I'm trying to determine whether a component in my PC is defective because I get severe stuttering after installing NVIDIA driver updates or rebuilding the shader cache. My system has an Intel i5-14400F, RTX 5060 Ti 8GB, ASUS Prime B760M-A motherboard, 32GB of G.Skill DDR5, and a Kingston SNVS250G NVMe boot drive. I tested the memory at 7600, 6400, 6000, 5600 CL36, and the default 4800 MT/s.
Games including Overwatch, Valorant, and Elden Ring stutter heavily during the first several minutes after a driver update, then usually become much smoother. In Elden Ring, I have occasionally seen FPS fall to roughly 17–30, GPU utilization drop very low, and the screen briefly go black or blink as though the NVIDIA driver restarted.
MSI Afterburner and HWiNFO show frametime spikes over 100 ms. This is not simply low overall FPS: GPU usage falls sharply while individual CPU threads become heavily loaded, even though total CPU usage remains well below 100%. Overwatch shows a similar pattern immediately after driver updates.
I have already tried DDU and a clean driver installation, reinstalling Windows, updating the BIOS and chipset/Intel ME drivers, running SFC, disabling E-cores, and using both 5600 CL36 and 4800 MT/s memory. None of those changes fixed the issue. Task Manager also sometimes reports the Kingston NVMe at 100% active time despite only about 86 KB/s of reads, 1 MB/s of writes, and roughly 22 ms response time.
Does this sound like normal shader compilation and cache rebuilding, or could the CPU, GPU, RAM, or SSD be faulty? I'm trying to identify the actual failing component before replacing hardware.
1 Answer
The stuttering during the first few runs after a driver update is consistent with shader compilation and cache rebuilding. Games may compile shaders as new effects, areas, or materials appear, which can cause temporary frametime spikes and thread-specific CPU activity even when total CPU usage is low. It should generally improve after the relevant shaders have been compiled and cached.
However, a black-screen blink followed by apparent driver recovery is not something to dismiss as ordinary shader compilation. That could indicate a driver timeout or instability, so it is worth checking Windows Event Viewer for display-driver errors and testing with a known-stable driver version. Also check the GPU power connections, temperatures, GPU stability, and the SSD’s health and firmware. The NVMe showing 100% active time at a low transfer rate can happen during small random I/O, but repeated high latency or drive-health warnings would justify testing the drive separately.
I’m familiar with shader-compilation stutter, but the driver recovery and black-screen behavior seem less normal, especially if the previous system did not show them. I would investigate those symptoms separately rather than assuming every problem is caused by shader caching.