diff --git a/nvml.cpp b/nvml.cpp
index 843d7e8..ff4d7e0 100644
--- a/nvml.cpp
+++ b/nvml.cpp
@@ -139,6 +139,8 @@ wrap_nvml_handle * wrap_nvml_create()
 		wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerUsage");
 	nvmlh->nvmlDeviceGetPowerUsage = (wrap_nvmlReturn_t (*)(wrap_nvmlDevice_t, unsigned int *))
 		wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerUsage");
+	nvmlh->nvmlSystemGetDriverVersion = (wrap_nvmlReturn_t (*)(char *, unsigned int))
+		wrap_dlsym(nvmlh->nvml_dll, "nvmlSystemGetDriverVersion");
 	nvmlh->nvmlErrorString = (char* (*)(wrap_nvmlReturn_t))
 		wrap_dlsym(nvmlh->nvml_dll, "nvmlErrorString");
 	nvmlh->nvmlShutdown = (wrap_nvmlReturn_t (*)())
@@ -146,6 +148,8 @@ wrap_nvml_handle * wrap_nvml_create()
 
 	if (nvmlh->nvmlInit == NULL ||
 			nvmlh->nvmlShutdown == NULL ||
+			nvmlh->nvmlErrorString == NULL ||
+			nvmlh->nvmlSystemGetDriverVersion == NULL ||
 			nvmlh->nvmlDeviceGetCount == NULL ||
 			nvmlh->nvmlDeviceGetHandleByIndex == NULL ||
 			nvmlh->nvmlDeviceGetPciInfo == NULL ||
@@ -161,6 +165,7 @@ wrap_nvml_handle * wrap_nvml_create()
 	}
 
 	nvmlh->nvmlInit();
+	nvmlh->nvmlSystemGetDriverVersion(driver_version, sizeof(driver_version));
 	nvmlh->nvmlDeviceGetCount(&nvmlh->nvml_gpucount);
 
 	/* Query CUDA device count, in case it doesn't agree with NVML, since  */
diff --git a/nvml.h b/nvml.h
index 6a60ab9..74bbe19 100644
--- a/nvml.h
+++ b/nvml.h
@@ -71,6 +71,7 @@ typedef struct {
 	wrap_nvmlReturn_t (*nvmlDeviceGetFanSpeed)(wrap_nvmlDevice_t, unsigned int *);
 	wrap_nvmlReturn_t (*nvmlDeviceGetPerformanceState)(wrap_nvmlDevice_t, int *); /* enum */
 	wrap_nvmlReturn_t (*nvmlDeviceGetPowerUsage)(wrap_nvmlDevice_t, unsigned int *);
+	wrap_nvmlReturn_t (*nvmlSystemGetDriverVersion)(char *version, unsigned int len);
 	char* (*nvmlErrorString)(wrap_nvmlReturn_t);
 	wrap_nvmlReturn_t (*nvmlShutdown)(void);
 } wrap_nvml_handle;