Get temperature from NVidia GPU using NVAPI

问题:

I have been trying for the last few days to get the temperature of my GPU using C++

using the NVAPI i have the following code

#include "stdafx.h"
#include "nvapi.h"


int _tmain(int argc, _TCHAR* argv[])
{
    NvAPI_Status ret = NVAPI_OK;
    int i=0;

    NvDisplayHandle hDisplay_a[NVAPI_MAX_PHYSICAL_GPUS*2] = {0};

    ret = NvAPI_Initialize();

    if (!ret == NVAPI_OK){
        NvAPI_ShortString string;
        NvAPI_GetErrorMessage(ret, string);
        printf("NVAPI NvAPI_Initialize: %s\n", string);
    }

    NvAPI_ShortString ver;

    NvAPI_GetInterfaceVersionString(ver);
    printf("NVAPI Version: %s\n", ver);

    NvU32 cnt;

    NvPhysicalGpuHandle phys;

    ret = NvAPI_EnumPhysicalGPUs(&phys, &cnt);

    if (!ret == NVAPI_OK){
        NvAPI_ShortString string;
        NvAPI_GetErrorMessage(ret, string);
        printf("NVAPI NvAPI_EnumPhysicalGPUs: %s\n", string);
    }

    NvAPI_ShortString name;

    NV_GPU_THERMAL_SETTINGS thermal;

    ret = NvAPI_GPU_GetFullName(phys, name);
    if (!ret == NVAPI_OK){
        NvAPI_ShortString string;
        NvAPI_GetErrorMessage(ret, string);
        printf("NVAPI NvAPI_GPU_GetFullName: %s\n", string);
    }

    printf("Name: %s\n", name);
    thermal.version =NV_GPU_THERMAL_SETTINGS_VER;
    ret = NvAPI_GPU_GetThermalSettings(phys,0, &thermal);

    if (!ret == NVAPI_OK){
        NvAPI_ShortString string;
        NvAPI_GetErrorMessage(ret, string);
        printf("NVAPI NvAPI_GPU_GetThermalSettings: %s\n", string);
    }

    printf("Temp: %l C\n", thermal.sensor[0].currentTemp);

    return 0;
}

The output i get is

NVAPI Version: NVidia Complete Version 1.10
Name: GeForce GTX 680
Temp:  C

So NVAPI is initializing, and it can find my GPU which is a GTX680, but it wont return a temperature

回答1:

printf is very sensitive to the size and type of parameters it is passed.

Since currentTemp is fixed as a 32 bit unsigned (and C++ does not guarantee specific bitcounts for the types printf likes), cast it to a type that printf can anticipate.

printf("Temp: %u C\n", static_cast<unsigned>(thermal.sensor[0].currentTemp) );

回答2:

I think that the problem is that you are required to set thermal settings in the following way:

NV_GPU_THERMAL_SETTINGS ThermalInfo;
ThermalInfo.version = NV_GPU_THERMAL_SETTINGS_VER_2;
ThermalInfo.sensor[0].controller = NVAPI_THERMAL_CONTROLLER_GPU_INTERNAL;
ThermalInfo.sensor[0].target = NVAPI_THERMAL_TARGET_GPU;

Also you should also write:

NvPhysicalGpuHandle  nvGPUHandle[NVAPI_MAX_PHYSICAL_GPUS];

because you could corrupt memory. (Note look at NvAPI_EnumPhysicalGPUs in the documentation supplied with NVAPI).

回答3:

Your code almost works, just typos on these lines ;)

(!ret == NVAPI_OK)

change them to :

(ret != NVAPI_OK)

回答4:

You are already able to get correct value for thermal.sensor[0].currentTemp,

just you are mistaking while printing.

Please Look at thermal.sensor structure in Nvapi:

struct
{
    NV_THERMAL_CONTROLLER       controller;         //!< internal, ADM1032, MAX6649...
    NvS32                       defaultMinTemp;     //!< Minimum default temperature value of the thermal sensor in degree Celsius
    NvS32                       defaultMaxTemp;     //!< Maximum default temperature value of the thermal sensor in degree Celsius
    NvS32                       currentTemp;        //!< Current temperature value of the thermal sensor in degree Celsius
    NV_THERMAL_TARGET           target;             //!< Thermal sensor targeted - GPU, memory, chipset, powersupply, Visual Computing Device, etc
} sensor[NVAPI_MAX_THERMAL_SENSORS_PER_GPU];

It shows that NvS32 currentTemp is Signed 32 bit integer.