diff --git a/src/nvidia.cc b/src/nvidia.cc
index 5417e730..73b904a3 100644
--- a/src/nvidia.cc
+++ b/src/nvidia.cc
@@ -101,6 +101,8 @@
 #include "temphelper.h"
 #include "x11.h"
 
+#include <memory>
+
 // Separators for nvidia string parsing
 // (sample: "perf=0, nvclock=324, nvclockmin=324, nvclockmax=324 ; perf=1,
 // nvclock=549, nvclockmin=549, nvclockmax=549")
@@ -366,6 +368,14 @@ struct nvidia_c_string {
 const int MAXNUMGPU = 64;
 
 namespace {
+
+// Deleter for nv display to use with std::unique_ptr
+void close_nvdisplay(Display * dp) {
+  XCloseDisplay(dp);
+}
+
+using unique_display_t = std::unique_ptr<Display, decltype(&close_nvdisplay)>;
+
 class nvidia_display_setting
     : public conky::simple_config_setting<std::string> {
   typedef conky::simple_config_setting<std::string> Base;
@@ -378,7 +388,7 @@ class nvidia_display_setting
 
  public:
   nvidia_display_setting() : Base("nvidia_display", std::string(), false) {}
-  virtual Display *get_nvdisplay();
+  virtual unique_display_t get_nvdisplay();
 };
 
 void nvidia_display_setting::lua_setter(lua::state &l, bool init) {
@@ -391,28 +401,21 @@ void nvidia_display_setting::lua_setter(lua::state &l, bool init) {
   ++s;
 }  // namespace
 
-Display *nvidia_display_setting::get_nvdisplay() {
+unique_display_t nvidia_display_setting::get_nvdisplay() {
   if (!nvdisplay.empty()) {
-    Display *nvd = XOpenDisplay(nvdisplay.c_str());
-    if (nvd == nullptr) {
+    unique_display_t nvd(XOpenDisplay(nvdisplay.c_str()), &close_nvdisplay);
+    if (!nvd) {
       NORM_ERR(nullptr, NULL, "can't open nvidia display: %s",
                XDisplayName(nvdisplay.c_str()));
     }
     return nvd;
   }
-  return nullptr;
+  return unique_display_t(nullptr, &close_nvdisplay);
 }  // namespace
 
 void nvidia_display_setting::cleanup(lua::state &l) {
   lua::stack_sentry s(l, -1);
 
-  Display *nvdisplay = get_nvdisplay();
-
-  if (nvdisplay && nvdisplay != display) {
-    XCloseDisplay(nvdisplay);
-    nvdisplay = nullptr;
-  }
-
   l.pop();
 }
 
@@ -754,8 +757,8 @@ static int cache_nvidia_value(TARGET_ID tid, ATTR_ID aid, Display *dpy,
 // Retrieve attribute value via nvidia interface
 static int get_nvidia_value(TARGET_ID tid, ATTR_ID aid, int gid,
                             const char *arg) {
-  Display *dpy = nvidia_display.get_nvdisplay();
-  if (!dpy) dpy = display;
+  auto nvdpy = nvidia_display.get_nvdisplay();
+  Display * dpy = nvdpy ? nvdpy.get() : display;
   int value;
 
   // Check if the aid is cacheable
@@ -785,8 +788,8 @@ static int get_nvidia_value(TARGET_ID tid, ATTR_ID aid, int gid,
 // Retrieve attribute string via nvidia interface
 static char *get_nvidia_string(TARGET_ID tid, ATTR_ID aid, int gid,
                                const char *arg) {
-  Display *dpy = nvidia_display.get_nvdisplay();
-  if (!dpy) dpy = display;
+  auto nvdpy = nvidia_display.get_nvdisplay();
+  Display * dpy = nvdpy ? nvdpy.get() : display;
   char *str;
 
   // Query nvidia interface
@@ -958,8 +961,8 @@ void print_nvidia_value(struct text_object *obj, char *p,
   int event_base;
   int error_base;
 
-  Display *dpy = nvidia_display.get_nvdisplay();
-  if (!dpy) dpy = display;
+  auto nvdpy = nvidia_display.get_nvdisplay();
+  Display * dpy = nvdpy ? nvdpy.get() : display;
 
   if (!dpy) {
     NORM_ERR("%s: no display set (try setting nvidia_display)", __func__);
@@ -1062,8 +1065,8 @@ double get_nvidia_barval(struct text_object *obj) {
   int event_base;
   int error_base;
 
-  Display *dpy = nvidia_display.get_nvdisplay();
-  if (!dpy) dpy = display;
+  auto nvdpy = nvidia_display.get_nvdisplay();
+  Display * dpy = nvdpy ? nvdpy.get() : display;
 
   if (!dpy) {
     NORM_ERR("%s: no display set (try setting nvidia_display)", __func__);