diff --git a/src/common.h b/src/common.h
index 3b15300..938b9dd 100644
--- a/src/common.h
+++ b/src/common.h
@@ -385,7 +385,7 @@ bool flash_erase_sectors(uint32_t baseaddr, unsigned sectsize, unsigned sectcoun
 void flash_read(uint32_t baseaddr, uint8_t *buf, unsigned size);
 bool flash_check_erased(uintptr_t addr, unsigned size);
 bool flash_program(uint32_t baseaddr, const uint8_t *buf, unsigned size);
-bool flash_program_buffered(uint32_t baseaddr, const uint8_t *buf, unsigned size, unsigned bufsize);
+bool flash_program_buffered(uint32_t baseaddr, const uint8_t *buf, unsigned size, unsigned bufsize, bool precopy);
 bool flash_verify(uint32_t baseaddr, const uint8_t *buf, unsigned size);
 void flash_erase_fsm_start(t_flash_erase_state *st, uint32_t baseaddr, unsigned sectsize, unsigned sectorcnt);
 int flash_erase_fsm_step(t_flash_erase_state *st);
diff --git a/src/flash.c b/src/flash.c
index 3c3cc76..7a9f0c8 100644
--- a/src/flash.c
+++ b/src/flash.c
@@ -336,20 +336,28 @@ bool flash_program(uint32_t baseaddr, const uint8_t *buf, unsigned size) {
 }
 
 // Programs the built-in flash memory using the internal write buffer.
-bool flash_program_buffered(uint32_t baseaddr, const uint8_t *buf, unsigned size, unsigned bufsize) {
+bool flash_program_buffered(uint32_t baseaddr, const uint8_t *buf, unsigned size, unsigned bufsize, bool precopy) {
   // Reset any previous command that might be ongoing.
   FLASH_WE_MODE();
   SLOT2_BASE_U16[0] = 0x00F0;
   const unsigned wrsize = MIN(bufsize, 512);
+  union {
+    uint16_t b16[256];
+    uint32_t b32[128];
+  } tmp;
 
-  for (unsigned i = 0; i < size; i += 512) {
-    union {
-      uint16_t b16[256];
-      uint32_t b32[128];
-    } tmp;
+  if (precopy) {
     set_supercard_mode(MAPPED_SDRAM, true, true);
-    dma_memcpy32(tmp.b32, &buf[i], sizeof(tmp) / 4);
+    dma_memcpy32(tmp.b32, &buf[0], sizeof(tmp) / 4);
     FLASH_WE_MODE();
+  }
+
+  for (unsigned i = 0; i < size; i += 512) {
+    if (!precopy) {
+      set_supercard_mode(MAPPED_SDRAM, true, true);
+      dma_memcpy32(tmp.b32, &buf[i], sizeof(tmp) / 4);
+      FLASH_WE_MODE();
+    }
 
     for (unsigned off = 0; off < 512 && i+off < size; off += wrsize) {
       const uint32_t toff = (i + off);
@@ -366,6 +374,15 @@ bool flash_program_buffered(uint32_t baseaddr, const uint8_t *buf, unsigned size
 
       *(ptr-1) = 0x29;     // Confirm write buffer operation.
 
+      if (precopy) {
+        set_supercard_mode(MAPPED_SDRAM, true, true);
+        // off, off+wrsize is programmed, copy next chunk off, off+wrsize
+        unsigned next_chunk = i + 512;
+        if (next_chunk < size) {
+          dma_memcpy32(tmp.b32+(off/4), &buf[next_chunk + off], wrsize / 4);
+        }
+        FLASH_WE_MODE();
+      }
       // Wait a bit for the operation to finish.
       for (unsigned j = 0; j < 32*1024; j++) {
         if (SLOT2_BASE_U16[0] == SLOT2_BASE_U16[0])
diff --git a/src/loader.c b/src/loader.c
index c5d3b09..b0a9993 100644
--- a/src/loader.c
+++ b/src/loader.c
@@ -400,6 +400,8 @@ unsigned flash_gba_nor(
   // Map the game to the base 32MiB address space.
   set_superchis_normap(blkmap);
 
+  slowsd = use_slowld;
+
   t_flash_erase_state erst;
   for (uint32_t bigoff = 0; bigoff < fs; bigoff += ssize) {
     // Start clearing the flash block we will be writing to!
@@ -412,13 +414,14 @@ unsigned flash_gba_nor(
         if (progress)
           progress((bigoff + offset / 4) >> 8, fs >> 8);
 
-        flash_erase_fsm_step(&erst);
+        if (!slowsd) flash_erase_fsm_step(&erst);
       }
 
       unsigned toread = MIN(LOAD_BS, fs - absoff);
       UINT rdbytes;
       uint32_t tmp[LOAD_BS/4];
       if (FR_OK != f_read(&fd, tmp, toread, &rdbytes)) {
+        slowsd = true;
         f_close(&fd);
         reset_superchis_normap();
         return ERR_LOAD_BADROM;
@@ -457,7 +460,7 @@ unsigned flash_gba_nor(
         progress((bigoff + ssize / 4 + offset * 3/4) >> 8, fs >> 8);
 
       unsigned toflash = MIN(flashinfo.blksize, fs - absoff);
-      if (!flash_program_buffered(flashaddr, &scratch[offset], toflash, flashinfo.blkwrite)) {
+      if (!flash_program_buffered(flashaddr, &scratch[offset], toflash, flashinfo.blkwrite, !slowsd)) {
         f_close(&fd);
         reset_superchis_normap();
         return ERR_FLASH_OP;
@@ -465,6 +468,9 @@ unsigned flash_gba_nor(
     }
   }
 
+  slowsd = true;
+
+  f_close(&fd);
   reset_superchis_normap();
   return 0;
 }
diff --git a/src/menu.c b/src/menu.c
index fd352ad..5d077c2 100644
--- a/src/menu.c
+++ b/src/menu.c
@@ -2279,7 +2279,7 @@ void start_flash_update(const char *fn, unsigned fwsize, bool validate_superfw)
         bool programmed_ok;
         #ifdef SUPPORT_NORGAMES
         if (flashinfo.size && flashinfo.blksize && flashinfo.blkcount && flashinfo.blkwrite)
-          programmed_ok = flash_program_buffered(ROM_FLASHFIRMW_ADDR, sdr_state->scratch, fwsize, flashinfo.blkwrite);
+          programmed_ok = flash_program_buffered(ROM_FLASHFIRMW_ADDR, sdr_state->scratch, fwsize, flashinfo.blkwrite, !use_slowld);
         else
         #endif
           programmed_ok = flash_program(ROM_FLASHFIRMW_ADDR, sdr_state->scratch, fwsize);