diff --git a/src/pipewire/context.c b/src/pipewire/context.c
index f3e1b4d76..e30a2ce73 100644
--- a/src/pipewire/context.c
+++ b/src/pipewire/context.c
@@ -1637,6 +1637,7 @@ again:
 		uint64_t quantum_stamp = 0, rate_stamp = 0;
 		bool force_rate, force_quantum, restore_rate = false, restore_quantum = false;
 		bool do_reconfigure = false, need_resume, was_target_pending;
+		bool have_request = false;
 		const uint32_t *node_rates;
 		uint32_t node_n_rates, node_def_rate;
 		uint32_t node_max_quantum, node_min_quantum, node_def_quantum, node_rate_quantum;
@@ -1705,6 +1706,9 @@ again:
 				context, s, running, s->runnable, rate.num, rate.denom,
 				latency.num, latency.denom, s->name);
 
+			if (running && s != n && s->supports_request > 0)
+				have_request = true;
+
 			s->moved = false;
 		}
 
@@ -1861,6 +1865,9 @@ again:
 			n->target_rate = n->rt.position->clock.target_rate;
 		}
 
+		SPA_FLAG_UPDATE(n->rt.position->clock.flags,
+				SPA_IO_CLOCK_FLAG_LAZY, have_request && n->supports_lazy > 0);
+
 		pw_log_debug("%p: driver %p running:%d runnable:%d quantum:%u rate:%u (%"PRIu64"/%u)'%s'",
 				context, n, running, n->runnable, target_quantum, target_rate,
 				n->rt.position->clock.target_duration,
diff --git a/src/pipewire/impl-node.c b/src/pipewire/impl-node.c
index 8a03d7250..c1e79cb66 100644
--- a/src/pipewire/impl-node.c
+++ b/src/pipewire/impl-node.c
@@ -324,6 +324,9 @@ static int start_node(struct pw_impl_node *this)
 	pw_log_debug("%p: start node driving:%d driver:%d prepared:%d", this,
 			this->driving, this->driver, this->rt.prepared);
 
+	this->lazy = this->rt.position && SPA_FLAG_IS_SET(this->rt.position->clock.flags,
+			SPA_IO_CLOCK_FLAG_LAZY);
+
 	if (!(this->driving && this->driver)) {
 		impl->pending_play = true;
 		res = spa_node_send_command(this->node,
@@ -758,6 +761,14 @@ static inline void insert_driver(struct pw_context *context, struct pw_impl_node
 	spa_list_for_each_safe(n, t, &context->driver_list, driver_link) {
 		if (n->priority_driver < node->priority_driver)
 			break;
+		if (n->priority_driver == 0 && node->priority_driver == 0) {
+			/* no priority is set, we prefer the driver that does
+			 * lazy scheduling. */
+			if (n->supports_request > 0 && node->supports_lazy > 0) {
+				if (n->supports_request <= node->supports_lazy)
+					break;
+			}
+		}
 	}
 	spa_list_append(&n->driver_link, &node->driver_link);
 	pw_context_emit_driver_added(context, node);
diff --git a/src/pipewire/private.h b/src/pipewire/private.h
index e14d79f42..fa9fa8dea 100644
--- a/src/pipewire/private.h
+++ b/src/pipewire/private.h
@@ -782,6 +782,7 @@ struct pw_impl_node {
 	unsigned int sync:1;		/**< the sync-groups are active */
 	unsigned int transport:1;	/**< the transport is active */
 	unsigned int async:1;		/**< async processing, one cycle latency */
+	unsigned int lazy:1;		/**< the graph is lazy scheduling */
 
 	uint32_t port_user_data_size;	/**< extra size for port user data */