Add section for resultant frequency during GPU workloads. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@xxxxxxxxxxxxxxx> --- Documentation/admin-guide/pm/intel_pstate.rst | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/Documentation/admin-guide/pm/intel_pstate.rst b/Documentation/admin-guide/pm/intel_pstate.rst index d2b6fda..a7c6308 100644 --- a/Documentation/admin-guide/pm/intel_pstate.rst +++ b/Documentation/admin-guide/pm/intel_pstate.rst @@ -482,6 +482,25 @@ on the following rules, regardless of the current operation mode of the driver: 3. The global and per-policy limits can be set independently. +One important point to note is that all the above limit controls including +``no_turbo``, only limits the P-state requests by the intel_pstate driver. The +actual performance delivered by the processor is also influenced by GPU +(Graphics processing unit) for GPU bound workloads. So, it is possible +that, while GPU is requesting higher frequencies, the resultant P-state +(CPU frequency) can be higher than the limits set via sysfs. +Equivalent limits must, therefore, also be set for the GPU. + +An example where an Intel i915 GPU is used: + +#cat /sys/kernel/debug/dri/0/i915_ring_freq_table + +displays effective CPU frequency for a GPU frequency. To limit +a CPU frequency, a corresponding GPU frequency limit must also be placed: + +/sys/class/drm/card0/gt_max_freq_mhz and + +/sys/class/drm/card0/gt_boost_freq_mhz + If the `HWP feature is enabled in the processor <Active Mode With HWP_>`_, the resulting effective values are written into its registers whenever the limits change in order to request its internal P-state selection logic to always set -- 2.7.5 -- To unsubscribe from this list: send the line "unsubscribe linux-doc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
