diff --git a/rpcs3/Emu/RSX/RSXOffload.cpp b/rpcs3/Emu/RSX/RSXOffload.cpp
index 3356326b4b..4e1eb2542f 100644
--- a/rpcs3/Emu/RSX/RSXOffload.cpp
+++ b/rpcs3/Emu/RSX/RSXOffload.cpp
@@ -84,14 +84,10 @@ namespace rsx
 		static constexpr auto thread_name = "RSX Offloader"sv;
 	};
 
-
-	using dma_thread = named_thread<dma_manager::offload_thread>;
-
-	static_assert(std::is_default_constructible_v<dma_thread>);
-
 	// initialization
 	void dma_manager::init()
 	{
+		m_thread = std::make_shared<named_thread<offload_thread>>();
 	}
 
 	// General transport
@@ -103,8 +99,8 @@ namespace rsx
 		}
 		else
 		{
-			g_fxo->get<dma_thread>().m_enqueued_count++;
-			g_fxo->get<dma_thread>().m_work_queue.push(dst, src, length);
+			m_thread->m_enqueued_count++;
+			m_thread->m_work_queue.push(dst, src, length);
 		}
 	}
 
@@ -116,8 +112,8 @@ namespace rsx
 		}
 		else
 		{
-			g_fxo->get<dma_thread>().m_enqueued_count++;
-			g_fxo->get<dma_thread>().m_work_queue.push(dst, src, length);
+			m_thread->m_enqueued_count++;
+			m_thread->m_work_queue.push(dst, src, length);
 		}
 	}
 
@@ -131,8 +127,8 @@ namespace rsx
 		}
 		else
 		{
-			g_fxo->get<dma_thread>().m_enqueued_count++;
-			g_fxo->get<dma_thread>().m_work_queue.push(dst, primitive, count);
+			m_thread->m_enqueued_count++;
+			m_thread->m_work_queue.push(dst, primitive, count);
 		}
 	}
 
@@ -141,16 +137,16 @@ namespace rsx
 	{
 		ensure(g_cfg.video.multithreaded_rsx);
 
-		g_fxo->get<dma_thread>().m_enqueued_count++;
-		g_fxo->get<dma_thread>().m_work_queue.push(request_code, args);
+		m_thread->m_enqueued_count++;
+		m_thread->m_work_queue.push(request_code, args);
 	}
 
 	// Synchronization
-	bool dma_manager::is_current_thread()
+	bool dma_manager::is_current_thread() const
 	{
 		if (auto cpu = thread_ctrl::get_current())
 		{
-			return g_fxo->get<dma_thread>().current_thread_ == cpu;
+			return m_thread->current_thread_ == cpu;
 		}
 
 		return false;
@@ -158,7 +154,7 @@ namespace rsx
 
 	bool dma_manager::sync() const
 	{
-		auto& _thr = g_fxo->get<dma_thread>();
+		auto& _thr = *m_thread;
 
 		if (_thr.m_enqueued_count.load() <= _thr.m_processed_count.load()) [[likely]]
 		{
@@ -191,8 +187,8 @@ namespace rsx
 
 	void dma_manager::join()
 	{
-		g_fxo->get<dma_thread>() = thread_state::aborting;
 		sync();
+		*m_thread = thread_state::aborting;
 	}
 
 	void dma_manager::set_mem_fault_flag()
@@ -208,9 +204,9 @@ namespace rsx
 	}
 
 	// Fault recovery
-	utils::address_range dma_manager::get_fault_range(bool writing)
+	utils::address_range dma_manager::get_fault_range(bool writing) const
 	{
-		const auto m_current_job = ensure(g_fxo->get<dma_thread>().m_current_job);
+		const auto m_current_job = ensure(m_thread->m_current_job);
 
 		void *address = nullptr;
 		u32 range = m_current_job->length;
diff --git a/rpcs3/Emu/RSX/RSXOffload.h b/rpcs3/Emu/RSX/RSXOffload.h
index f8b12194ce..8622134962 100644
--- a/rpcs3/Emu/RSX/RSXOffload.h
+++ b/rpcs3/Emu/RSX/RSXOffload.h
@@ -6,6 +6,9 @@
 
 #include <vector>
 
+template <typename T>
+class named_thread;
+
 namespace rsx
 {
 	class dma_manager
@@ -50,6 +53,9 @@ namespace rsx
 
 		atomic_t<bool> m_mem_fault_flag = false;
 
+		struct offload_thread;
+		std::shared_ptr<named_thread<offload_thread>> m_thread;
+
 		// TODO: Improved benchmarks here; value determined by profiling on a Ryzen CPU, rounded to the nearest 512 bytes
 		const u32 max_immediate_transfer_size = 3584;
 
@@ -64,21 +70,19 @@ namespace rsx
 		void copy(void *dst, void *src, u32 length) const;
 
 		// Vertex utilities
-		static void emulate_as_indexed(void *dst, rsx::primitive_type primitive, u32 count);
+		void emulate_as_indexed(void *dst, rsx::primitive_type primitive, u32 count);
 
 		// Renderer callback
-		static void backend_ctrl(u32 request_code, void* args);
+		void backend_ctrl(u32 request_code, void* args);
 
 		// Synchronization
-		static bool is_current_thread();
+		bool is_current_thread() const;
 		bool sync() const;
 		void join();
 		void set_mem_fault_flag();
 		void clear_mem_fault_flag();
 
 		// Fault recovery
-		static utils::address_range get_fault_range(bool writing);
-
-		struct offload_thread;
+		utils::address_range get_fault_range(bool writing) const;
 	};
 }