PC Freezing, Crashing, or Lagging in Games? Check This First

A client called me because Overwatch had become nearly unplayable. The same PC had previously run it at around 240 FPS. Now it was dropping into the 20–30 FPS range, stuttering badly, and sometimes crashing.

The system had an Intel Core i7-7820X, 16 GB of RAM, and an NVIDIA GTX 1080 Ti. The graphics card was old enough to look suspicious. Suspicious is not the same as guilty.

The client had also seen a low GPU percentage in Task Manager and interpreted it as something like "8% GPU life." That is not what the number means. Task Manager reports GPU utilization: how much of the GPU's available work is being used at that moment. Microsoft's explanation of Task Manager's GPU counters describes the percentage as the utilization of the busiest GPU engine. It does not estimate the card's health or remaining lifespan.

It would have been easy to blame the old graphics card, wipe the driver, and hope for the best. I started by making the problem happen again and watching the machine while it happened.

Can a nearly full SSD make games stutter or crash?

Yes. A nearly full Windows drive can contribute to slow performance, stuttering, freezing, and crashes—especially when the computer is also running low on physical memory. It is one possible cause, not a diagnosis by itself.

Windows uses a hidden system file called a page file alongside physical RAM. People often describe it as "backup RAM." That is close enough for a quick explanation, provided you remember that storage is much slower than actual RAM. More precisely, the page file lets Windows move less-active memory pages out of physical RAM, extends the system's committed-memory limit, and can support crash dumps.

Microsoft notes that a system-managed page file needs enough free disk space when it has to grow. If memory demand approaches the system's commit limit, Windows and applications may freeze, crash, or otherwise malfunction. That made the combination on this PC important: memory was near 97% and the C: drive had almost no working room left.

There is no universal Microsoft rule saying every gaming PC must keep exactly 50 GB free. I do want more than the last few gigabytes available on a Windows drive, though. As a practical working margin, I usually free at least several dozen gigabytes before retesting a machine in this condition. That is a troubleshooting rule of thumb, not a Windows specification.

Microsoft provides built-in drive cleanup options under Settings → System → Storage, including Cleanup recommendations and Storage Sense. Review what Windows proposes before deleting it. The Downloads folder and Recycle Bin have a habit of containing things people still wanted.

What I found on this PC

These were the useful observations:

• Overwatch had fallen from roughly 240 FPS to 20–30 FPS.
• The problem could be reproduced instead of occurring randomly once a week.
• System memory use reached about 97% while the problem was happening.
• The main Windows SSD had only around 10 GB of free space.
• HWiNFO did not show an obvious GPU overheating problem.
• Nothing observed at that point proved that the GTX 1080 Ti was failing.

The nearly full drive mattered more because the PC was already under heavy memory pressure. If the page file needed room to grow, Windows had very little space available to give it. Updates, temporary files, logs, and caches were all competing for the same remaining space as well.

That was a much better lead than the age of the graphics card or a misunderstood percentage in Task Manager.

How I diagnose sudden game lag, freezes, and crashes

1. Reproduce the problem first

Open the game and recreate the slowdown if you can. A computer sitting at the desktop tells you very little about what happens under load.

Open Task Manager with Ctrl + Shift + Esc, then watch the Performance tab while the issue occurs. I start with:

• Memory: Is RAM use close to 100%?
• Disk: Is active time pinned near 100%, and which process is using it?
• CPU: Is one process saturating it, or are clock speeds falling under load?
• GPU: Is the 3D engine being used, and are dedicated GPU memory limits being reached?
• Processes: Is a browser, update, antivirus scan, capture tool, or launcher consuming the missing resources?

High GPU utilization during a game is not automatically bad. Low GPU utilization is not proof that the GPU is dying either. These numbers describe workload, not health. Their value comes from reading them together with FPS, temperatures, clocks, memory use, and disk activity.

2. Check free space on the Windows drive

Open File Explorer → This PC and check the C: drive. Even when the game is installed on another drive, Windows normally keeps the page file, temporary data, logs, driver caches, and other system files on C: unless someone deliberately changed the configuration.

If the drive is down to its last few gigabytes, free space before doing anything invasive. Restart the PC and repeat the same test under the same conditions. Changing one thing and retesting is how you learn whether it mattered.

3. Check temperatures and clock speeds under load

I use HWiNFO for detailed sensor readings. Start sensor monitoring, play long enough to reproduce the issue, then check:

• CPU and GPU temperatures
• CPU and GPU clock speeds
• Thermal- or power-throttling indicators
• GPU power draw and utilization
• Physical and committed memory use
• Drive temperature and activity

Idle temperatures are not enough. A graphics card can look perfectly healthy on the desktop and throttle ten minutes into a game. In this case, the readings did not give me a clear overheating or throttling trail back to the GPU.

4. Match crashes to Windows logs

If the game or the whole PC actually crashes, record the exact time. Then check Reliability Monitor and Event Viewer → Windows Logs → System for events at that time.

Do not assume every red icon is the cause. Windows logs a great deal of noise. A useful event repeats with the failure, matches its timestamp, and names a relevant driver, process, hardware error, or unexpected shutdown.

5. Rule out the game and its extras

Before replacing hardware, I would also test the simple software variables:

• Repair the game files through the launcher.
• Reset the game's graphics settings.
• Disable Discord, NVIDIA, Xbox Game Bar, Steam, and recording overlays temporarily.
• Close browsers and other memory-heavy applications.
• Test another game or a repeatable benchmark.

One broken game points you toward that game or its configuration. Several unrelated games failing in the same way points you back toward drivers, Windows, temperatures, power, memory, or hardware.

6. Reinstall the graphics driver when the evidence points there

A clean graphics-driver reinstall is reasonable when crashes name the display driver, the issue began after a driver update, the driver installation appears damaged, or simpler checks have gone nowhere. It is not where I start every performance problem.

If a normal NVIDIA reinstall does not help and the evidence still points toward the driver, I use the more thorough DDU clean-install process. That removes more variables, but it also takes more time and changes more of the system. Measure first.

The fix in this case

I freed a meaningful amount of space on the Windows SSD, restarted the computer, and retested Overwatch. The performance problem improved immediately and the game returned to normal operation. The GTX 1080 Ti did not need to be replaced.

That before-and-after result is why low free space was the strongest diagnosis in this case. The 97% memory use explains why the lack of disk headroom could become especially painful. It does not prove that every dropped frame came from one specific Windows operation.

The PC still had only 16 GB of RAM, so I would keep an eye on memory use as the client's normal workload changes. Freeing space addressed the immediate failure. More RAM may reduce paging in workloads that genuinely exceed 16 GB, but buying RAM before observing the pressure would have been another guess.

What about NVIDIA Shader Cache Size?

I also reviewed the NVIDIA shader-cache setting after the drive had room again. Shader compilation can cause short hitches. NVIDIA's driver stores compiled shaders on disk so it does not have to compile the same ones every time the game runs. Shader Cache Size controls how much disk space the driver may use for that cache; it does not add video memory to the graphics card.

You can find the option under NVIDIA Control Panel → Manage 3D settings → Shader Cache Size. NVIDIA advises choosing a limit large enough for the games you use without exceeding your available drive space.

I would not tell everyone to set it to 100 GB. Bigger is not automatically faster, and changing a cache limit does not fix low RAM, overheating, a bad driver, or a failing SSD. If the Windows drive is nearly full, making another disk cache larger is also a fairly strange opening move.

In this case, freeing storage produced the clear immediate result. The shader cache was a secondary setting to review, not proof of the original cause.

When I would take a possible GPU failure seriously

An older graphics card can fail. I just want evidence before condemning it.

Evidence worth investigating includes:

• Artifacts or corruption across multiple games and benchmarks
• Repeatable driver resets at the same type of GPU load
• Abnormal temperatures, clocks, fan behavior, or power readings
• Crashes that follow the card into another known-good PC
• A known-good graphics card fixing the same workload in the original PC

None of those alone is perfect. Artifacts can come from unstable overclocks, drivers, cables, or video memory. Power problems can look like GPU problems. The point is to build a consistent trail instead of turning one symptom into an expensive conclusion.

The order I would use on another PC

1. Reproduce the exact problem.
2. Watch RAM, CPU, GPU, and disk activity in Task Manager.
3. Check free space on C: and the drive containing the game.
4. Monitor temperatures, clock speeds, and throttling under load.
5. Match actual crashes to Reliability Monitor and Event Viewer.
6. Repair the game, reset its settings, and disable overlays.
7. Test another game or benchmark.
8. Only then move into clean driver reinstalls and hardware substitution.

The order matters. Each step is cheap, observable, and tells you what to do next. Replacing a graphics card is neither cheap nor diagnostic if the problem is still sitting on the C: drive.

The existing PC crash and slowdown diagnostic walks through the same basic process if you want a guided version.

Frequently asked questions

Can a full SSD cause low FPS or stuttering?

Yes, it can contribute—particularly when Windows also needs room for a page file, temporary files, updates, or caches. It is not the only possible cause. Compare performance before and after freeing space under the same workload.

Does 8% GPU usage mean the graphics card has 8% life left?

No. The GPU percentage shown in Task Manager is utilization, not a health score or life-expectancy estimate. It tells you how busy the GPU is at that moment.

Is 100% GPU usage bad?

Not by itself. A game may use nearly all available GPU capacity normally. Look for poor performance, overheating, throttling, artifacts, driver resets, or crashes rather than treating a high utilization number as a fault.

How much free space should a gaming PC keep on C:?

There is no single correct number for every drive and workload. I do not like a Windows gaming PC operating on its last few gigabytes. Freeing several dozen gigabytes is a reasonable first troubleshooting target, but it is a practical margin rather than an official Windows minimum.

Should I increase NVIDIA Shader Cache Size?

Only if the drive has adequate free space and shader-compilation stutter is a plausible issue. The setting limits disk space used by compiled shaders. It does not increase GPU memory or repair unrelated crashes.

Should I reinstall Windows or buy a new GPU?

Not as an opening move. Reproduce the issue, check resource use and free space, monitor temperatures, read the relevant logs, and test simpler software causes first. Reinstall or replace something when the evidence justifies it.