diff --git a/subprojects/gst-plugins-bad/gst-libs/gst/va/gstvaallocator.c b/subprojects/gst-plugins-bad/gst-libs/gst/va/gstvaallocator.c
index e418e6be07..b9a38684c6 100644
--- a/subprojects/gst-plugins-bad/gst-libs/gst/va/gstvaallocator.c
+++ b/subprojects/gst-plugins-bad/gst-libs/gst/va/gstvaallocator.c
@@ -1372,13 +1372,6 @@ _va_map_unlocked (GstVaMemory * mem, GstMapFlags flags)
     use_derived = FALSE;
   } else {
     switch (gst_va_display_get_implementation (display)) {
-      case GST_VA_IMPLEMENTATION_INTEL_IHD:
-        /* On Gen7+ Intel graphics the memory is mappable but not
-         * cached, so normal memcpy() access is very slow to read, but
-         * it's ok for writing. So let's assume that users won't prefer
-         * direct-mapped memory if they request read access. */
-        use_derived = va_allocator->use_derived && !(flags & GST_MAP_READ);
-        break;
       case GST_VA_IMPLEMENTATION_INTEL_I965:
         /* YUV derived images are tiled, so writing them is also
          * problematic */
@@ -1386,12 +1379,6 @@ _va_map_unlocked (GstVaMemory * mem, GstMapFlags flags)
             || ((flags & GST_MAP_WRITE)
                 && GST_VIDEO_INFO_IS_YUV (&va_allocator->info)));
         break;
-      case GST_VA_IMPLEMENTATION_MESA_GALLIUM:
-        /* Reading RGB derived images, with non-standard resolutions,
-         * looks like tiled too. TODO(victor): fill a bug in Mesa. */
-        use_derived = va_allocator->use_derived && !((flags & GST_MAP_READ)
-            && GST_VIDEO_INFO_IS_RGB (&va_allocator->info));
-        break;
       default:
         use_derived = va_allocator->use_derived;
         break;