diff --git a/CHANGES.WII b/CHANGES.WII
index bff7dd6..e995b4c 100644
--- a/CHANGES.WII
+++ b/CHANGES.WII
@@ -1,4 +1,10 @@
 version 3:
+        * Skip SDL_Delay and SDL_GetTicks() which seem to give strange
+          results. Use gettime and usleep instead.
+
+        * Improve performance a bit more by doing a custom
+          blit-to-double-size implementation
+
         * Fixed changelog format
 
         * Bought a Classic controller and fixed so that it actually works...
diff --git a/Makefile b/Makefile
index 72a7c50..513a9b0 100644
--- a/Makefile
+++ b/Makefile
@@ -26,7 +26,7 @@ INCLUDES	:=
 #---------------------------------------------------------------------------------
 
 PCFLAGS = -DPRECISE_CPU_CYCLES=1 -DPRECISE_CIA_CYCLES=1 -DPC_IS_POINTER=0
-CFLAGS	= -O2 -g -Wall $(MACHDEP) $(INCLUDE) -I$(DEVKITPRO)/SDL/include -U__unix -DHAVE_SDL
+CFLAGS	= -O3 -g -Wall $(MACHDEP) $(INCLUDE) -I$(DEVKITPRO)/SDL/include -U__unix -DHAVE_SDL
 CXXFLAGS = $(CFLAGS)
 
 LDFLAGS	= -L$(DEVKITPRO)/SDL/lib -g $(MACHDEP) -Wl,-Map,$(notdir $@).map
diff --git a/Src/C64_SDL.i b/Src/C64_SDL.i
index 5898ebb..5af7e4e 100644
--- a/Src/C64_SDL.i
+++ b/Src/C64_SDL.i
@@ -12,6 +12,7 @@
 
 #if defined(GEKKO)
 #include <wiiuse/wpad.h>
+#include <ogc/lwp_watchdog.h>
 #define FONT_PATH "/apps/frodo/FreeMono.ttf"
 #define SAVES_PATH "/apps/frodo/saves"
 #define IMAGE_PATH "/apps/frodo/images"
@@ -22,7 +23,7 @@
 #define IMAGE_PATH "images"
 #define TMP_PATH "tmp"
 #endif
-#define MS_PER_FRAME 27
+#define MS_PER_FRAME 38
 
 static struct timeval tv_start;
 static int MENU_SIZE_X, MENU_SIZE_Y;
@@ -578,10 +579,14 @@ void C64::VBlank(bool draw_frame)
 	}
 	/* From Acorn port */
 	static uint64_t lastFrame;
+#if defined(GEKKO)
+        uint32_t now = ticks_to_millisecs(gettime());
+#else
         uint32_t now = SDL_GetTicks();
+#endif
 
         if ( (now - lastFrame) < MS_PER_FRAME ) {
-        	SDL_Delay( MS_PER_FRAME - (now - lastFrame) );
+        	usleep( (MS_PER_FRAME - (now - lastFrame)) * 1000);
         }
         lastFrame = now;
 }
diff --git a/Src/Display_SDL.i b/Src/Display_SDL.i
index 0c2e8bb..60c1911 100644
--- a/Src/Display_SDL.i
+++ b/Src/Display_SDL.i
@@ -11,7 +11,6 @@
 
 #include <SDL.h>
 
-
 // Display surface
 SDL_Surface *screen = NULL;
 SDL_Surface *real_screen = NULL;
@@ -161,17 +160,33 @@ void C64Display::Update(void)
 	if (ThePrefs.DisplayOption == 0) {
 		const int x_border = (DISPLAY_X - FULL_DISPLAY_X / 2) / 2;
 		const int y_border = (DISPLAY_Y - FULL_DISPLAY_Y / 2) / 2;
+		Uint8 *src_pixels = (Uint8*)screen->pixels;
+		Uint8 *dst_pixels = (Uint8*)real_screen->pixels;
+		const Uint16 src_pitch = screen->pitch;
+		const Uint16 dst_pitch = real_screen->pitch;
 
 		/* Center, double size */
-		srcrect = (SDL_Rect){x_border, y_border, FULL_DISPLAY_X / 2, FULL_DISPLAY_Y / 2};
-		dstrect = (SDL_Rect){0, 0, FULL_DISPLAY_X, FULL_DISPLAY_Y};
+		for (int y = y_border; y < (FULL_DISPLAY_Y/2) + y_border; y++)
+		{
+			for (int x = x_border; x < (FULL_DISPLAY_X / 2 + x_border); x++)
+			{
+				int src_off = y * src_pitch + x;
+				int dst_off = (y * 2 - y_border * 2) * dst_pitch + (x * 2 - x_border * 2);
+				Uint8 v = src_pixels[src_off];
+
+				dst_pixels[ dst_off ] = v;
+				dst_pixels[ dst_off + 1 ] = v;
+				dst_pixels[ dst_off + dst_pitch ] = v;
+				dst_pixels[ dst_off + dst_pitch + 1] = v;
+			}
+		}
 	}
 	else {
 		/* Stretch */
 		srcrect = (SDL_Rect){0, 0, DISPLAY_X, DISPLAY_Y};
 		dstrect = (SDL_Rect){0, 0, FULL_DISPLAY_X, FULL_DISPLAY_Y};
+		SDL_SoftStretch(screen, &srcrect, real_screen, &dstrect);                                                                     
 	}
-	SDL_SoftStretch(screen, &srcrect, real_screen, &dstrect);                                                                     
 	SDL_Flip(real_screen);
 }