68 lines (52 with data), 3.1 kB
FAQ - Frequently asked questions
I. Common errors and some more information about them
1. "Cannot find mmiotrace in the list of available tracers"
Generally this error can mean one or all of the following
- The current kernel does not have support for debugfs
- The debugfs has already been mounted somewhere else than the place stated
- The kernel has not been built with CONFIG_MMIOTRACE
2. "mmiotrace is not the current tracer"
The current tracer is not mmiotrace, the user did not run init_trace.sh
3. "(cat | grep | gawk | sed | perl | lspci | tar | xz | gzip | git | timeout) not found"
The corresponding program does not appear to be available on the system
i.e. the program returns unusual error code
- Verify that the program exist and it is not a ancient version
4. "Cannot create temporary directory"
- Make sure that "mktemp" exist on the system
- The user does not sufficient rights to write in the current directory
5. "The driver is not operating in Adaptive clock state"
- Take a look in nvidia-settings and set the current Clock state to Adaptive
6. "The PowerMizer mode is not Adaptive"
- Use nvidia-setting to change the mode
7. "Timeout while waiting for the driver to reach default 2D frequencies"
There is a number of reasons for this message to happen please take a look
section "II. "ERROR - Timeout..." and what does it mean"
II. "ERROR - Timeout..." and what does it mean
By default the program monitor_downclocking check for how long it has been
run in order to prevent unwanted endless loop. It may occur if the ddx/driver
does not go to the lowest frequencies - the ones returned by
1. The card is integrated/built into the motherboard
For some reason the driver (NV-Control) does not return always return
the correct value of NV_CTRL_GPU_DEFAULT_2D_CLOCK_FREQS
(The memory returned by the function is often zero)
- Use gdb to find the correct value and "hardcode it into monitor_downclocking.c
XXX: Add instructions how to do it exactly, step by step
2. There is more than one display connected/There is a HDMI/DP display connected
From NVIDIA's forum
"PowerMizer often needs to keep the performance level high when there
are two screens connected to make sure there is enough internal bandwidth
in the chip to drive both displays..." 
- Use a single display for that session, preferably only a VGA (the LVDS if the system is a laptop). DVI should work as well
3. None of the above
The card does reach NV_CTRL_GPU_CURRENT_CLOCK_FREQS but it takes more than
the currently coded timeout.
- Edit monitor_downclocking.c by changing NOUVEAU_TIME_WAIT to a more
appropriate value. Follow the steps in README from 2 onwards
XXX: Add instructions how to check/confirm the above statement