diff --git a/general/package/goke-osdrv-gk7205v200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini b/general/package/goke-osdrv-gk7205v200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini
index 5ed3bed847..ccf6d5cda9 100644
--- a/general/package/goke-osdrv-gk7205v200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini
+++ b/general/package/goke-osdrv-gk7205v200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini
@@ -1,10 +1,31 @@
-; IMX335 boost full-frame mode, 2592x1944 @ 45 fps — full 5M readout
-; at 1.5x the stock 30 fps rate. Dispatched by Isp_SnsMode=6.
-; Useful when full sensor resolution is required at higher motion rate.
+; IMX335 boost full-frame mode, 2592x1944 — full 5M readout at the
+; sensor's boost rate (~40 fps internally vs the stock 30 fps).
+; Dispatched by Isp_FrameRate > 30 at 5M (auto-promoted inside
+; cmos_fps_set).
+;
+; This preset's value is **latency**, not wire fps. The encoder
+; pipeline caps wire-rate at ~21 fps for full-5M h.264 4 Mbps
+; regardless of video0.size (downscale doesn't help at 4:3 output).
+; But every delivered frame is fresher than stock 5M because:
+;   - sensor readout per frame is ~25 ms (vs ~33 ms stock 5M)
+;   - encoder picks a more recent frame from a 40 fps stream
+;     instead of a 30 fps stream
+; Combined: ~16 ms shorter glass-to-glass latency than stock 5M.
+; Useful for FPV / machine-vision where freshness > fps count.
+;
 ; clock=37.125MHz overrides gk's vendor blob default of 27 MHz —
 ; without it the sensor INCK is wrong and the rate is ~73% of ev's.
+;
+; To enable (output at full 4:3 native, max FoV):
+;   cli -s .isp.sensorConfig /etc/sensors/high-fps/imx335_2592x1944_45fps.ini
+;   cli -s .video0.size 2592x1944
+;   cli -s .video0.fps 45
+; killall -HUP majestic   # or reboot
+;
 ; Full per-mode notes:
 ;   openhisilicon/docs/imx335-v4-high-fps.md
+; How to measure the latency win (this claim is analytical, unverified):
+;   openhisilicon/docs/sensor-pipeline-latency-metering.md
 [sensor]
 Sensor_type=stSnsImx335Obj
 Mode=WDR_MODE_NONE
diff --git a/general/package/hisilicon-osdrv-hi3516ev200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini b/general/package/hisilicon-osdrv-hi3516ev200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini
index 1e0be37d34..b451fa33d8 100644
--- a/general/package/hisilicon-osdrv-hi3516ev200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini
+++ b/general/package/hisilicon-osdrv-hi3516ev200/files/sensor/config/high-fps/imx335_2592x1944_45fps.ini
@@ -1,8 +1,28 @@
-; IMX335 boost full-frame mode, 2592x1944 @ 45 fps — full 5M readout
-; at 1.5x the stock 30 fps rate. Dispatched by Isp_SnsMode=6.
-; Useful when full sensor resolution is required at higher motion rate.
+; IMX335 boost full-frame mode, 2592x1944 — full 5M readout at the
+; sensor's boost rate (~40 fps internally vs the stock 30 fps).
+; Dispatched by Isp_FrameRate > 30 at 5M (auto-promoted inside
+; cmos_fps_set).
+;
+; This preset's value is **latency**, not wire fps. The encoder
+; pipeline caps wire-rate at ~21 fps for full-5M h.264 4 Mbps
+; regardless of video0.size (downscale doesn't help at 4:3 output).
+; But every delivered frame is fresher than stock 5M because:
+;   - sensor readout per frame is ~25 ms (vs ~33 ms stock 5M)
+;   - encoder picks a more recent frame from a 40 fps stream
+;     instead of a 30 fps stream
+; Combined: ~16 ms shorter glass-to-glass latency than stock 5M.
+; Useful for FPV / machine-vision where freshness > fps count.
+;
+; To enable (output at full 4:3 native, max FoV):
+;   cli -s .isp.sensorConfig /etc/sensors/high-fps/imx335_2592x1944_45fps.ini
+;   cli -s .video0.size 2592x1944
+;   cli -s .video0.fps 45
+; killall -HUP majestic   # or reboot
+;
 ; Full per-mode notes:
 ;   openhisilicon/docs/imx335-v4-high-fps.md
+; How to measure the latency win (this claim is analytical, unverified):
+;   openhisilicon/docs/sensor-pipeline-latency-metering.md
 [sensor]
 Sensor_type=stSnsImx335Obj
 Mode=WDR_MODE_NONE