I'm troubleshooting a major performance drop after changing my memory settings. My system has an Intel i5-12600K, RTX 4070 Ti Super, Gigabyte B660 DS3H AX DDR4 motherboard, and 32GB of G.Skill Ripjaws DDR4-3600 running as a 2x16GB kit in slots 2 and 4.
With XMP enabled at 3600MHz, I was getting roughly 100–120+ FPS in games such as Marvel Rivals and Deadlock, but I started experiencing MEMORY_MANAGEMENT blue screens, especially in Unreal Engine 5 games. I changed the memory speed to 3200MHz in the BIOS, which stopped the crashes and passed a short TM5 test, but gaming performance dropped dramatically. I now get around 40–60 FPS, occasional dips to about 10 FPS, and only 35–50% GPU usage. CPU usage is above 75% across the P-cores.
The GPU still boosts to approximately 2835MHz, and GPU-Z reports PCIe 4.0 x16 under load. All P-core threads are active, so the system does not appear to be parking cores or using only the E-cores. I also re-enabled Resizable BAR and Above 4G Decoding after the BIOS reset, disabled CSM, returned my GPU settings to stock, and cleared the game configuration files, but none of that changed performance.
The system worked normally with this CPU and board for several years. The memory is relatively new, and I previously had instability when mixing a 2x16GB kit with a 2x32GB kit, so I'm now using only the 32GB kit. Could a BIOS reset or memory retraining have changed a CPU power, boost, or chipset setting? Should I reset the CMOS, update or restore the BIOS, run DDU, check CPU temperatures and boost clocks, or test another memory kit?
4 Answers
A 200MHz memory reduction should not cause a 50% FPS loss by itself. I’d first load BIOS defaults, update the BIOS if needed, and test with everything stock. Then check CPU temperatures, effective clocks, and whether the processor is hitting its expected boost frequency. A CPU power-limit or thermal-throttling problem could explain low GPU usage much better than the memory speed change.
Run a CPU benchmark such as Cinebench and compare the score and clock behavior with results from other i5-12600K systems. Monitor package temperature, effective P-core clocks, power consumption, and throttling flags with a tool such as HWiNFO. If the score is abnormally low, investigate BIOS power limits, cooler mounting, CPU temperatures, and chipset or motherboard settings before focusing on the GPU.
The MEMORY_MANAGEMENT crashes should also be investigated instead of assuming they are only caused by the board’s memory layout. Check the Windows minidump files for the actual faulting driver or module, run a longer memory test, and test the two DIMMs individually if possible. Ten minutes of TM5 is useful as a quick check but is not enough to prove the kit is stable.
Try a proper CMOS reset, then configure only the basic settings and test again. Confirm that the memory is installed in the recommended slots, use one known-good kit, and avoid mixing kits. If performance remains poor at BIOS defaults, testing another memory kit or another CPU-compatible board would help determine whether the DIMMs, motherboard, or processor’s memory controller is involved.
The games mentioned can be CPU-limited, but the size of this performance drop still points to another setting or hardware issue rather than DDR4-3200 alone.