diff --git a/wgpu-core/src/device/queue.rs b/wgpu-core/src/device/queue.rs
index d3d0e67704..d9a613fe1d 100644
--- a/wgpu-core/src/device/queue.rs
+++ b/wgpu-core/src/device/queue.rs
@@ -239,10 +239,11 @@ impl<G: GlobalIdentityHandlerFactory> Global<G> {
             .bits as u32
             / BITS_PER_BYTE;
 
-        let stage_bytes_per_row = get_lowest_common_denom(
+        let bytes_per_row_alignment = get_lowest_common_denom(
             device.hal_limits.optimal_buffer_copy_pitch_alignment as u32,
             bytes_per_texel,
         );
+        let stage_bytes_per_row = align_to(bytes_per_texel * size.width, bytes_per_row_alignment);
         let stage_size = stage_bytes_per_row as u64
             * ((size.depth - 1) * data_layout.rows_per_image + size.height) as u64;
         let mut stage = device.prepare_stage(stage_size);
@@ -532,6 +533,13 @@ fn get_greatest_common_divisor(mut a: u32, mut b: u32) -> u32 {
     }
 }
 
+fn align_to(value: u32, alignment: u32) -> u32 {
+    match value % alignment {
+        0 => value,
+        other => value - other + alignment,
+    }
+}
+
 #[test]
 fn test_lcd() {
     assert_eq!(get_lowest_common_denom(2, 2), 2);