GpuCoreTemp attribute wrong in SLI mode.

[Spyke]

In SLI mode only, the second GPU's temp in the GPUCoreTemp attribute always reports 0. Disabling SLI gives the correct reporting.

See here:

Attribute 'GPUCoreTemp' (darius.spykes.net:0[gpu:0]): 50.

Attribute 'GPUCoreTemp' (darius.spykes.net:0[gpu:1]): 0.


Second GPU's temperature sensor can still be read from the 'ThermalSensorReading' attribute. (In this case 0 and 2 are the GPU core's and 1 and 3 are the PCB sensors)

Attribute 'ThermalSensorReading' (darius.spykes.net:0[thermalsensor:0]): 50.

Attribute 'ThermalSensorReading' (darius.spykes.net:0[thermalsensor:1]): 44.

Attribute 'ThermalSensorReading' (darius.spykes.net:0[thermalsensor:2]): 52.

Attribute 'ThermalSensorReading' (darius.spykes.net:0[thermalsensor:3]): 43.

[Spyke]

I made a clever hack to make gkrellm (or whatever monitoring program that uses nvidia-settings) report the proper temperature of the second GPU.

File: /usr/local/bin/nvidia-settings

#!/bin/sh
if [ "$1" != "-q" ]; then
        /usr/bin/nvidia-settings "$@"
else
        if [ "$2" = "[gpu:1]/GPUCoreTemp" ]; then
                /usr/bin/nvidia-settings -q "[thermalsensor:2]/ThermalSensorReading" | sed -e "s/thermalsensor:2/gpu:1/g" -e "s/ThermalSensorReading/GPUCoreTemp/g"
        else
                /usr/bin/nvidia-settings "$@"
        fi
fi

Broken nvidia-settings:
spike@darius:~> /usr/bin/nvidia-settings -q [gpu:1]/GPUCoreTemp

Attribute 'GPUCoreTemp' (darius.spykes.net:0[gpu:1]): 0.
'GPUCoreTemp' is an integer attribute.
'GPUCoreTemp' is a read-only attribute.
'GPUCoreTemp' can use the following target types: X Screen, GPU.

Fixed with script:

spike@darius:~> /usr/local/bin/nvidia-settings -q [gpu:1]/GPUCoreTemp

Attribute 'GPUCoreTemp' (darius.spykes.net:0[gpu:1]): 49.
The valid values for 'GPUCoreTemp' are in the range 0 - 191 (inclusive).
'GPUCoreTemp' is a read-only attribute.
'GPUCoreTemp' can use the following target types: Thermal Sensor.

[pigeon]

I have the same problem with my two Gigabyte GTX 460 cards. But I don't have SLI enabled. Second GPU always gives a 0 temperature.

I don't have those ThermalSensorReading attributes, are they only available in a newer version of the driver? I'm running 256.53.

Thanks.

[CuChullain]

I also have this problem with 2x9600gt in SLI with the 260.19.12 drivers and kernel 2.6.35.1. I load the nvidia kernel module with "options nvidia NVreg_EnableMSI=1". If I update the kernel to 2.6.36(latest ATM) X will not even start and I get the following:

(**) NVIDIA(0): Option "NoLogo" "1"
(**) NVIDIA(0): Option "Coolbits" "1"
(**) NVIDIA(0): Option "SLI" "AA"
(**) NVIDIA(0): Option "MultiGPU" "AA"
(**) NVIDIA(0): Option "TripleBuffer" "True"
(**) Oct 23 23:34:24 NVIDIA(0): Enabling RENDER acceleration
(**) Oct 23 23:34:24 NVIDIA(0): NVIDIA SLI antialiasing selected.
(**) Oct 23 23:34:24 NVIDIA(0): NVIDIA Multi-GPU antialiasing selected.
(II) Oct 23 23:34:24 NVIDIA(0): Support for GLX with the Damage and Composite X extensions is
(II) Oct 23 23:34:24 NVIDIA(0):     enabled.
(EE) Oct 23 23:34:29 NVIDIA(0): The NVIDIA kernel module does not appear to be receiving
(EE) Oct 23 23:34:29 NVIDIA(0):     interrupts generated by the NVIDIA graphics device
(EE) Oct 23 23:34:29 NVIDIA(0):     PCI:2:0:0.  Please see Chapter 8: Common Problems in the
(EE) Oct 23 23:34:29 NVIDIA(0):     README for additional information.
(EE) Oct 23 23:34:29 NVIDIA(0): Failed to initialize one NVIDIA graphics device!
(WW) Oct 23 23:34:29 NVIDIA(0): Failed to initialize SLI configuration.  Reason: One GPU
(WW) Oct 23 23:34:29 NVIDIA(0):     failed to initialize; Only one GPU will be used for this X
(WW) Oct 23 23:34:29 NVIDIA(0):     screen.
(EE) Oct 23 23:34:34 NVIDIA(0): The NVIDIA kernel module does not appear to be receiving
(EE) Oct 23 23:34:34 NVIDIA(0):     interrupts generated by the NVIDIA graphics device
(EE) Oct 23 23:34:34 NVIDIA(0):     PCI:4:0:0.  Please see Chapter 8: Common Problems in the
(EE) Oct 23 23:34:34 NVIDIA(0):     README for additional information.
(EE) Oct 23 23:34:34 NVIDIA(0): Failed to initialize the NVIDIA graphics device!
(II) UnloadModule: "nvidia"
(II) UnloadModule: "wfb"
(II) UnloadModule: "fb"
(EE) Screen(s) found, but none have a usable configuration.