diff --git a/src/amd/common/ac_nir.c b/src/amd/common/ac_nir.c
index 2503f5acd0c..d06207804bd 100644
--- a/src/amd/common/ac_nir.c
+++ b/src/amd/common/ac_nir.c
@@ -250,10 +250,11 @@ ac_nir_export_position(nir_builder *b,
     * the pixel shader starts.
     */
    if (gfx_level >= GFX10 && no_param_export && b->shader->info.writes_memory) {
-      nir_intrinsic_instr *wait_instr =
-         nir_intrinsic_instr_create(b->shader, nir_intrinsic_memory_barrier_buffer);
-
-      nir_instr_insert_before(&final_exp->instr, &wait_instr->instr);
+      nir_cursor cursor = b->cursor;
+      b->cursor = nir_before_instr(&final_exp->instr);
+      nir_scoped_memory_barrier(b, NIR_SCOPE_DEVICE, NIR_MEMORY_RELEASE,
+                                nir_var_mem_ssbo | nir_var_mem_global | nir_var_image);
+      b->cursor = cursor;
    }
 }
 
diff --git a/src/amd/common/ac_nir_lower_ngg.c b/src/amd/common/ac_nir_lower_ngg.c
index a5ab1d4c184..0799dbc9956 100644
--- a/src/amd/common/ac_nir_lower_ngg.c
+++ b/src/amd/common/ac_nir_lower_ngg.c
@@ -2012,6 +2012,7 @@ ngg_build_streamout_vertex(nir_builder *b, nir_xfb_info *info,
                            vtx_buffer_offsets[out->buffer],
                            zero, zero,
                            .base = out->offset,
+                           .memory_modes = nir_var_mem_ssbo,
                            .access = ACCESS_NON_TEMPORAL);
    }
 }
@@ -2071,7 +2072,7 @@ ngg_nogs_build_streamout(nir_builder *b, lower_ngg_nogs_state *s)
     * TODO: not sure if we need this barrier when late prim export, as I
     *       can't observe test fail without this barrier.
     */
-   nir_memory_barrier_buffer(b);
+   nir_scoped_memory_barrier(b, NIR_SCOPE_DEVICE, NIR_MEMORY_RELEASE, nir_var_mem_ssbo);
 }
 
 static unsigned