diff --git a/ports/raspberrypi/Makefile b/ports/raspberrypi/Makefile
index 6348dc941f7..13499baf0f7 100755
--- a/ports/raspberrypi/Makefile
+++ b/ports/raspberrypi/Makefile
@@ -217,7 +217,10 @@ CFLAGS += \
 	-DPICO_NO_BI_BOOTSEL_VIA_DOUBLE_RESET=0 \
 	-DDVI_1BPP_BIT_REVERSE=0
 
-OPTIMIZATION_FLAGS ?= -O3
+# -O2 plus the two -O3 flags that measure: -funswitch-loops (16 KB) carries the sprite
+# blits, -fvect-cost-model=dynamic (4 KB) the fills and memcpy. Same speed as -O3 on both
+# cores for about 125 KB less flash. Boards may override this in mpconfigboard.mk.
+OPTIMIZATION_FLAGS ?= -O2 -funswitch-loops -fvect-cost-model=dynamic
 # TinyUSB defines
 CFLAGS += \
 	-DCFG_TUSB_OS=OPT_OS_PICO \
diff --git a/ports/raspberrypi/boards/pajenicko_picopad/mpconfigboard.mk b/ports/raspberrypi/boards/pajenicko_picopad/mpconfigboard.mk
index b5df0e6ecb5..03a203dc2de 100644
--- a/ports/raspberrypi/boards/pajenicko_picopad/mpconfigboard.mk
+++ b/ports/raspberrypi/boards/pajenicko_picopad/mpconfigboard.mk
@@ -50,8 +50,3 @@ CFLAGS += \
 
 # Must be accompanied by a linker script change
 CFLAGS += -DCIRCUITPY_FIRMWARE_SIZE='(1536 * 1024)'
-
-# The rp2 port default is -O3; on this Cortex-M0+ (no SIMD/FPU, 16 KB XIP cache) -O2 plus
-# these five loop passes measures within +-1% of -O3 across the picogame render kernels
-# while using ~150 KB less flash (gc.o/vm.o stay -O3 via SUPEROPT regardless).
-OPTIMIZATION_FLAGS = -O2 -funswitch-loops -fpredictive-commoning -fgcse-after-reload -ftree-partial-pre -fsplit-paths
diff --git a/ports/raspberrypi/boards/pimoroni_picosystem/mpconfigboard.mk b/ports/raspberrypi/boards/pimoroni_picosystem/mpconfigboard.mk
index 28a3c6a641a..38ee11df70b 100644
--- a/ports/raspberrypi/boards/pimoroni_picosystem/mpconfigboard.mk
+++ b/ports/raspberrypi/boards/pimoroni_picosystem/mpconfigboard.mk
@@ -18,7 +18,3 @@ CIRCUITPY_PICOGAME_FAST_DISPLAY = 1
 CIRCUITPY_AUDIOIO = 1
 
 FROZEN_MPY_DIRS += $(TOP)/frozen/circuitpython-stage/picosystem
-
-# The default is -O3. picogame does not fit at -O3; these loop passes keep the render
-# kernels within 1% of it.
-OPTIMIZATION_FLAGS = -O2 -funswitch-loops -fpredictive-commoning -fgcse-after-reload -ftree-partial-pre -fsplit-paths
diff --git a/ports/raspberrypi/boards/raspberry_pi_pico/mpconfigboard.mk b/ports/raspberrypi/boards/raspberry_pi_pico/mpconfigboard.mk
index 20607d50e41..c53ddbbd7e4 100644
--- a/ports/raspberrypi/boards/raspberry_pi_pico/mpconfigboard.mk
+++ b/ports/raspberrypi/boards/raspberry_pi_pico/mpconfigboard.mk
@@ -9,4 +9,8 @@ CHIP_FAMILY = rp2
 EXTERNAL_FLASH_DEVICES = "W25Q16JVxQ"
 
 CIRCUITPY__EVE = 1
+
+CIRCUITPY_PICOGAME = 1
+CIRCUITPY_PICOGAME_FAST_DISPLAY = 1
+
 CIRCUITPY_PICODVI = 1
diff --git a/ports/raspberrypi/boards/raspberry_pi_pico_w/mpconfigboard.mk b/ports/raspberrypi/boards/raspberry_pi_pico_w/mpconfigboard.mk
index e658db65f4d..bb7a9208899 100644
--- a/ports/raspberrypi/boards/raspberry_pi_pico_w/mpconfigboard.mk
+++ b/ports/raspberrypi/boards/raspberry_pi_pico_w/mpconfigboard.mk
@@ -12,6 +12,9 @@ CIRCUITPY_USB_HOST = 0
 
 CIRCUITPY__EVE = 1
 
+CIRCUITPY_PICOGAME = 1
+CIRCUITPY_PICOGAME_FAST_DISPLAY = 1
+
 CIRCUITPY_CYW43 = 1
 CIRCUITPY_SSL = 1
 CIRCUITPY_HASHLIB = 1