diff --git a/.travis.yml b/.travis.yml
deleted file mode 100644
index f34e348..0000000
--- a/.travis.yml
+++ /dev/null
@@ -1,32 +0,0 @@
-language: generic
-os: linux
-dist: trusty
-sudo: required
-addons:
-  apt:
-    packages:
-      - g++-7
-    sources:
-      - ubuntu-toolchain-r-test
-env:
-  global:
-    - CORE=sameboy
-    - COMPILER_NAME=gcc CXX=g++-7 CC=gcc-7
-  matrix:
-    - PLATFORM=linux_x64
-before_script:
-  - pwd
-  - mkdir -p ~/bin
-  - ln -s /usr/bin/gcc-7 ~/bin/gcc
-  - ln -s /usr/bin/g++-7 ~/bin/g++
-  - ln -s /usr/bin/cpp-7 ~/bin/cpp
-  - export PATH=~/bin:$PATH
-  - ls -l ~/bin
-  - echo $PATH
-  - g++-7 --version
-  - g++ --version
-script:
-  - cd ~/
-  - git clone --depth=50 https://github.com/libretro/libretro-super
-  - cd libretro-super/travis
-  - ./build.sh
diff --git a/BootROMs/dmg_boot.asm b/BootROMs/dmg_boot.asm
index e22dee0..d4176a8 100644
--- a/BootROMs/dmg_boot.asm
+++ b/BootROMs/dmg_boot.asm
@@ -80,14 +80,14 @@ Start:
     ; Play first sound
     ld a, $83
     call PlaySound
-    ld b, 15
+    ld b, 5
     call WaitBFrames
     ; Play second sound
     ld a, $c1
     call PlaySound
 
-; Wait ~2.5 seconds
-    ld b, 150
+; Wait ~1.15 seconds
+    ld b, 70
     call WaitBFrames
     
 ; Set registers to match the original DMG boot
@@ -123,9 +123,13 @@ DoubleBitsAndWriteRow:
     ret
 
 WaitFrame:
-    ldh a, [$44]
-    cp $90
-    jr nz, WaitFrame
+    push hl
+    ld hl, $FF0F
+  res 0, [hl]
+.wait
+    bit 0, [hl]
+    jr z, .wait
+    pop hl
     ret
 
 WaitBFrames:
diff --git a/BootROMs/prebuilt/cgb_boot.bin b/BootROMs/prebuilt/cgb_boot.bin
deleted file mode 100644
index 42d88fc..0000000
Binary files a/BootROMs/prebuilt/cgb_boot.bin and /dev/null differ
diff --git a/BootROMs/prebuilt/dmg_boot.bin b/BootROMs/prebuilt/dmg_boot.bin
deleted file mode 100644
index a7b51d4..0000000
Binary files a/BootROMs/prebuilt/dmg_boot.bin and /dev/null differ
diff --git a/Core/apu.c b/Core/apu.c
index c74bb6d..8d5b3e7 100755
--- a/Core/apu.c
+++ b/Core/apu.c
@@ -27,11 +27,11 @@ static void update_sample(GB_gameboy_t *gb, unsigned index, uint8_t value, unsig
     if (gb->apu_output.sample_rate) {
         unsigned left_volume = 0;
         if (gb->io_registers[GB_IO_NR51] & (1 << index)) {
-            left_volume = gb->io_registers[GB_IO_NR50] & 7;
+            left_volume = (gb->io_registers[GB_IO_NR50] & 7) + 1;
         }
         unsigned right_volume = 0;
         if (gb->io_registers[GB_IO_NR51] & (0x10 << index)) {
-            right_volume = (gb->io_registers[GB_IO_NR50] >> 4) & 7;
+            right_volume = ((gb->io_registers[GB_IO_NR50] >> 4) & 7) + 1;
         }
         GB_sample_t output = {(0xf - value) * left_volume, (0xf - value) * right_volume};
         if (*(uint32_t *)&(gb->apu_output.current_sample[index]) != *(uint32_t *)&output) {
@@ -575,6 +575,14 @@ void GB_apu_write(GB_gameboy_t *gb, uint8_t reg, uint8_t value)
                     gb->apu.square_channels[index].sample_countdown = (gb->apu.square_channels[index].sample_length ^ 0x7FF) * 2 + 4 - gb->apu.lf_div;
                 }
                 gb->apu.square_channels[index].current_volume = gb->io_registers[index == GB_SQUARE_1 ? GB_IO_NR12 : GB_IO_NR22] >> 4;
+                
+                /* The volume changes caused by NRX4 sound start take effect instantly (i.e. the effect the previously
+                   started sound). The playback itself is not instant which is why we don't update the sample for other
+                   cases. */
+                if (gb->apu.is_active[index]) {
+                    update_square_sample(gb, index);
+                }
+                
                 gb->apu.square_channels[index].volume_countdown = gb->io_registers[index == GB_SQUARE_1 ? GB_IO_NR12 : GB_IO_NR22] & 7;
                 
                 if ((gb->io_registers[index == GB_SQUARE_1 ? GB_IO_NR12 : GB_IO_NR22] & 0xF8) != 0) {
@@ -599,7 +607,6 @@ void GB_apu_write(GB_gameboy_t *gb, uint8_t reg, uint8_t value)
                     if (!gb->apu.square_sweep_countdown) gb->apu.square_sweep_countdown = 8;
                 }
                 
-                /* Note that we don't change the sample just yet! This was verified on hardware. */
             }
             
             /* APU glitch - if length is enabled while the DIV-divider's LSB is 1, tick the length once. */
@@ -718,7 +725,10 @@ void GB_apu_write(GB_gameboy_t *gb, uint8_t reg, uint8_t value)
             }
             else if (gb->apu.is_active[GB_NOISE]){
                 nrx2_glitch(&gb->apu.noise_channel.current_volume, value, gb->io_registers[reg]);
-                update_square_sample(gb, GB_NOISE);
+                update_sample(gb, GB_NOISE,
+                              (gb->apu.noise_channel.lfsr & 1) ?
+                              gb->apu.noise_channel.current_volume : 0,
+                              0);
             }
             break;
         }
@@ -757,6 +767,17 @@ void GB_apu_write(GB_gameboy_t *gb, uint8_t reg, uint8_t value)
                 }
                 
                 gb->apu.noise_channel.current_volume = gb->io_registers[GB_IO_NR42] >> 4;
+                
+                /* The volume changes caused by NRX4 sound start take effect instantly (i.e. the effect the previously
+                 started sound). The playback itself is not instant which is why we don't update the sample for other
+                 cases. */
+                if (gb->apu.is_active[GB_NOISE]) {
+                    update_sample(gb, GB_NOISE,
+                                  (gb->apu.noise_channel.lfsr & 1) ?
+                                  gb->apu.noise_channel.current_volume : 0,
+                                  0);
+                }
+                
                 gb->apu.noise_channel.volume_countdown = gb->io_registers[GB_IO_NR42] & 7;
                 
                 if ((gb->io_registers[GB_IO_NR42] & 0xF8) != 0) {
@@ -767,8 +788,6 @@ void GB_apu_write(GB_gameboy_t *gb, uint8_t reg, uint8_t value)
                     gb->apu.noise_channel.pulse_length = 0x40;
                     gb->apu.noise_channel.length_enabled = false;
                 }
-                
-                /* Note that we don't change the sample just yet! This was verified on hardware. */
             }
             
             /* APU glitch - if length is enabled while the DIV-divider's LSB is 1, tick the length once. */
diff --git a/Core/apu.h b/Core/apu.h
index bace09a..44b3d45 100644
--- a/Core/apu.h
+++ b/Core/apu.h
@@ -6,9 +6,9 @@
 
 
 #ifdef GB_INTERNAL
-/* Divides nicely and never overflows with 4 channels and 8 volume levels */
-#define MAX_CH_AMP 0x1FFE
-#define CH_STEP (MAX_CH_AMP/0xF/7)
+/* Divides nicely and never overflows with 4 channels and 8 (1-8) volume levels */
+#define MAX_CH_AMP 0x1FE0
+#define CH_STEP (MAX_CH_AMP/0xF/8)
 #endif
 
 /* APU ticks are 2MHz, triggered by an internal APU clock. */
diff --git a/SDL/drop.bmp b/SDL/drop.bmp
deleted file mode 100644
index fbdc374..0000000
Binary files a/SDL/drop.bmp and /dev/null differ
diff --git a/libretro/Makefile b/libretro/Makefile
index feb19bd..ff62b7a 100644
--- a/libretro/Makefile
+++ b/libretro/Makefile
@@ -6,6 +6,8 @@ ifneq ($(GIT_VERSION)," unknown")
    CFLAGS += -DGIT_VERSION=\"$(GIT_VERSION)\"
 endif
 
+CFLAGS += -DSAMEBOY_CORE_VERSION=\"$(VERSION)\"
+
 ifeq ($(platform),)
 platform = unix
 ifeq ($(shell uname -a),)
@@ -139,7 +141,7 @@ CFLAGS   += -Wall -D__LIBRETRO__ $(fpic) $(INCFLAGS) -std=gnu11 -D_GNU_SOURCE -D
 
 all: $(TARGET)
 
-$(CORE_DIR)/libretro/%_boot.c: $(CORE_DIR)/BootROMs/prebuilt/%_boot.bin
+$(CORE_DIR)/libretro/%_boot.c: $(CORE_DIR)/build/bin/BootROMs/%_boot.bin
 	echo "/* AUTO-GENERATED */" > $@
 	echo "const unsigned char $(notdir $(@:%.c=%))[] = {" >> $@
 	hexdump -v -e '/1 "0x%02x, "' $< >> $@