diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index 11dca552356..a04c4a3f9fc 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -94,6 +94,7 @@ static const struct debug_named_value debug_options[] = {
    /* 3D engine options: */
    {"nogfx", DBG(NO_GFX), "Disable graphics. Only multimedia compute paths can be used."},
    {"nongg", DBG(NO_NGG), "Disable NGG and use the legacy pipeline."},
+   {"nofastlaunch", DBG(NO_FAST_LAUNCH), "Disable NGG GS fast launch."},
    {"nggc", DBG(ALWAYS_NGG_CULLING_ALL), "Always use NGG culling even when it can hurt."},
    {"nggctess", DBG(ALWAYS_NGG_CULLING_TESS), "Always use NGG culling for tessellation."},
    {"nonggc", DBG(NO_NGG_CULLING), "Disable NGG culling."},
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 20230556744..d14fe6916dd 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -191,6 +191,7 @@ enum
    DBG_ALWAYS_NGG_CULLING_ALL,
    DBG_ALWAYS_NGG_CULLING_TESS,
    DBG_NO_NGG_CULLING,
+   DBG_NO_FAST_LAUNCH,
    DBG_ALWAYS_PD,
    DBG_PD,
    DBG_NO_PD,
diff --git a/src/gallium/drivers/radeonsi/si_shader.h b/src/gallium/drivers/radeonsi/si_shader.h
index 767247b75b6..1f06324e562 100644
--- a/src/gallium/drivers/radeonsi/si_shader.h
+++ b/src/gallium/drivers/radeonsi/si_shader.h
@@ -433,8 +433,8 @@ struct si_shader_selector {
    ubyte num_vs_inputs;
    ubyte num_vbos_in_user_sgprs;
    unsigned pa_cl_vs_out_cntl;
-   unsigned ngg_cull_vert_threshold; /* 0 = disabled */
-   unsigned ngg_cull_nonindexed_fast_launch_vert_threshold; /* 0 = disabled */
+   unsigned ngg_cull_vert_threshold; /* UINT32_MAX = disabled */
+   unsigned ngg_cull_nonindexed_fast_launch_vert_threshold; /* UINT32_MAX = disabled */
    ubyte clipdist_mask;
    ubyte culldist_mask;
    ubyte rast_prim;
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index 2acb1337b85..831555a0f54 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -2169,7 +2169,9 @@ static void si_draw_vbo(struct pipe_context *ctx,
       /* Use NGG fast launch for certain primitive types.
        * A draw must have at least 1 full primitive.
        */
-      if (ngg_culling && min_direct_count >= 3 && !sctx->tes_shader.cso &&
+      if (ngg_culling &&
+          hw_vs->ngg_cull_nonindexed_fast_launch_vert_threshold < UINT32_MAX &&
+          min_direct_count >= 3 && !sctx->tes_shader.cso &&
           !sctx->gs_shader.cso) {
          if (prim == PIPE_PRIM_TRIANGLES && !index_size) {
             ngg_culling |= SI_NGG_CULL_GS_FAST_LAUNCH_TRI_LIST;
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index f822229d836..e57727bb4d3 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -2817,7 +2817,8 @@ static void *si_create_shader_selector(struct pipe_context *ctx,
          /* 1000 non-indexed vertices (roughly 8 primgroups) are needed
           * per draw call (no TES/GS) to enable NGG culling by default.
           */
-         sel->ngg_cull_nonindexed_fast_launch_vert_threshold = 1000;
+         if (!(sscreen->debug_flags & DBG(NO_FAST_LAUNCH)))
+            sel->ngg_cull_nonindexed_fast_launch_vert_threshold = 1000;
 
          if (sscreen->debug_flags & DBG(ALWAYS_NGG_CULLING_ALL))
             sel->ngg_cull_vert_threshold = 0; /* always enabled */