames make it look like 180 FPS on screen, but the responsiveness and the underlying limit still belong to the CPU. When testing with frame generation on, judge the bottleneck by the base rendered frame rate and the CPU graphs, not the final displayed number. NVIDIA documents the technology in detail on its DLSS 4 page. For diagnosis purposes, consider running your test with frame generation off first, then on, and compare.
Shader stutter is getting a platform-level fix. As discussed in the frametime section, Microsoft’s Advanced Shader Delivery is expanding to all NVIDIA RTX GPUs on Windows 11 24H2 and newer this month, shipping precompiled shaders with supported games to cut first-launch compilation and in-game shader stutter. For your testing, the practical point stands: frametime spikes during a game’s first session are increasingly a solved software problem, not a verdict on your CPU.
Practical takeaway: watch power and temperature on high-wattage cards, test with frame generation off for the cleanest diagnosis, and do not mistake first-run shader stutter for a CPU bottleneck.
What to Do After the Test Finds Your Limit
A diagnosis is only useful if it leads to a decision. Here is how to act on each outcome.
If the test shows a GPU limit, that is the normal, healthy state for a gaming PC. The graphics card is doing exactly what you bought it for. The next step depends on your goals: if the FPS is fine, do nothing and enjoy the game. If you want more performance, the levers are lower settings, lower resolution, or upscaling technologies like DLSS and FSR, which shift work off the GPU. For a structured way to think about what counts as acceptable, our guide on good bottleneck percentages gives you the numbers to aim for: in short, a small measured bottleneck is normal and expected, and only large, consistent imbalances deserve action. Read it here: What Is a Good Bottleneck Percentage.
If the test shows a CPU limit, start with the free fixes. Close background apps, lower CPU-heavy settings like draw distance, crowd density, and physics, and make sure your RAM is running at its rated speed with XMP or EXPO enabled in the BIOS, since slow memory can starve a CPU. If the numbers still show pinned cores after that, the honest answer is a CPU upgrade, and pairing advice matters here. Two warning signs that specifically point at a CPU holding back a graphics card are covered in our signs guide, which lists the observable symptoms, from low GPU usage to stuttering in busy scenes, so you can cross-check your overlay readings against real-world behavior: 10 Signs Your CPU Is Bottlenecking Your GPU.
If the test shows neither component maxed out but FPS is poor, revisit the misreads section. The usual suspects are a hidden frame cap, thermal throttling, a driver problem, or the game itself being poorly optimized. Our broader low-FPS troubleshooting guide walks through the full checklist beyond the overlay numbers. And if you want to compare your results against other measurement tools, our roundup of free benchmarking software lists the best companions to Afterburner, including options that specialize in frametime capture and 1% low analysis: Best Free Benchmark Software.
Practical takeaway: GPU-bound with good FPS means enjoy your PC. CPU-bound means try the free fixes first, then consider an upgrade. Neither-bound with bad FPS means hunt for caps, heat, or driver issues.
FAQs
My CPU usage is almost 100% and my GPU is around 40% in every game. Is it a bottleneck?
Yes, that is the textbook shape of a CPU bottleneck, and it is one of the most asked questions on Quora. When the processor is saturated while the graphics card waits at low usage in game after game, the CPU cannot prepare frames fast enough for the GPU to draw. Confirm it with the resolution test from the routine above: drop the resolution one step, and if FPS barely moves, the CPU is confirmed as the limit. The fixes are closing background apps, lowering CPU-heavy settings, and, if the pattern persists, a CPU upgrade.
How much CPU usage is normal for gaming?
There is no single normal number, which is why this Quora question gets asked so often. In well-balanced systems, total CPU usage of 30 to 70 percent with the GPU near 90 to 100 percent is typical. The number that actually matters is per-core usage. If any single core sits near 100 percent during gameplay while the GPU waits, you have a CPU limit regardless of what the total says. Judge the CPU core by core, not by the average.
How do I detect a CPU bottleneck with MSI Afterburner?
Enable GPU usage and per-core CPU usage in the Afterburner overlay, play a demanding scene for at least ten minutes, and watch the pair. GPU usage consistently near 95 to 100 percent means the GPU is the limit and the CPU is keeping up. One or more CPU cores pinned near 100 percent while GPU usage stays well below 90 percent means the CPU is bottlenecking the GPU. That is the entire test in one paragraph.
Does low GPU usage always mean a CPU bottleneck?
No, and this is the misread that generates half the confusion threads on Tom’s Hardware forums. Low GPU usage with FPS exactly at a frame cap, V-Sync limit, or your monitor’s refresh rate is completely normal. Low GPU usage in menus is normal. Low GPU usage with high GPU temperatures means thermal throttling. Only low GPU usage during uncapped, demanding gameplay, with CPU cores pinned, means a CPU bottleneck.
Should I enable every monitoring metric in Afterburner?
No. Enable the nine metrics in this guide’s table and send them to the overlay. Enabling everything turns the overlay into unreadable clutter and makes the diagnosis harder, not easier. If you want deeper data later, use the log file Afterburner writes to disk, which records every metric for after-the-fact analysis.
Can MSI Afterburner itself lower my FPS?
The monitoring overlay has a very small performance cost, and for bottleneck testing it is negligible compared to the value of the data. If you are chasing maximum benchmark numbers rather than diagnosing, you can close the overlay and rely on the log file instead. Do not let a one or two percent measurement cost stop you from measuring.
Key Takeaways
- Install Afterburner and RTSS only from msi.com or Guru3D, leave all overclocking sliders at default, and use the program purely as a measuring tool.
- Enable GPU usage, per-core CPU usage, framerate, frametime, both temperatures, GPU power, CPU clock, and RAM usage in the overlay.
- GPU usage at 95 to 100 percent with CPU headroom is the healthy GPU-bound state. CPU cores pinned near 100 percent with the GPU waiting is a CPU bottleneck.
- Judge the CPU core by core. Total CPU percentage is an average that hides pinned cores.
- Read the frametime graph, not just FPS. Repeating spikes under load point at the CPU. First-run spikes point at shader compilation.
- Run the 15-minute routine: ten minutes across two scenes, then the resolution-drop confirmation test.
- Rule out the seven misreads, frame caps, menus, shader stutter, heat, background apps, misleading averages, and single-game tests, before spending money.
- Test with frame generation off for the cleanest diagnosis, and watch power and temperature graphs on high-wattage cards.
Conclusion
The msi afterburner bottleneck test turns bottleneck hunting from guesswork into measurement. Fifteen focused minutes with the right overlay tells you more about your PC than any online calculator, because you are watching your actual hardware handle your actual games. Set up the metrics once, learn the two signatures, and you will never again have to wonder whether your CPU or your GPU is the limit.
Run the test tonight on the games you actually play. Write down the numbers. Then, whether the answer is a comfortable GPU limit, a CPU asking for help, or one of the classic misreads, you will know exactly what your next move should be. That confidence is the whole point of measuring instead of guessing.
Leave a Reply