diff --git a/src/GPU.cpp b/src/GPU.cpp
index acbc9f17..53069784 100644
--- a/src/GPU.cpp
+++ b/src/GPU.cpp
@@ -880,9 +880,9 @@ void GPU::StartHBlank(u32 line) noexcept
     DispStat[0] |= (1<<1);
     DispStat[1] |= (1<<1);
     
-    // not the correct timing, but... close enough i guess?
+    // TODO: not quite the correct update time, but... close enough i guess?
     int scanline = (VCount == 262 ? 0 : (line+1));
-    GPU3D.ScanlineSync(scanline);
+    if (!(scanline & 1)) GPU3D.ScanlineSync(scanline);
     if (GPU3D.UnderflowFlagVCount == scanline) GPU3D.DispCnt |= (1<<12);
 
     if (VCount < 192)
diff --git a/src/GPU3D_Soft.cpp b/src/GPU3D_Soft.cpp
index eca3fa5b..fc2f0c13 100644
--- a/src/GPU3D_Soft.cpp
+++ b/src/GPU3D_Soft.cpp
@@ -2158,18 +2158,21 @@ void SoftRenderer::RenderThreadFunc(GPU& gpu)
 }
 void SoftRenderer::ScanlineSync(int line)
 {
+    // only used in accurate mode (timings must be emulated)
     if (Accuracy && RenderThreadRunning.load(std::memory_order_relaxed))
     {
         if (line < 192)
-            // We need a scanline, so let's wait for the render thread to finish it.
-            // (both threads process scanlines from top-to-bottom,
-            // so we don't need to wait for a specific row)
+        {
+            // wait for two scanlines here, since scanlines render in pairs.
             Platform::Semaphore_Wait(Sema_ScanlineCount);
+            Platform::Semaphore_Wait(Sema_ScanlineCount);
+        }
     }
 }
 
 u32* SoftRenderer::GetLine(int line)
 {
+    // only wait in in-accurate mode (we've already waited for scanlines in accurate mode)
     if (!Accuracy && RenderThreadRunning.load(std::memory_order_relaxed))
     {
         if (line < 192)