diff --git a/pcsx2/MTGS.cpp b/pcsx2/MTGS.cpp
index 5b0ccc3535..26d3ddb1a6 100644
--- a/pcsx2/MTGS.cpp
+++ b/pcsx2/MTGS.cpp
@@ -400,7 +400,6 @@ void SysMtgsThread::ExecuteTaskInThread()
 				case GS_RINGTYPE_MTVU_GSPACKET:
 				{
 					MTVU_LOG("MTGS - Waiting on semaXGkick!");
-					vu1Thread.KickStart(true);
 					if (!vu1Thread.semaXGkick.TryWait())
 					{
 						busy.PartialRelease();
diff --git a/pcsx2/MTVU.cpp b/pcsx2/MTVU.cpp
index f50e85b638..46750f2f3b 100644
--- a/pcsx2/MTVU.cpp
+++ b/pcsx2/MTVU.cpp
@@ -105,10 +105,7 @@ VU_Thread::~VU_Thread()
 
 void VU_Thread::Reset()
 {
-	ScopedLock lock(mtxBusy);
-
 	vuCycleIdx = 0;
-	isBusy = false;
 	m_ato_write_pos = 0;
 	m_write_pos = 0;
 	m_ato_read_pos = 0;
@@ -133,8 +130,7 @@ void VU_Thread::ExecuteRingBuffer()
 {
 	for (;;)
 	{
-		semaEvent.WaitWithoutYield();
-		ScopedLockBool lock(mtxBusy, isBusy);
+		semaEvent.WaitForWork();
 		while (m_ato_read_pos.load(std::memory_order_relaxed) != GetWritePos())
 		{
 			u32 tag = Read();
@@ -419,10 +415,9 @@ void VU_Thread::Get_MTVUChanges()
 	}
 }
 
-void VU_Thread::KickStart(bool forceKick)
+void VU_Thread::KickStart()
 {
-	if ((forceKick && !semaEvent.Count()) || (!isBusy.load(std::memory_order_acquire) && GetReadPos() != m_ato_write_pos.load(std::memory_order_relaxed)))
-		semaEvent.Post();
+	semaEvent.NotifyOfWork();
 }
 
 bool VU_Thread::IsDone()
@@ -433,16 +428,7 @@ bool VU_Thread::IsDone()
 void VU_Thread::WaitVU()
 {
 	MTVU_LOG("MTVU - WaitVU!");
-	for (;;)
-	{
-		if (IsDone())
-			break;
-		//DevCon.WriteLn("WaitVU()");
-		//pxAssert(THREAD_VU1);
-		KickStart();
-		std::this_thread::yield(); // Give a chance to the MTVU thread to actually start
-		ScopedLock lock(mtxBusy);
-	}
+	semaEvent.WaitForEmpty();
 }
 
 void VU_Thread::ExecuteVU(u32 vu_addr, u32 vif_top, u32 vif_itop, u32 fbrst)
diff --git a/pcsx2/MTVU.h b/pcsx2/MTVU.h
index dbfad2506e..d743224199 100644
--- a/pcsx2/MTVU.h
+++ b/pcsx2/MTVU.h
@@ -31,13 +31,11 @@ class VU_Thread : public pxThread {
 
 	u32 buffer[buffer_size];
 	// Note: keep atomic on separate cache line to avoid CPU conflict
-	alignas(64) std::atomic<bool> isBusy;   // Is thread processing data?
 	alignas(64) std::atomic<int> m_ato_read_pos; // Only modified by VU thread
 	alignas(64) std::atomic<int> m_ato_write_pos;    // Only modified by EE thread
 	alignas(64) int  m_read_pos; // temporary read pos (local to the VU thread)
 	int  m_write_pos; // temporary write pos (local to the EE thread)
-	Mutex     mtxBusy;
-	Semaphore semaEvent;
+	WorkSema semaEvent;
 	BaseVUmicroCPU*& vuCPU;
 	VURegs&          vuRegs;
 
@@ -67,7 +65,7 @@ public:
 	void Reset();
 
 	// Get MTVU to start processing its packets if it isn't already
-	void KickStart(bool forceKick = false);
+	void KickStart();
 
 	// Used for assertions...
 	bool IsDone();