ONLYBOOST
1. CPU and GPU temperatures
Every processor only works well within certain thermal limits. When the CPU exceeds 85-90 °C, thermal throttling kicks in: the processor slows itself down to avoid damage. Result: fewer FPS and unexplained stuttering.
To monitor temperatures, use HWInfo in "Sensor Only" mode. Note the idle values (desktop, no games running), then run a stress test with OCCT or play for a few minutes and check the maximum peaks.
If the CPU exceeds 90 °C without overclocking:
- Check that the cooler is mounted correctly and all fans are spinning
- Clean the dust inside the case (especially the heatsink fins)
- Replace the thermal paste if you haven't changed it in a while
- If you have an entry-level cooler, consider an upgrade
Note for Intel 12th/13th/14th gen (i7 and i9)
Peaks above 90 °C under load are normal for these processors. To lower them, you can reduce the Power Limit in BIOS (make sure you have the latest BIOS version installed).
Also check GPU temperatures in HWInfo. If they're high, try increasing fan speed with MSI Afterburner, or check that dust isn't blocking airflow.
2. GPU PCI-E interface
GPU performance also depends on the PCI-Express mode detected by the motherboard. Open GPU-Z and look at "Bus Interface": the value before the @ is the expected mode, the one after the @ is the current one. They should match.
If the GPU shows x8 instead of x16:
- Update the BIOS and check the PCI-E configuration
- Verify that NVMe or SATA SSDs are not sharing bandwidth with the GPU
- Make sure the GPU is in the slot closest to the CPU (the first x16 slot)
- If you use a vertical mount, check that the riser cable supports the same PCI-E generation
The number after "x16" (e.g. 4.0) indicates the generation. At idle you might see "1.0" — that's normal, it goes up to 3.0 or 4.0 under load. To verify, click the "?" in GPU-Z and run the "Render Test": if the value goes up, everything is fine.
3. Proper GPU power delivery
Stable power delivery is essential, especially with NVIDIA 3000/4000 series and AMD 6000/7000 series GPUs.
Do not use a single daisy-chained PCI-E cable
Connect a separate PCI-E cable from the PSU for each connector on the GPU. A daisy-chained cable doesn't guarantee enough current and causes stuttering or sudden crashes.
For NVIDIA 4000 series GPUs, verify that the 12VHPWR connector is firmly seated and compatible with your power supply.
4. Correct RAM slots
If you have two RAM sticks, they should be in slots 2 and 4 (counting from the left) to enable Dual Channel, which doubles memory bandwidth and significantly improves performance.
You can verify the configuration in Task Manager → Performance → Memory or by physically checking the motherboard.
If you have more than two sticks or a motherboard with a non-standard layout, consult your motherboard manual for the recommended installation scheme.
Want me to check everything directly on your PC?
Book a session