I'm running an RTX 5070 Ti with a Core i5-13600KF at 1440p. A couple of weeks ago, Plague Tale: Requiem was only reaching around 75 FPS with roughly 90% GPU usage and 180 W of power draw, while similar systems were getting closer to 95 FPS in the same area. After trying several software fixes, I reseated the graphics card and the performance immediately improved to around 95–105 FPS, with full GPU utilization and higher power draw.
I recently saw something similar in Dawnwalker. At 1440p with DLAA and Ultra settings, I was getting only 24–35 FPS instead of the roughly 50 FPS I expected. I used DDU and was considering reinstalling Windows, but reseating the GPU fixed the issue again. Before reseating, I was also seeing occasional WHEA-17 power-related shutdown events, which stopped afterward.
However, shortly after changing the refresh rate from 60 Hz to 144 Hz, the system froze and restarted. I'm trying to figure out what could cause performance to suddenly drop and then return to normal after physically removing and reinstalling the GPU. The system has otherwise worked well since January, including in games where the GPU has reached around 300 W without crashing.
The main hardware concerns I can think of are my Cooler Master MWE V2 Gold 650 W power supply and the H610 motherboard. NVIDIA recommends a higher-wattage PSU for this GPU, and I'm also wondering whether the basic motherboard's power delivery could be overheating or becoming unstable. I already use a GPU support bracket, but I'm considering running the case horizontally to see whether GPU sag or the PCIe connection is involved. What should I test first?
3 Answers
The 650 W PSU is below NVIDIA’s typical recommendation for this class of GPU, so it shouldn’t be dismissed, especially if the card has transient power spikes. Also verify that the GPU power cable is fully inserted and isn’t connected through an inappropriate daisy-chain arrangement. The H610 board could contribute too, particularly if its VRM cooling is minimal.
The H610 motherboard is a fairly basic platform for a 13600KF. If the board has weak or poorly cooled VRMs, the CPU could throttle under load or the system could become unstable. Use HWiNFO in sensor-only mode, run a CPU-heavy workload, and monitor VRM temperatures, CPU clocks, power limits, and throttling flags. Check those readings while the GPU performance is poor as well.
Even if the game looks GPU-limited, a throttling CPU or unstable motherboard can still affect frame pacing, GPU utilization, and system stability. VRM temperatures and clock behavior would help confirm or rule this out.
Because reseating the card fixes the problem more than once, I’d investigate the physical PCIe connection rather than immediately reinstalling Windows. Inspect the slot and GPU contacts for dust or damage, make sure the card is fully level and secured to the case, and check whether the support bracket is putting upward pressure on the card. Also test the system with the case horizontal and, if available, try another PCIe slot or another known-good PSU.
A loose connection or slight card sag could explain reduced link quality or intermittent instability, although you should confirm the PCIe link width and speed in GPU-Z or HWiNFO instead of relying only on benchmark results.
Even if the system has worked for months, that doesn’t completely rule out power delivery or connection problems. Intermittent behavior can appear only under certain loads or after thermal expansion and contraction.