diff --git a/desmume/src/GPU_osd.cpp b/desmume/src/GPU_osd.cpp
index bb515c831..392553dbf 100644
--- a/desmume/src/GPU_osd.cpp
+++ b/desmume/src/GPU_osd.cpp
@@ -504,7 +504,7 @@ void DrawHUD()
 
 	if (CommonSettings.hud.FpsDisplay) 
 	{
-		osd->addFixed(Hud.FpsDisplay.x, Hud.FpsDisplay.y, "Fps:%02d/%02d%s", Hud.fps, Hud.fps3d, driver->EMU_IsEmulationPaused() ? " (paused)" : "");
+		osd->addFixed(Hud.FpsDisplay.x, Hud.FpsDisplay.y, "Fps:%02d/%02d (%02d%%)%s", Hud.fps, Hud.fps3d, Hud.arm9load, driver->EMU_IsEmulationPaused() ? " (paused)" : "");
 	}
 
 	if (CommonSettings.hud.FrameCounterDisplay) 
diff --git a/desmume/src/GPU_osd.h b/desmume/src/GPU_osd.h
index 872c2eb3f..722642ce4 100644
--- a/desmume/src/GPU_osd.h
+++ b/desmume/src/GPU_osd.h
@@ -56,6 +56,7 @@ public:
 	{
 		fps = 0;
 		fps3d = 0;
+		arm9load = 0;
 		clicked = false;
 	}
 
@@ -71,7 +72,7 @@ public:
 	HudCoordinates &hud(int i) { return ((HudCoordinates*)this)[i]; }
 	void reset();
 
-	int fps, fps3d;
+	int fps, fps3d, arm9load;
 	bool clicked;
 };
 
diff --git a/desmume/src/NDSSystem.cpp b/desmume/src/NDSSystem.cpp
index b3c7a1c9f..4b80542d7 100644
--- a/desmume/src/NDSSystem.cpp
+++ b/desmume/src/NDSSystem.cpp
@@ -1946,6 +1946,12 @@ static void execHardware_hstart_vblankStart()
 
 	//trigger vblank dmas
 	triggerDma(EDMAMode_VBlank);
+
+	//tracking for arm9 load average
+	nds.runCycleCollector[nds.idleFrameCounter] = 1120380-nds.idleCycles;
+	nds.idleFrameCounter++;
+	nds.idleFrameCounter &= 15;
+	nds.idleCycles = 0;
 }
 
 static void execHardware_hstart_vcount()
@@ -2232,7 +2238,9 @@ static /*donotinline*/ std::pair<s32,s32> armInnerLoop(
 			}
 			else
 			{
+				s32 temp = arm9;
 				arm9 = min(s32next, arm9 + kIrqWait);
+				nds.idleCycles += arm9-temp;
 			}
 		}
 		if(doarm7 && (!doarm9 || arm7 <= timer))
diff --git a/desmume/src/windows/main.cpp b/desmume/src/windows/main.cpp
index 31114bfe3..9fc2fcd9f 100644
--- a/desmume/src/windows/main.cpp
+++ b/desmume/src/windows/main.cpp
@@ -1277,8 +1277,7 @@ static void StepRunLoop_User()
 		for(int i=0;i<16;i++)
 			load = load/8 + nds.runCycleCollector[(i+nds.idleFrameCounter)&15]*7/8;
 		load = std::min(100,std::max(0,(int)(load*100/1120380)));
-		//sprintf(txt,"(%02d%%) %s", load, DESMUME_NAME_AND_VERSION);
-		SetWindowText(mainLoopData.hwnd, DESMUME_NAME_AND_VERSION);
+		Hud.arm9load = load;
 	}
 }