diff --git a/src/GPU.cpp b/src/GPU.cpp
index 92f05d87..af8c5d8c 100644
--- a/src/GPU.cpp
+++ b/src/GPU.cpp
@@ -651,8 +651,11 @@ void StartHBlank(u32 line)
     {
         // draw
         // note: this should start 48 cycles after the scanline start
-        GPU2D_A->DrawScanline(line);
-        GPU2D_B->DrawScanline(line);
+        if (line < 192)
+        {
+            GPU2D_A->DrawScanline(line);
+            GPU2D_B->DrawScanline(line);
+        }
 
         NDS::CheckDMAs(0, 0x02);
     }
diff --git a/src/GPU3D.cpp b/src/GPU3D.cpp
index d818c5ca..7521952e 100644
--- a/src/GPU3D.cpp
+++ b/src/GPU3D.cpp
@@ -607,6 +607,10 @@ int ClipPolygon(Vertex* vertices, int nverts, int clipstart)
 
     // TODO: check for 1-dot polygons
     // TODO: the hardware seems to use a different algorithm. it reacts differently to vertices with W=0
+    // some vertices that should get Y=-0x1000 get Y=0x1000 for some reason on hardware. it doesn't make sense.
+    // clipping seems to process the Y plane before the X plane.
+
+    // also, polygons with any negative W are completely rejected. (TODO)
 
     // X clipping
     nverts = ClipAgainstPlane<0, attribs>(vertices, nverts, clipstart);
diff --git a/src/GPU3D_Soft.cpp b/src/GPU3D_Soft.cpp
index 01653e94..cce5d709 100644
--- a/src/GPU3D_Soft.cpp
+++ b/src/GPU3D_Soft.cpp
@@ -1522,8 +1522,7 @@ void ScanlineFinalPass(s32 y)
     if (RenderDispCnt & (1<<5))
     {
         // edge marking
-
-        // TODO: is it applied to bottom pixels?
+        // only applied to topmost pixels
 
         for (int x = 0; x < 256; x++)
         {
@@ -1842,7 +1841,10 @@ void RenderThreadFunc()
 void RequestLine(int line)
 {
     if (RenderThreadRunning)
-        Platform::Semaphore_Wait(Sema_ScanlineCount);
+    {
+        if (line < 192)
+            Platform::Semaphore_Wait(Sema_ScanlineCount);
+    }
 }
 
 u32* GetLine(int line)
diff --git a/src/NDS.cpp b/src/NDS.cpp
index 10f3d640..271b6185 100644
--- a/src/NDS.cpp
+++ b/src/NDS.cpp
@@ -400,6 +400,7 @@ u32 RunFrame()
         s32 ndscyclestorun;
         s32 ndscycles = 0;
 
+        // TODO: give it some margin, so it can directly do 17 cycles instead of 16 then 1
         CalcIterationCycles();
 
         if (CPUStop & 0xFFFF)
diff --git a/src/wx/main.cpp b/src/wx/main.cpp
index 3c2375b5..302f5dae 100644
--- a/src/wx/main.cpp
+++ b/src/wx/main.cpp
@@ -496,8 +496,12 @@ wxThread::ExitCode EmuThread::Entry()
                 u32 fps = (nframes * 1000) / diff;
                 nframes = 0;
 
+                float fpstarget;
+                if (framerate < 1) fpstarget = 999;
+                else fpstarget = 1000.0f/framerate;
+
                 char melontitle[100];
-                sprintf(melontitle, "%d FPS - melonDS " MELONDS_VERSION, fps);
+                sprintf(melontitle, "%d/%.0f FPS | melonDS " MELONDS_VERSION, fps, fpstarget);
                 SDL_SetWindowTitle(sdlwin, melontitle);
             }
         }