diff --git a/Src/Base/AMReX_Arena.cpp b/Src/Base/AMReX_Arena.cpp
index 19f4f6f5459..2082014d634 100644
--- a/Src/Base/AMReX_Arena.cpp
+++ b/Src/Base/AMReX_Arena.cpp
@@ -305,7 +305,7 @@ Arena::Initialize ()
     the_async_arena = new PArena(the_async_arena_release_threshold);
 
 #ifdef AMREX_USE_GPU
-    if (the_arena->isDevice() || the_arena->isManaged()) {
+    if (the_arena->isDevice()) {
         the_device_arena = the_arena;
     } else {
         the_device_arena = new CArena(0, ArenaInfo{}.SetDeviceMemory().SetReleaseThreshold
diff --git a/Src/Base/AMReX_GpuContainers.H b/Src/Base/AMReX_GpuContainers.H
index 0992411b552..faccec1d2ef 100644
--- a/Src/Base/AMReX_GpuContainers.H
+++ b/Src/Base/AMReX_GpuContainers.H
@@ -19,9 +19,8 @@ namespace Gpu {
 
     /**
      * \brief A PODVector that uses the standard memory Arena.
-     * Note that, on NVIDIA architectures, this Arena is actually
-     * managed.
-     *
+     * Note that the memory might or might not be managed depending
+     * on the amrex.the_arena_is_managed ParmParse parameter.
      */
     template <class T>
     using DeviceVector = PODVector<T, ArenaAllocator<T> >;