I'm troubleshooting a strange performance problem with an RTX 5060 Ti 8GB while playing Hogwarts Legacy and Expedition 33. My system has a Ryzen 5 7500F, 32GB of RAM, an XPG Core Reactor II 850W Gold PSU, Windows 11, two monitors, a USB headset and microphone, a wireless gamepad dongle, and an Ethernet connection. I'm using NVIDIA driver 610.88. When I'm not streaming or sharing my screen, performance is normal. However, Discord screen sharing—and previously TikTok Live Studio and Streamlabs—can cause the game to drop to extremely low FPS with major frametime spikes. GPU utilization remains close to 99%, VRAM usage is around 7GB, but GPU power consumption suddenly falls. 720p30 is usually fine initially, while 720p60 and 1080p30 stutter more noticeably, and 1080p60 is the worst. Closing Discord immediately restores normal performance. I've already tried clean driver installations with DDU, removing NVIDIA HD Audio, default GPU settings, several undervolt and clock configurations, different PBO Curve Optimizer values, Frame Generation on and off, HAGS, disabling Discord hardware acceleration and Clips, and multiple capture resolutions and frame rates. LatencyMon showed roughly 4.7 ms of ISR time from Wdf01000.sys and about 24 ms of DPC time from nvlddmkm.sys during a particularly bad episode. Could this be an interaction between the NVIDIA driver, capture, and encoding pipeline rather than a hardware, VRAM, CPU tuning, or undervolt problem?
1 Answer
One possibility is a data-transfer bottleneck that becomes more noticeable when the game and capture software are both moving large amounts of data. Asset streaming in Unreal Engine games can already stress the CPU, storage, memory, and PCIe path during area transitions; adding recording or screen capture may expose the problem. That could explain why the issue appears more often while entering new areas and why GPU usage stays high even though power and frame rates fall. Your TeamGroup MP44L NVMe drive and motherboard would be worth checking, along with PCIe link speed, chipset drivers, storage health, BIOS settings, and background disk activity. However, this explanation is not conclusive—capture-driver or encoder scheduling conflicts can produce similar symptoms, so testing with a different capture method and monitoring CPU, SSD, PCIe, and encoder activity would help distinguish between them.
The stutters do seem more common during asset or area transitions in both games. I’m using a TeamGroup MP44L 1TB NVMe drive and a Colorful Battle-AX B650M-PLUS V14 motherboard. I’ll check the PCIe configuration, BIOS, chipset drivers, and storage activity, although I’m still wondering whether upgrading the CPU or changing capture settings would be the more likely fix.