I'm trying to diagnose random but very noticeable frametime spikes on my Windows 11 PC. The system has a Ryzen 7 7800X3D, RTX 4070 ROG Strix OC, 32GB of DDR5-6000, and a Gigabyte X870E AORUS Elite WiFi 7 motherboard. Average FPS is generally high and overall 1% lows are reasonable, but I periodically get severe hitches lasting around 20–60 ms. PresentMon shows that the spikes are mostly CPU-side. For example, a Valheim hitch measured 60.87 ms frametime, 60.80 ms CPU Busy, 0.10 ms CPU Wait, 7.20 ms GPU Busy, and 54.19 ms GPU Wait. RDR2 showed a similar result: 53.52 ms frametime, 53.06 ms CPU Busy, 0.46 ms CPU Wait, 9.06 ms GPU Busy, and 42.98 ms GPU Wait. It looks like the CPU or game thread stalls while the GPU sits idle waiting for it. I've already disabled overlays, removed GPU tuning and overclocks, and updated my drivers. Oddly, heavier games such as Cyberpunk with path tracing can sometimes run more smoothly than lighter games. Could this be caused by a background process, interrupt or DPC latency, Windows scheduling, storage activity, or some other system-level issue?
3 Answers
The measurements point more toward a CPU-side interruption or stall than a GPU performance problem. Since the same behavior appears in unrelated games, I’d check interrupt and DPC latency while reproducing the hitch. A tool such as LatencyMon can help identify a driver, but Windows security or driver restrictions can sometimes prevent it from running. You can also monitor DPC and ISR execution with Windows Performance Recorder/Analyzer or another hardware monitoring utility. Look for a device driver or background service that spikes at the exact time the frametime jumps.
A recent driver or Windows update is worth treating as a possible trigger, especially since this only started after a year or two of normal behavior. Check what changed during the last week, then test by clean-installing the graphics driver or rolling back to the previous known-good version. Also temporarily disable unnecessary startup apps, hardware-monitoring utilities, RGB controllers, and motherboard vendor services. Those can interrupt the game thread even when CPU usage does not look high overall.
That timing is what makes me suspicious of a recent NVIDIA or Windows update. The games were stable until roughly the last week, so I’ll compare recent changes and test a clean driver install while checking background utilities.
Changing the number of active CPU cores is not a good first diagnosis, but it can be a temporary experiment if you suspect a scheduling or firmware issue. Before doing that, check BIOS defaults, chipset drivers, power settings, memory stability, and whether the system is logging WHEA or other hardware errors. If the hitch disappears when background services are stopped in a clean boot, re-enable them one at a time to find the offender. The fact that the GPU is mostly waiting means lowering graphics settings probably won’t address the root cause.
I’m going to investigate interrupts and DPCs next. I tried removing RGB-related software without success, and the same GPU-waiting-for-CPU pattern occurs in Subnautica and even occasionally in Rocket League. LatencyMon wouldn’t run because of a driver or security issue, so I’m looking for another reliable monitoring method.