diff --git a/csrc/u8x8.h b/csrc/u8x8.h
index a295f3df..41bec5d3 100644
--- a/csrc/u8x8.h
+++ b/csrc/u8x8.h
@@ -631,6 +631,7 @@ uint8_t u8x8_cad_011(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
 uint8_t u8x8_cad_100(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
 uint8_t u8x8_cad_st7920_spi(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
 uint8_t u8x8_cad_ssd13xx_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
+uint8_t u8x8_cad_ssd13xx_fast_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
 uint8_t u8x8_cad_st75256_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
 uint8_t u8x8_cad_ld7032_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
 uint8_t u8x8_cad_uc16xx_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr);
diff --git a/csrc/u8x8_cad.c b/csrc/u8x8_cad.c
index 8d6b50f9..0a407cd8 100644
--- a/csrc/u8x8_cad.c
+++ b/csrc/u8x8_cad.c
@@ -407,6 +407,7 @@ static void u8x8_i2c_data_transfer(u8x8_t *u8x8, uint8_t arg_int, void *arg_ptr)
     u8x8_byte_EndTransfer(u8x8);
 }
 
+/* classic version: will put a start/stop condition around each command and arg */
 uint8_t u8x8_cad_ssd13xx_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr)
 {
   uint8_t *p;
@@ -456,6 +457,83 @@ uint8_t u8x8_cad_ssd13xx_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *a
   return 1;
 }
 
+
+/* fast version with reduced data start/stops, issue 735 */
+uint8_t u8x8_cad_ssd13xx_fast_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr)
+{
+  static uint8_t in_transfer = 0;
+  uint8_t *p;
+  switch(msg)
+  {
+    case U8X8_MSG_CAD_SEND_CMD:
+      /* improved version, takeover from ld7032 */
+      /* assumes, that the args of a command is not longer than 31 bytes */
+      /* speed improvement is about 4% compared to the classic version */
+      if ( in_transfer != 0 )
+	 u8x8_byte_EndTransfer(u8x8); 
+      
+      u8x8_byte_StartTransfer(u8x8);
+      u8x8_byte_SendByte(u8x8, 0x000);	/* cmd byte for ssd13xx controller */
+      u8x8_byte_SendByte(u8x8, arg_int);
+      in_transfer = 1;
+      /* lightning version: can replace the improved version from above */
+      /* the drawback of the lightning version is this: The complete init sequence */
+      /* must fit into the 32 byte Arduino Wire buffer, wbich might not always be the case */
+      /* speed improvement is about 6% compared to the classic version */
+      // if ( in_transfer == 0 )
+	// {
+	//   u8x8_byte_StartTransfer(u8x8);
+	//   u8x8_byte_SendByte(u8x8, 0x000);	/* cmd byte for ssd13xx controller */
+	//   in_transfer = 1;
+	// }
+	//u8x8_byte_SendByte(u8x8, arg_int);
+      break;
+    case U8X8_MSG_CAD_SEND_ARG:
+      u8x8_byte_SendByte(u8x8, arg_int);
+      break;      
+    case U8X8_MSG_CAD_SEND_DATA:
+      if ( in_transfer != 0 )
+	u8x8_byte_EndTransfer(u8x8); 
+      
+    
+      /* the FeatherWing OLED with the 32u4 transfer of long byte */
+      /* streams was not possible. This is broken down to */
+      /* smaller streams, 32 seems to be the limit... */
+      /* I guess this is related to the size of the Wire buffers in Arduino */
+      /* Unfortunately, this can not be handled in the byte level drivers, */
+      /* so this is done here. Even further, only 24 bytes will be sent, */
+      /* because there will be another byte (DC) required during the transfer */
+      p = arg_ptr;
+       while( arg_int > 24 )
+      {
+	u8x8_i2c_data_transfer(u8x8, 24, p);
+	arg_int-=24;
+	p+=24;
+      }
+      u8x8_i2c_data_transfer(u8x8, arg_int, p);
+      in_transfer = 0;
+      break;
+    case U8X8_MSG_CAD_INIT:
+      /* apply default i2c adr if required so that the start transfer msg can use this */
+      if ( u8x8->i2c_address == 255 )
+	u8x8->i2c_address = 0x078;
+      return u8x8->byte_cb(u8x8, msg, arg_int, arg_ptr);
+    case U8X8_MSG_CAD_START_TRANSFER:
+      in_transfer = 0;
+      break;
+    case U8X8_MSG_CAD_END_TRANSFER:
+      if ( in_transfer != 0 )
+	u8x8_byte_EndTransfer(u8x8); 
+      in_transfer = 0;
+      break;
+    default:
+      return 0;
+  }
+  return 1;
+}
+
+
+
 /* the st75256 i2c driver is a copy of the ssd13xx driver, but with arg=1 */
 /* modified from cad001 (ssd13xx) to cad011 */
 uint8_t u8x8_cad_st75256_i2c(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr)
diff --git a/sys/arduino/u8g2_full_buffer/FPS/FPS.ino b/sys/arduino/u8g2_full_buffer/FPS/FPS.ino
index 8dba61ac..2f1f8d5e 100644
--- a/sys/arduino/u8g2_full_buffer/FPS/FPS.ino
+++ b/sys/arduino/u8g2_full_buffer/FPS/FPS.ino
@@ -153,7 +153,13 @@
   Uno U8G2_ST7920_128X64_F_HW_SPI			HW SPI 1.8MHz FPS: Clip=23.0 Box=22.8  @=7.4 Pix=10.4
   Uno U8G2_ST7920_128X64_F_SW_SPI			SW SPI		FPS: Clip=22.7 Box=21.0  @=7.1 Pix=9.6
   Due U8G2_ST7920_128X64_F_SW_SPI			SW SPI		FPS: Clip=3.3 Box=3.2  @=3.1 Pix=3.1
-  Due U8G2_ST7920_128X64_F_SW_SPI			SW SPI		FPS: Clip=18.0 Box=17.8  @=13.5 Pix=14.6		/* Due optimized, 1000ns */
+  Due U8G2_ST7920_128X64_F_SW_SPI			SW SPI		FPS: Clip=18.0 Box=17.8  @=13.5 Pix=14.6		Due optimized, 1000ns
+  
+  20 Dec 2018
+  Uno U8G2_SSD1306_128X64_NONAME_F_SW_I2C	SW I2C		FPS: Clip=1.8 Box=1.9  @=1.6 Pix=1.7
+  Uno U8G2_SSD1306_128X64_NONAME_F_HW_I2C	HW I2C		FPS: Clip=17.3 Box=19.5  @=6.6 Pix=8.8		old SSD13xx cad procedure
+  Uno U8G2_SSD1306_128X64_NONAME_F_HW_I2C	HW I2C		FPS: Clip=18.0 Box=20.3  @=6.6 Pix=9.0		new SSD13xx cad procedure
+  Uno U8G2_SSD1306_128X64_NONAME_F_HW_I2C	HW I2C		FPS: Clip=18.3 Box=20.8  @=6.7 Pix=9.1		new SSD13xx cad procedure (lightning version)
 */
 
 
@@ -184,7 +190,7 @@
 //U8G2_SSD1306_128X64_NONAME_F_4W_HW_SPI u8g2(U8G2_R0, /* cs=*/ 12, /* dc=*/ 4, /* reset=*/ 6);	// Arduboy (Production, Kickstarter Edition)
 //U8G2_SSD1306_128X64_NONAME_F_4W_HW_SPI u8g2(U8G2_R0, /* cs=*/ 10, /* dc=*/ 9, /* reset=*/ 8);
 //U8G2_SSD1306_128X64_NONAME_F_3W_SW_SPI u8g2(U8G2_R0, /* clock=*/ 13, /* data=*/ 11, /* cs=*/ 10, /* reset=*/ 8);
-//U8G2_SSD1306_128X64_NONAME_F_HW_I2C u8g2(U8G2_R0, /* reset=*/ U8X8_PIN_NONE);
+U8G2_SSD1306_128X64_NONAME_F_HW_I2C u8g2(U8G2_R0, /* reset=*/ U8X8_PIN_NONE);
 //U8G2_SSD1306_128X64_ALT0_F_HW_I2C u8g2(U8G2_R0, /* reset=*/ U8X8_PIN_NONE);   // same as the NONAME variant, but may solve the "every 2nd line skipped" problem
 //U8G2_SSD1306_128X64_NONAME_F_SW_I2C u8g2(U8G2_R0, /* clock=*/ 13, /* data=*/ 11, /* reset=*/ 8);
 //U8G2_SSD1306_128X64_NONAME_F_SW_I2C u8g2(U8G2_R0, /* clock=*/ SCL, /* data=*/ SDA, /* reset=*/ U8X8_PIN_NONE);   // All Boards without Reset of the Display
diff --git a/tools/codebuild/codebuild.c b/tools/codebuild/codebuild.c
index a06b4696..cb548b0d 100644
--- a/tools/codebuild/codebuild.c
+++ b/tools/codebuild/codebuild.c
@@ -2,6 +2,14 @@
 
   generate and update c/c++ files
   this also replaces the buildcpp tool
+  
+  
+  Dec 2018:
+  There are now two SSD13xx cad procedures:
+    u8x8_cad_ssd13xx_i2c			Put a I2C start/stop around each command and each argument	--> many start/stop commands
+    u8x8_cad_ssd13xx_fast_i2c		Put a I2C start/stop around each command+arg sequence		--> start/stop is probably halfed --> 4% faster
+
+  
 
 */
 
@@ -119,7 +127,7 @@ struct controller controller_list[] =
     }
   },
   {
-    "ssd1306", 	16, 	8, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_i2c", "i2c", COM_I2C,
+    "ssd1306", 	16, 	8, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_fast_i2c", "i2c", COM_I2C,
     "", /* is_generate_u8g2_class= */ 1,
     {
       { "128x64_noname" },
@@ -278,7 +286,7 @@ struct controller controller_list[] =
     }
   },
   {
-    "ssd1306", 	16, 	4, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_i2c", "i2c", COM_I2C,
+    "ssd1306", 	16, 	4, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_fast_i2c", "i2c", COM_I2C,
     "", /* is_generate_u8g2_class= */ 1,
     {
       { "128x32_univision" },
@@ -295,7 +303,7 @@ struct controller controller_list[] =
     }
   },
   {
-    "ssd1306", 	8, 	6, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_i2c", "i2c", COM_I2C,
+    "ssd1306", 	8, 	6, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_fast_i2c", "i2c", COM_I2C,
     "", /* is_generate_u8g2_class= */ 1,
     {
       { "64x48_er" },
@@ -312,7 +320,7 @@ struct controller controller_list[] =
     }
   },
   {
-    "ssd1306", 	6, 	8, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_i2c", "i2c", COM_I2C,
+    "ssd1306", 	6, 	8, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_fast_i2c", "i2c", COM_I2C,
     "", /* is_generate_u8g2_class= */ 1,
     {
       { "48x64_winstar" },
@@ -331,7 +339,7 @@ struct controller controller_list[] =
     }
   },
   {
-    "ssd1306", 	8, 	4, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_i2c", "i2c", COM_I2C,
+    "ssd1306", 	8, 	4, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_fast_i2c", "i2c", COM_I2C,
     "", /* is_generate_u8g2_class= */ 1,
     {
       { "64x32_noname" },
@@ -349,7 +357,7 @@ struct controller controller_list[] =
     }
   },
   {
-    "ssd1306", 	12, 	2, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_i2c", "i2c", COM_I2C,
+    "ssd1306", 	12, 	2, 	"u8g2_ll_hvline_vertical_top_lsb", "u8x8_cad_ssd13xx_fast_i2c", "i2c", COM_I2C,
     "", /* is_generate_u8g2_class= */ 1,
     {
       { "96x16_er" },