diff --git a/cub/cub/util_math.cuh b/cub/cub/util_math.cuh
index 60c81364b03..9578c84319b 100644
--- a/cub/cub/util_math.cuh
+++ b/cub/cub/util_math.cuh
@@ -72,23 +72,6 @@ _CCCL_HOST_DEVICE _CCCL_FORCEINLINE OffsetT safe_add_bound_to_max(OffsetT lhs, O
 
 } // namespace detail
 
-/**
- * Divide n by d, round up if any remainder, and return the result.
- *
- * Effectively performs `(n + d - 1) / d`, but is robust against the case where
- * `(n + d - 1)` would overflow.
- * deprecated [Since 2.8.0] `cub::DivideAndRoundUp` is deprecated. Use `cuda::ceil_div` instead.
- */
-template <typename NumeratorT, typename DenominatorT>
-CCCL_DEPRECATED_BECAUSE("Use cuda::ceil_div instead")
-_CCCL_HOST_DEVICE _CCCL_FORCEINLINE constexpr NumeratorT DivideAndRoundUp(NumeratorT n, DenominatorT d)
-{
-  static_assert(
-    cub::detail::is_integral_or_enum<NumeratorT>::value && cub::detail::is_integral_or_enum<DenominatorT>::value,
-    "DivideAndRoundUp is only intended for integral types.");
-  return ::cuda::ceil_div(n, d);
-}
-
 constexpr _CCCL_HOST_DEVICE int Nominal4BItemsToItemsCombined(int nominal_4b_items_per_thread, int combined_bytes)
 {
   return (cub::min)(nominal_4b_items_per_thread, (cub::max)(1, nominal_4b_items_per_thread * 8 / combined_bytes));