Before you press scan: baseline thermals and drivers
Update to a known-stable driver for your GPU generation, then confirm idle and load temperatures look sane in a light workload. If your GPU is already near thermal limits, Scanner’s proposal may be optimistic because the scan window is not identical to a 45-minute ray-traced session.
What to watch during and after the scan
Watch power limits, hotspot behavior, and fan headroom. If the scan completes but games crash immediately, treat memory clocks separately: VRAM instability often masquerades as “bad scan” when it is actually an offset issue.
Validation order that catches real-world instability
Use a layered approach: a short synthetic load, then a shader-compilation-heavy game, then your normal title mix. If you only run a short test, you may approve a curve that fails under sustained VRAM traffic.
When Scanner is not the right tool
Laptops with aggressive EC throttling, unsupported GPUs, or driver installs that still have remnants from major upgrades can produce odd curves. In those cases, manual VF tuning with smaller steps can be faster than fighting an automated proposal.
Next steps
If your curve is stable, document the settings and export a profile. If unstable, revert one change at a time and read the undervolting article for transient-load pitfalls.
Elsewhere on this site
Keep context when you leave this page
Long guides read best with the FAQ for quick recovery steps and the glossary for vocabulary. The home page links the full screenshot tour and download reminders.