diff --git a/rpcs3/Emu/Cell/SPUThread.cpp b/rpcs3/Emu/Cell/SPUThread.cpp
index 2468e72c09..42e186a2e3 100644
--- a/rpcs3/Emu/Cell/SPUThread.cpp
+++ b/rpcs3/Emu/Cell/SPUThread.cpp
@@ -5659,6 +5659,24 @@ s64 spu_thread::get_ch_value(u32 ch)
 				// Don't busy-wait with TSX - memory is sensitive
 				if (g_use_rtm || !reservation_busy_waiting)
 				{
+					if (u32 max_threads = std::min<u32>(g_cfg.core.max_spurs_threads, group->max_num); group->max_run != max_threads)
+					{
+						constexpr std::string_view spurs_suffix = "CellSpursKernelGroup"sv;
+
+						if (group->name.ends_with(spurs_suffix) && !group->name.substr(0, group->name.size() - spurs_suffix.size()).ends_with("_libsail"))
+						{
+							// Hack: don't run more SPURS threads than specified.
+							if (u32 old = atomic_storage<u32>::exchange(group->max_run, max_threads); old > max_threads)
+							{
+								spu_log.success("HACK: '%s' (0x%x) limited to %u threads.", group->name, group->id, max_threads);
+							}
+							else if (u32 running = group->spurs_running; old < max_threads && running >= old && running < max_threads)
+							{
+								group->spurs_running.notify_all();
+							}
+						}
+					}
+
 					if (u32 work_count = g_spu_work_count)
 					{
 						const u32 true_free = utils::sub_saturate<u32>(utils::get_thread_count(), 10);
diff --git a/rpcs3/Emu/Cell/lv2/sys_spu.cpp b/rpcs3/Emu/Cell/lv2/sys_spu.cpp
index 150d755536..87cff4ea34 100644
--- a/rpcs3/Emu/Cell/lv2/sys_spu.cpp
+++ b/rpcs3/Emu/Cell/lv2/sys_spu.cpp
@@ -731,14 +731,16 @@ error_code sys_spu_thread_initialize(ppu_thread& ppu, vm::ptr<u32> thread, u32 g
 
 	if (++group->init == group->max_num)
 	{
-		if (g_cfg.core.max_spurs_threads < 6 && group->max_num > 0u + g_cfg.core.max_spurs_threads)
+		if (u32 max_threads = std::min<u32>(g_cfg.core.max_spurs_threads, group->max_num); group->max_num > max_threads)
 		{
-			if (group->name.ends_with("CellSpursKernelGroup") && !group->name.ends_with("_libsailCellSpursKernelGroup"))
+			constexpr std::string_view spurs_suffix = "CellSpursKernelGroup"sv;
+
+			if (group->name.ends_with(spurs_suffix) && !group->name.substr(0, group->name.size() - spurs_suffix.size()).ends_with("_libsail"))
 			{
 				// Hack: don't run more SPURS threads than specified.
-				group->max_run = g_cfg.core.max_spurs_threads;
+				group->max_run = max_threads;
 
-				spu_log.success("HACK: '%s' (0x%x) limited to %u threads.", group->name, group_id, +g_cfg.core.max_spurs_threads);
+				spu_log.success("HACK: '%s' (0x%x) limited to %u threads.", group->name, group_id, max_threads);
 			}
 		}
 
diff --git a/rpcs3/Emu/system_config.h b/rpcs3/Emu/system_config.h
index 236c4269d7..3fbefc8df7 100644
--- a/rpcs3/Emu/system_config.h
+++ b/rpcs3/Emu/system_config.h
@@ -39,7 +39,7 @@ struct cfg_root : cfg::node
 		cfg::_int<0, 6> preferred_spu_threads{ this, "Preferred SPU Threads", 0, true }; // Number of hardware threads dedicated to heavy simultaneous spu tasks
 		cfg::_int<0, 16> spu_delay_penalty{ this, "SPU delay penalty", 3 }; // Number of milliseconds to block a thread if a virtual 'core' isn't free
 		cfg::_bool spu_loop_detection{ this, "SPU loop detection", false }; // Try to detect wait loops and trigger thread yield
-		cfg::_int<0, 6> max_spurs_threads{ this, "Max SPURS Threads", 6 }; // HACK. If less then 6, max number of running SPURS threads in each thread group.
+		cfg::_int<0, 6> max_spurs_threads{ this, "Max SPURS Threads", 6, true }; // HACK. If less then 6, max number of running SPURS threads in each thread group.
 		cfg::_enum<spu_block_size_type> spu_block_size{ this, "SPU Block Size", spu_block_size_type::safe };
 		cfg::_bool spu_accurate_dma{ this, "Accurate SPU DMA", false };
 		cfg::_bool spu_accurate_reservations{ this, "Accurate SPU Reservations", true };