diff --git a/fpga_io.cpp b/fpga_io.cpp
index b5cff4f..555eaf4 100644
--- a/fpga_io.cpp
+++ b/fpga_io.cpp
@@ -519,6 +519,7 @@ void inline fpga_gpo_write(uint32_t value)
 	writel(value, (void*)(SOCFPGA_MGR_ADDRESS + 0x10));
 }
 
+#define fpga_gpo_writeN(value) writel((value), (void*)(SOCFPGA_MGR_ADDRESS + 0x10))
 #define fpga_gpo_read() gpo_copy //readl((void*)(SOCFPGA_MGR_ADDRESS + 0x10))
 #define fpga_gpi_read() (int)readl((void*)(SOCFPGA_MGR_ADDRESS + 0x14))
 
@@ -694,25 +695,304 @@ uint16_t fpga_spi_fast(uint16_t word)
 	uint32_t gpo = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE)) | word;
 	fpga_gpo_write(gpo);
 	fpga_gpo_write(gpo | SSPI_STROBE);
-	fpga_gpo_write(gpo | SSPI_STROBE);
 	fpga_gpo_write(gpo);
 	return (uint16_t)fpga_gpi_read();
 }
 
-uint32_t fpga_spi_fast_32(uint32_t dword)
+void fpga_spi_fast_block_write(const uint16_t *buf, uint32_t length)
 {
-	uint32_t gpo = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE)) | (uint16_t)dword;
+	uint32_t gpoH = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE));
+	uint32_t gpo = gpoH;
+
+	// should be optimized for speed by compiler automatically
+	while (length--)
+	{
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+	}
 	fpga_gpo_write(gpo);
-	fpga_gpo_write(gpo | SSPI_STROBE);
-
-	gpo = (gpo & ~0xFFFF) | (uint16_t)(dword>>16);
-	fpga_gpo_write(gpo);
-
-	uint16_t ret_low = (uint16_t)fpga_gpi_read();
-
-	fpga_gpo_write(gpo | SSPI_STROBE);
-	fpga_gpo_write(gpo | SSPI_STROBE);
-	fpga_gpo_write(gpo);
-
-	return (((uint16_t)fpga_gpi_read()) << 16) | ret_low;
+}
+
+void fpga_spi_fast_block_read(uint16_t *buf, uint32_t length)
+{
+	uint32_t gpo = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE));
+	uint32_t rem = length % 16;
+	length /= 16;
+
+	// not optimized by compiler automatically
+	// so do manual optimization for speed.
+	while (length--)
+	{
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+	}
+
+	while (rem--)
+	{
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint16_t)fpga_gpi_read();
+	}
+}
+
+void fpga_spi_fast_block_write_8(const uint8_t *buf, uint32_t length)
+{
+	uint32_t gpoH = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE));
+	uint32_t gpo = gpoH;
+	uint32_t rem = length % 16;
+	length /= 16;
+
+	// not optimized by compiler automatically
+	// so do manual optimization for speed.
+	while (length--)
+	{
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+	}
+
+	while (rem--)
+	{
+		gpo = gpoH | *buf++;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+	}
+
+	fpga_gpo_write(gpo);
+}
+
+void fpga_spi_fast_block_read_8(uint8_t *buf, uint32_t length)
+{
+	uint32_t gpo = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE));
+	uint32_t rem = length % 16;
+	length /= 16;
+
+	// not optimized by compiler automatically
+	// so do manual optimization for speed.
+	while (length--)
+	{
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+	}
+
+	while (rem--)
+	{
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		*buf++ = (uint8_t)fpga_gpi_read();
+	}
+}
+
+void fpga_spi_fast_block_write_be(const uint16_t *buf, uint32_t length)
+{
+	uint32_t gpoH = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE));
+	uint32_t gpo = gpoH;
+
+	// should be optimized for speed by compiler automatically
+	while (length--)
+	{
+		uint16_t tmp = *buf++;
+		tmp = (tmp << 8) | (tmp >> 8);
+		gpo = gpoH | tmp;
+		fpga_gpo_writeN(gpo);
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+	}
+	fpga_gpo_write(gpo);
+}
+
+void fpga_spi_fast_block_read_be(uint16_t *buf, uint32_t length)
+{
+	uint32_t gpo = (fpga_gpo_read() & ~(0xFFFF | SSPI_STROBE));
+
+	// should be optimized for speed by compiler automatically
+	while (length--)
+	{
+		fpga_gpo_writeN(gpo | SSPI_STROBE);
+		fpga_gpo_writeN(gpo);
+		uint16_t tmp = (uint16_t)fpga_gpi_read();
+		*buf++ = (tmp << 8) | (tmp >> 8);
+	}
 }
diff --git a/fpga_io.h b/fpga_io.h
index 9250c3a..53a3f43 100644
--- a/fpga_io.h
+++ b/fpga_io.h
@@ -12,7 +12,13 @@ int fpga_io_init();
 void fpga_spi_en(uint32_t mask, uint32_t en);
 uint16_t fpga_spi(uint16_t word);
 uint16_t fpga_spi_fast(uint16_t word);
-uint32_t fpga_spi_fast_32(uint32_t dword);
+
+void fpga_spi_fast_block_write(const uint16_t *buf, uint32_t length);
+void fpga_spi_fast_block_read(uint16_t *buf, uint32_t length);
+void fpga_spi_fast_block_write_8(const uint8_t *buf, uint32_t length);
+void fpga_spi_fast_block_read_8(uint8_t *buf, uint32_t length);
+void fpga_spi_fast_block_write_be(const uint16_t *buf, uint32_t length);
+void fpga_spi_fast_block_read_be(uint16_t *buf, uint32_t length);
 
 void fpga_set_led(uint32_t on);
 int  fpga_get_buttons();
diff --git a/spi.cpp b/spi.cpp
index cb59be5..a258ebb 100644
--- a/spi.cpp
+++ b/spi.cpp
@@ -192,52 +192,12 @@ void spi_write(const uint8_t *addr, uint16_t len, int wide)
 
 void spi_block_read(uint8_t *addr, int wide)
 {
-	if (wide)
-	{
-		uint16_t len16 = 256;
-		uint16_t *a16 = (uint16_t*)addr;
-		while (len16--) *a16++ = fpga_spi_fast(0);
-	}
-	else
-	{
-		uint16_t len = 512;
-		while (len--) *addr++ = fpga_spi_fast(0);
-	}
+	if (wide) fpga_spi_fast_block_read((uint16_t*)addr, 256);
+	else fpga_spi_fast_block_read_8(addr, 512);
 }
 
 void spi_block_write(const uint8_t *addr, int wide)
 {
-	if (wide)
-	{
-		uint16_t len16 = 256;
-		uint16_t *a16 = (uint16_t*)addr;
-		while (len16--) fpga_spi_fast(*a16++);
-	}
-	else
-	{
-		uint16_t len = 512;
-		while (len--) fpga_spi_fast(*addr++);
-	}
-}
-
-void spi_block_write_16be(const uint16_t *addr)
-{
-	uint16_t len = 256;
-	uint16_t tmp;
-	while (len--)
-	{
-		tmp = *addr++;
-		fpga_spi_fast(SWAPW(tmp));
-	}
-}
-
-void spi_block_read_16be(uint16_t *addr)
-{
-	uint16_t len = 256;
-	uint16_t tmp;
-	while (len--)
-	{
-		tmp = fpga_spi_fast(0xFFFF);
-		*addr++ = SWAPW(tmp);
-	}
+	if (wide) fpga_spi_fast_block_write((const uint16_t*)addr, 256);
+	else fpga_spi_fast_block_write_8(addr, 512);
 }
diff --git a/spi.h b/spi.h
index a12265a..04a8f83 100644
--- a/spi.h
+++ b/spi.h
@@ -43,8 +43,6 @@ void spi_read(uint8_t *addr, uint16_t len, int wide);
 void spi_write(const uint8_t *addr, uint16_t len, int wide);
 void spi_block_read(uint8_t *addr, int wide);
 void spi_block_write(const uint8_t *addr, int wide);
-void spi_block_write_16be(const uint16_t *addr);
-void spi_block_read_16be(uint16_t *addr);
 
 /* OSD related SPI functions */
 void EnableOsd_on(int target);
diff --git a/support/minimig/minimig_hdd.cpp b/support/minimig/minimig_hdd.cpp
index 863281d..715b10a 100644
--- a/support/minimig/minimig_hdd.cpp
+++ b/support/minimig/minimig_hdd.cpp
@@ -394,10 +394,10 @@ static void ATA_IdentifyDevice(uint8_t* tfr, hdfTYPE *hdf)
 	WriteTaskFile(0, tfr[2], tfr[3], tfr[4], tfr[5], tfr[6]);
 	WriteStatus(IDE_STATUS_RDY); // pio in (class 1) command type
 	EnableFpga();
-	spi_w(CMD_IDE_DATA_WR<<8); // write data command
-	spi_w(0);
-	spi_w(0);
-	spi_block_write_16be((uint16_t*)sector_buffer);
+	fpga_spi_fast(CMD_IDE_DATA_WR<<8); // write data command
+	fpga_spi_fast(0);
+	fpga_spi_fast(0);
+	fpga_spi_fast_block_write_be((uint16_t*)sector_buffer, 256);
 	DisableFpga();
 	WriteStatus(IDE_STATUS_END | IDE_STATUS_IRQ);
 }
@@ -506,20 +506,20 @@ static void ReadSector(hdfTYPE *hdf)
 static void SendSector()
 {
 	EnableFpga();
-	spi_w(CMD_IDE_DATA_WR << 8); // write data command
-	spi_w(0);
-	spi_w(0);
-	spi_block_write_16be((uint16_t*)sector_buffer);
+	fpga_spi_fast(CMD_IDE_DATA_WR << 8); // write data command
+	fpga_spi_fast(0);
+	fpga_spi_fast(0);
+	fpga_spi_fast_block_write_be((uint16_t*)sector_buffer, 256);
 	DisableFpga();
 }
 
 static void RecvSector()
 {
 	EnableFpga();
-	spi_w(CMD_IDE_DATA_RD << 8); // read data command
-	spi_w(0);
-	spi_w(0);
-	spi_block_read_16be((uint16_t*)sector_buffer);
+	fpga_spi_fast(CMD_IDE_DATA_RD << 8); // read data command
+	fpga_spi_fast(0);
+	fpga_spi_fast(0);
+	fpga_spi_fast_block_read_be((uint16_t*)sector_buffer, 256);
 	DisableFpga();
 }
 
diff --git a/support/x86/x86.cpp b/support/x86/x86.cpp
index 474bead..56fa66b 100644
--- a/support/x86/x86.cpp
+++ b/support/x86/x86.cpp
@@ -123,7 +123,7 @@ static void dma_sendbuf(uint32_t address, uint32_t length, uint32_t *data)
 	EnableIO();
 	spi8(UIO_DMA_WRITE);
 	spi32_w(address);
-	if(address == IMG_TYPE_HDD0_FAST || address == IMG_TYPE_HDD1_FAST) while (length--) fpga_spi_fast_32(*data++);
+	if (address == IMG_TYPE_HDD0_FAST || address == IMG_TYPE_HDD1_FAST) fpga_spi_fast_block_write((uint16_t*)data, length * 2);
 	else while (length--) spi32_w(*data++);
 	DisableIO();
 }
@@ -133,7 +133,7 @@ static void dma_recvbuf(uint32_t address, uint32_t length, uint32_t *data)
 	EnableIO();
 	spi8(UIO_DMA_READ);
 	spi32_w(address);
-	if (address == IMG_TYPE_HDD0_FAST || address == IMG_TYPE_HDD1_FAST) while (length--) *data++ = fpga_spi_fast_32(0);
+	if (address == IMG_TYPE_HDD0_FAST || address == IMG_TYPE_HDD1_FAST) fpga_spi_fast_block_read((uint16_t*)data, length * 2);
 	else while (length--) *data++ = spi32_w(0);
 	DisableIO();
 }