NV-CONTROL: *_THRESHOLD attributes readings always to 127

It seems that using NV-CONTROL to read the threshold temperature gives you always a values of 127.

The attributes affected are:

• NV_CTRL_GPU_CORE_THRESHOLD
• NV_CTRL_GPU_DEFAULT_CORE_THRESHOLD
• NV_CTRL_GPU_MAX_CORE_THRESHOLD

I modified the nv-control-targets.c to print those values, together with the core temp:

GPU 0 information:
   Product Name                    : GeForce RTX 2080 Ti
   Core temp                     > : 33
   Core threshold temp           > : 127
   Core default core threshold   > : 127
   Core max core threshold       > : 127
   GPU UUID                        : GPU-22569caf-c41b-be44-07e1-314d0904febb
   Coolers on GPU                  : COOLER-1, COOLER-0
   Thermal Sensors on GPU          : THERMAL-SENSOR-0
   Connected Display Devices       : DPY-5
   Framelock Devices               : 
   Number of X Screens on GPU 0    : 1

As you can see the core temp is correctly read but all the threshold are stuck to 127.

The same happens if I use the X Protocol to communicate with NV-CONTROL from Python.

nvidia-smi shows the correct values:

$ nvidia-smi -q -x | grep "<temperature>" -A7
		<temperature>
			<gpu_temp>33 C</gpu_temp>
			<gpu_temp_max_threshold>94 C</gpu_temp_max_threshold>
			<gpu_temp_slow_threshold>91 C</gpu_temp_slow_threshold>
			<gpu_temp_max_gpu_threshold>89 C</gpu_temp_max_gpu_threshold>
			<memory_temp>N/A</memory_temp>
			<gpu_temp_max_mem_threshold>N/A</gpu_temp_max_mem_threshold>
		</temperature>

I was able to reproduce this issue on all the GPU i tested it:

• 2 different GeForce GTX 2080 Ti (driver 415.25)
• GeForce GTX 1080 Ti (driver 415.25)
• GeForce GTX 1060 6GB (driver 396.54)

I am currently developing an alternative to MSI Afterburner for Linux so I would like to know if and when a fix for this will be provided (and also if I can get some help from Nvidia in testing the application):

Start with the attached nv-control-targets.c and the NV_CTRL_GPU_*_THRESHOLD attribute reads, then compare their results with the nvidia-smi XML values shown. Trace the NV-CONTROL/X Protocol path for these attributes and reproduce the issue on an affected GPU. Done means the three threshold attributes return their actual values instead of 127.