diff --git a/src/freedreno/vulkan/tu_cmd_buffer.c b/src/freedreno/vulkan/tu_cmd_buffer.c
index 5fd8c16ce7c..17ae859f621 100644
--- a/src/freedreno/vulkan/tu_cmd_buffer.c
+++ b/src/freedreno/vulkan/tu_cmd_buffer.c
@@ -556,41 +556,23 @@ tu_cs_emit_draw_state(struct tu_cs *cs, uint32_t id, struct tu_draw_state state)
    tu_cs_emit_qw(cs, state.iova);
 }
 
-static bool
-is_hw_binning_possible(const struct tu_framebuffer *fb)
-{
-   /* Similar to older gens, # of tiles per pipe cannot be more than 32.
-    * But there are no hangs with 16 or more tiles per pipe in either
-    * X or Y direction, so that limit does not seem to apply.
-    */
-   uint32_t tiles_per_pipe = fb->pipe0.width * fb->pipe0.height;
-   return tiles_per_pipe <= 32;
-}
-
 static bool
 use_hw_binning(struct tu_cmd_buffer *cmd)
 {
    const struct tu_framebuffer *fb = cmd->state.framebuffer;
 
-   /* XFB commands are emitted for BINNING || SYSMEM, which makes it incompatible
-    * with non-hw binning GMEM rendering. this is required because some of the
-    * XFB commands need to only be executed once
+   /* XFB commands are emitted for BINNING || SYSMEM, which makes it
+    * incompatible with non-hw binning GMEM rendering. this is required because
+    * some of the XFB commands need to only be executed once.
+    * use_sysmem_rendering() should have made sure we only ended up here if no
+    * XFB was used.
     */
    if (cmd->state.xfb_used) {
-      assert(is_hw_binning_possible(fb));
+      assert(fb->binning_possible);
       return true;
    }
 
-   if (!is_hw_binning_possible(fb))
-      return false;
-
-   if (unlikely(cmd->device->physical_device->instance->debug_flags & TU_DEBUG_NOBIN))
-      return false;
-
-   if (unlikely(cmd->device->physical_device->instance->debug_flags & TU_DEBUG_FORCEBIN))
-      return true;
-
-   return (fb->tile_count.width * fb->tile_count.height) > 2;
+   return fb->binning;
 }
 
 static bool
@@ -619,7 +601,7 @@ use_sysmem_rendering(struct tu_cmd_buffer *cmd,
       return true;
 
    /* XFB is incompatible with non-hw binning GMEM rendering, see use_hw_binning */
-   if (cmd->state.xfb_used && !is_hw_binning_possible(cmd->state.framebuffer))
+   if (cmd->state.xfb_used && !cmd->state.framebuffer->binning_possible)
       return true;
 
    if (unlikely(cmd->device->physical_device->instance->debug_flags & TU_DEBUG_GMEM))
diff --git a/src/freedreno/vulkan/tu_private.h b/src/freedreno/vulkan/tu_private.h
index 2c9f66b71e5..2e8110c93b2 100644
--- a/src/freedreno/vulkan/tu_private.h
+++ b/src/freedreno/vulkan/tu_private.h
@@ -1816,6 +1816,12 @@ struct tu_framebuffer
    /* number of VSC pipes */
    VkExtent2D pipe_count;
 
+   /* Whether binning should be used for gmem rendering using this framebuffer. */
+   bool binning;
+
+   /* Whether binning could be used for gmem rendering using this framebuffer. */
+   bool binning_possible;
+
    /* pipe register values */
    uint32_t pipe_config[MAX_VSC_PIPES];
    uint32_t pipe_sizes[MAX_VSC_PIPES];
diff --git a/src/freedreno/vulkan/tu_util.c b/src/freedreno/vulkan/tu_util.c
index 2d54af74e61..5be9100702b 100644
--- a/src/freedreno/vulkan/tu_util.c
+++ b/src/freedreno/vulkan/tu_util.c
@@ -208,6 +208,35 @@ tu_tiling_config_update_pipes(struct tu_framebuffer *fb,
           sizeof(uint32_t) * (max_pipe_count - used_pipe_count));
 }
 
+static bool
+is_hw_binning_possible(const struct tu_framebuffer *fb)
+{
+   /* Similar to older gens, # of tiles per pipe cannot be more than 32.
+    * But there are no hangs with 16 or more tiles per pipe in either
+    * X or Y direction, so that limit does not seem to apply.
+    */
+   uint32_t tiles_per_pipe = fb->pipe0.width * fb->pipe0.height;
+   return tiles_per_pipe <= 32;
+}
+
+static void
+tu_tiling_config_update_binning(struct tu_framebuffer *fb, const struct tu_device *device)
+{
+   fb->binning_possible = is_hw_binning_possible(fb);
+
+   if (fb->binning_possible) {
+      fb->binning = (fb->tile_count.width * fb->tile_count.height) > 2;
+
+      if (unlikely(device->physical_device->instance->debug_flags & TU_DEBUG_FORCEBIN))
+         fb->binning = true;
+      if (unlikely(device->physical_device->instance->debug_flags &
+                   TU_DEBUG_NOBIN))
+         fb->binning = false;
+   } else {
+      fb->binning = false;
+   }
+}
+
 void
 tu_framebuffer_tiling_config(struct tu_framebuffer *fb,
                              const struct tu_device *device,
@@ -216,6 +245,7 @@ tu_framebuffer_tiling_config(struct tu_framebuffer *fb,
    tu_tiling_config_update_tile_layout(fb, device, pass);
    tu_tiling_config_update_pipe_layout(fb, device);
    tu_tiling_config_update_pipes(fb, device);
+   tu_tiling_config_update_binning(fb, device);
 }
 
 void
@@ -268,4 +298,4 @@ tu_dbg_log_gmem_load_store_skips(struct tu_device *device)
    last_total_stores = current_total_stores;
 
    pthread_mutex_unlock(&device->submit_mutex);
-}
\ No newline at end of file
+}