diff --git a/core/hw/pvr/Renderer_if.cpp b/core/hw/pvr/Renderer_if.cpp
index 0e9b2696d..9032d1273 100644
--- a/core/hw/pvr/Renderer_if.cpp
+++ b/core/hw/pvr/Renderer_if.cpp
@@ -198,7 +198,7 @@ TA_context* read_frame(const char* file, u8* vram_ref) {
 bool rend_frame(TA_context* ctx, bool draw_osd) {
 	bool proc = renderer->Process(ctx);
 #if !defined(TARGET_NO_THREADS)
-	if (!ctx->rend.isRTT)
+	if (!proc || !ctx->rend.isRTT)
 		// If rendering to texture, continue locking until the frame is rendered
 		re.Set();
 #endif
diff --git a/core/hw/pvr/ta_ctx.cpp b/core/hw/pvr/ta_ctx.cpp
index a092ea844..72cf79f3a 100644
--- a/core/hw/pvr/ta_ctx.cpp
+++ b/core/hw/pvr/ta_ctx.cpp
@@ -134,17 +134,10 @@ bool QueueRender(TA_context* ctx)
 	} 
 
 	if (rqueue) {
-		// If the queued frame is for rendering to a texture, we can't skip it, so we wait
-		if (ctx->rend.isRTT) {
-			frame_finished.Wait();
-			verify(!rqueue);
-		}
-		else
-		{
-			tactx_Recycle(ctx);
-			fskip++;
-			return false;
-		}
+		// FIXME if the discarded render is a RTT we'll have a texture missing. But waiting for the current frame to finish kills performance...
+		tactx_Recycle(ctx);
+		fskip++;
+		return false;
 	}
 
 	frame_finished.Reset();