Details of x11-drivers/nvidia-drivers:

Currently loaded NVIDIA modules do not match the newly installed
libraries and may prevent launching GPU-accelerated applications.
Easiest way to fix this is normally to reboot. If still run into issues
(e.g. API mismatch messages in the \`dmesg\` output), please verify
that the running kernel is ${KV_FULL} and that (if used) the
initramfs does not include NVIDIA modules (or at least, not old ones).
Detected that the current kernel command line is using 'slub_debug=',
this may lead to system instability/freezes with this version of
${PN}. Bug: https://bugs.gentoo.org/796329
***WARNING***
You are installing a version of ${PN} known not to work
with a GPU of the current system. If unwanted, add the mask:
  echo '${NV_LEGACY_MASK}' > ${EROOT}/etc/portage/package.mask/${PN}
  echo '${NV_LEGACY_MASK}' >> ${EROOT}/etc/portage/package.mask
...then downgrade to a legacy[1] branch if possible (not all old versions
are available or fully functional, may need to consider nouveau[2]).
[1] https://www.nvidia.com/object/IO_32667.html
[2] https://wiki.gentoo.org/wiki/Nouveau
Open source variant of ${PN} was selected, note that it requires
Turing/Ampere+ GPUs (aka GTX 1650+). Try disabling if run into issues.
Also see: ${EROOT}/usr/share/doc/${PF}/html/kernel_open.html
!!! WARNING !!!
Detected HandleNvidiaSleep=yes in ${EROOT}/etc/elogind/sleep.conf.d/.
This 'could' cause issues if used in combination with the new hook
installed by the ebuild to handle sleep using the official upstream
script. It is recommended to disable the option.
!!! WARNING !!!
Detected a custom script at ${EROOT}{/etc,{/usr,}/lib*}/elogind/system-sleep
referencing NVIDIA. This version of ${PN} has installed its own
hook at ${EROOT}/usr/lib/elogind/system-sleep/nvidia and it is recommended
to remove the custom one to avoid potential issues.
Feel free to ignore this warning if you know the other NVIDIA-related
scripts can be used together. The warning will be removed in the future.

Currently loaded NVIDIA modules do not match the newly installed
libraries and may prevent launching GPU-accelerated applications.
Easiest way to fix this is normally to reboot. If still run into issues
(e.g. API mismatch messages in the \`dmesg\` output), please verify
that the running kernel is ${KV_FULL} and that (if used) the
initramfs does not include NVIDIA modules (or at least, not old ones).
Detected that the current kernel command line is using 'slub_debug=',
this may lead to system instability/freezes with this version of
${PN}. Bug: https://bugs.gentoo.org/796329
***WARNING***
You are installing a version of ${PN} known not to work
with a GPU of the current system. If unwanted, add the mask:
  echo '${NV_LEGACY_MASK}' > ${EROOT}/etc/portage/package.mask/${PN}
  echo '${NV_LEGACY_MASK}' >> ${EROOT}/etc/portage/package.mask
...then downgrade to a legacy[1] branch if possible (not all old versions
are available or fully functional, may need to consider nouveau[2]).
[1] https://www.nvidia.com/object/IO_32667.html
[2] https://wiki.gentoo.org/wiki/Nouveau
Open source variant of ${PN} was selected, be warned it is experimental
and only for modern GPUs (e.g. GTX 1650+). Try to disable if run into issues.
Please also see: ${EROOT}/usr/share/doc/${PF}/html/kernel_open.html
!!! WARNING !!!
Detected HandleNvidiaSleep=yes in ${EROOT}/etc/elogind/sleep.conf.d/.
This 'could' cause issues if used in combination with the new hook
installed by the ebuild to handle sleep using the official upstream
script. It is recommended to disable the option.
!!! WARNING !!!
Detected a custom script at ${EROOT}{/etc,{/usr,}/lib*}/elogind/system-sleep
referencing NVIDIA. This version of ${PN} has installed its own
hook at ${EROOT}/usr/lib/elogind/system-sleep/nvidia and it is recommended
to remove the custom one to avoid potential issues.
Feel free to ignore this warning if you know the other NVIDIA-related
scripts can be used together. The warning will be removed in the future.

Currently loaded NVIDIA modules do not match the newly installed
libraries and may prevent launching GPU-accelerated applications.
Easiest way to fix this is normally to reboot. If still run into issues
(e.g. API mismatch messages in the \`dmesg\` output), please verify
that the running kernel is ${KV_FULL} and that (if used) the
initramfs does not include NVIDIA modules (or at least, not old ones).
Detected that the current kernel command line is using 'slub_debug=',
this may lead to system instability/freezes with this version of
${PN}. Bug: https://bugs.gentoo.org/796329
***WARNING***
You are installing a version of ${PN} known not to work
with a GPU of the current system. If unwanted, add the mask:
  echo '${NV_LEGACY_MASK}' > ${EROOT}/etc/portage/package.mask/${PN}
  echo '${NV_LEGACY_MASK}' >> ${EROOT}/etc/portage/package.mask
...then downgrade to a legacy[1] branch if possible (not all old versions
are available or fully functional, may need to consider nouveau[2]).
[1] https://www.nvidia.com/object/IO_32667.html
[2] https://wiki.gentoo.org/wiki/Nouveau
Open source variant of ${PN} was selected, be warned it is experimental
and only for modern GPUs (e.g. GTX 1650+). Try to disable if run into issues.
Please also see: ${EROOT}/usr/share/doc/${PF}/html/kernel_open.html
!!! WARNING !!!
Detected HandleNvidiaSleep=yes in ${EROOT}/etc/elogind/sleep.conf.d/.
This 'could' cause issues if used in combination with the new hook
installed by the ebuild to handle sleep using the official upstream
script. It is recommended to disable the option.
!!! WARNING !!!
Detected a custom script at ${EROOT}{/etc,{/usr,}/lib*}/elogind/system-sleep
referencing NVIDIA. This version of ${PN} has installed its own
hook at ${EROOT}/usr/lib/elogind/system-sleep/nvidia and it is recommended
to remove the custom one to avoid potential issues.
Feel free to ignore this warning if you know the other NVIDIA-related
scripts can be used together. The warning will be removed in the future.

Currently loaded NVIDIA modules do not match the newly installed
libraries and may prevent launching GPU-accelerated applications.
Easiest way to fix this is normally to reboot. If still run into issues
(e.g. API mismatch messages in the \`dmesg\` output), please verify
that the running kernel is ${KV_FULL} and that (if used) the
initramfs does not include NVIDIA modules (or at least, not old ones).
Detected that the current kernel command line is using 'slub_debug=',
this may lead to system instability/freezes with this version of
${PN}. Bug: https://bugs.gentoo.org/796329
***WARNING***
You are installing a version of ${PN} known not to work
with a GPU of the current system. If unwanted, add the mask:
  echo '${NV_LEGACY_MASK}' > ${EROOT}/etc/portage/package.mask/${PN}
  echo '${NV_LEGACY_MASK}' >> ${EROOT}/etc/portage/package.mask
...then downgrade to a legacy[1] branch if possible (not all old versions
are available or fully functional, may need to consider nouveau[2]).
[1] https://www.nvidia.com/object/IO_32667.html
[2] https://wiki.gentoo.org/wiki/Nouveau
Open source variant of ${PN} was selected, be warned it is experimental
and only for modern GPUs (e.g. GTX 1650+). Try to disable if run into issues.
Please also see: ${EROOT}/usr/share/doc/${PF}/html/kernel_open.html
!!! WARNING !!!
Detected HandleNvidiaSleep=yes in ${EROOT}/etc/elogind/sleep.conf.d/.
This 'could' cause issues if used in combination with the new hook
installed by the ebuild to handle sleep using the official upstream
script. It is recommended to disable the option.
!!! WARNING !!!
Detected a custom script at ${EROOT}{/etc,{/usr,}/lib*}/elogind/system-sleep
referencing NVIDIA. This version of ${PN} has installed its own
hook at ${EROOT}/usr/lib/elogind/system-sleep/nvidia and it is recommended
to remove the custom one to avoid potential issues.
Feel free to ignore this warning if you know the other NVIDIA-related
scripts can be used together. The warning will be removed in the future.

Currently loaded NVIDIA modules do not match the newly installed
libraries and may prevent launching GPU-accelerated applications.
Easiest way to fix this is normally to reboot. If still run into issues
(e.g. API mismatch messages in the \`dmesg\` output), please verify
that the running kernel is ${KV_FULL} and that (if used) the
initramfs does not include NVIDIA modules (or at least, not old ones).
Detected that the current kernel command line is using 'slub_debug=',
this may lead to system instability/freezes with this version of
${PN}. Bug: https://bugs.gentoo.org/796329
***WARNING***
You are installing a version of ${PN} known not to work
with a GPU of the current system. If unwanted, add the mask:
  echo '${NV_LEGACY_MASK}' > ${EROOT}/etc/portage/package.mask/${PN}
  echo '${NV_LEGACY_MASK}' >> ${EROOT}/etc/portage/package.mask
...then downgrade to a legacy[1] branch if possible (not all old versions
are available or fully functional, may need to consider nouveau[2]).
[1] https://www.nvidia.com/object/IO_32667.html
[2] https://wiki.gentoo.org/wiki/Nouveau
!!! WARNING !!!
Detected HandleNvidiaSleep=yes in ${EROOT}/etc/elogind/sleep.conf.d/.
This 'could' cause issues if used in combination with the new hook
installed by the ebuild to handle sleep using the official upstream
script. It is recommended to disable the option.
!!! WARNING !!!
Detected a custom script at ${EROOT}{/etc,{/usr,}/lib*}/elogind/system-sleep
referencing NVIDIA. This version of ${PN} has installed its own
hook at ${EROOT}/usr/lib/elogind/system-sleep/nvidia and it is recommended
to remove the custom one to avoid potential issues.
Feel free to ignore this warning if you know the other NVIDIA-related
scripts can be used together. The warning will be removed in the future.

Currently loaded NVIDIA modules do not match the newly installed
libraries and may prevent launching GPU-accelerated applications.
Easiest way to fix this is normally to reboot. If still run into issues
(e.g. API mismatch messages in the \`dmesg\` output), please verify
that the running kernel is ${KV_FULL} and that (if used) the
initramfs does not include NVIDIA modules (or at least, not old ones).
Be warned/reminded that the 390.xx branch reached end-of-life and
NVIDIA is no longer fixing issues (including security). Free to keep
using (for now) but it is recommended to either switch to nouveau or
replace hardware. Will be kept in-tree while possible, but expect it
to be removed likely in early 2027 or earlier if major issues arise.
Note that there is no plans to patch in support for kernels branches
newer than 6.1.x which will be supported upstream until December 2026.