diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 399fbc7551d..300e320fd2a 100644
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -7823,20 +7823,18 @@ visit_emit_vertex_with_counter(isel_context* ctx, nir_intrinsic_instr* instr)
                const_offset %= 4096u;
             }
 
-            aco_ptr<MTBUF_instruction> mtbuf{create_instruction<MTBUF_instruction>(
-               aco_opcode::tbuffer_store_format_x, Format::MTBUF, 4, 0)};
-            mtbuf->operands[0] = Operand(gsvs_ring);
-            mtbuf->operands[1] = vaddr_offset;
-            mtbuf->operands[2] = Operand(get_arg(ctx, ctx->args->ac.gs2vs_offset));
-            mtbuf->operands[3] = Operand(ctx->outputs.temps[i * 4u + j]);
-            mtbuf->offen = !vaddr_offset.isUndefined();
-            mtbuf->dfmt = V_008F0C_BUF_DATA_FORMAT_32;
-            mtbuf->nfmt = V_008F0C_BUF_NUM_FORMAT_UINT;
-            mtbuf->offset = const_offset;
-            mtbuf->glc = ctx->program->gfx_level < GFX11;
-            mtbuf->slc = true;
-            mtbuf->sync = memory_sync_info(storage_vmem_output, semantic_can_reorder);
-            bld.insert(std::move(mtbuf));
+            aco_ptr<MUBUF_instruction> mubuf{create_instruction<MUBUF_instruction>(
+               aco_opcode::buffer_store_dword, Format::MUBUF, 4, 0)};
+            mubuf->operands[0] = Operand(gsvs_ring);
+            mubuf->operands[1] = vaddr_offset;
+            mubuf->operands[2] = Operand(get_arg(ctx, ctx->args->ac.gs2vs_offset));
+            mubuf->operands[3] = Operand(ctx->outputs.temps[i * 4u + j]);
+            mubuf->offen = !vaddr_offset.isUndefined();
+            mubuf->offset = const_offset;
+            mubuf->glc = ctx->program->gfx_level < GFX11;
+            mubuf->slc = true;
+            mubuf->sync = memory_sync_info(storage_vmem_output, semantic_can_reorder);
+            bld.insert(std::move(mubuf));
          }
 
          offset += ctx->shader->info.gs.vertices_out;
diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
index 1a3bb842d93..7c44781ad6f 100644
--- a/src/amd/vulkan/radv_device.c
+++ b/src/amd/vulkan/radv_device.c
@@ -4061,6 +4061,10 @@ radv_fill_shader_rings(struct radv_device *device, uint32_t *map, bool add_sampl
       } else if (device->physical_device->rad_info.gfx_level >= GFX10) {
          desc[3] |= S_008F0C_FORMAT(V_008F0C_GFX10_FORMAT_32_FLOAT) |
                     S_008F0C_OOB_SELECT(V_008F0C_OOB_SELECT_DISABLED) | S_008F0C_RESOURCE_LEVEL(1);
+      } else if (device->physical_device->rad_info.gfx_level >= GFX8) {
+         /* DATA_FORMAT is STRIDE[14:17] for MUBUF with ADD_TID_ENABLE=1 */
+         desc[3] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+                    S_008F0C_DATA_FORMAT(0) | S_008F0C_ELEMENT_SIZE(1);
       } else {
          desc[3] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
                     S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) | S_008F0C_ELEMENT_SIZE(1);
@@ -4133,6 +4137,10 @@ radv_fill_shader_rings(struct radv_device *device, uint32_t *map, bool add_sampl
       } else if (device->physical_device->rad_info.gfx_level >= GFX10) {
          desc[7] |= S_008F0C_FORMAT(V_008F0C_GFX10_FORMAT_32_FLOAT) |
                     S_008F0C_OOB_SELECT(V_008F0C_OOB_SELECT_DISABLED) | S_008F0C_RESOURCE_LEVEL(1);
+      } else if (device->physical_device->rad_info.gfx_level >= GFX8) {
+         /* DATA_FORMAT is STRIDE[14:17] for MUBUF with ADD_TID_ENABLE=1 */
+         desc[7] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+                    S_008F0C_DATA_FORMAT(0) | S_008F0C_ELEMENT_SIZE(1);
       } else {
          desc[7] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
                     S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) | S_008F0C_ELEMENT_SIZE(1);