diff --git a/Source/Core/Core/HW/GPFifo.cpp b/Source/Core/Core/HW/GPFifo.cpp
index ef05946c2b..4de989b05b 100644
--- a/Source/Core/Core/HW/GPFifo.cpp
+++ b/Source/Core/Core/HW/GPFifo.cpp
@@ -56,9 +56,23 @@ void Init()
   memset(s_gather_pipe, 0, sizeof(s_gather_pipe));
 }
 
-bool IsEmpty()
+bool IsBNE()
 {
-  return GetGatherPipeCount() == 0;
+  // TODO: It's not clear exactly when the BNE (buffer not empty) bit is set.
+  // The PPC 750cl manual says in section 2.1.2.12 "Write Pipe Address Register (WPAR)" (page 78):
+  // "A mfspr WPAR is used to read the BNE bit to check for any outstanding data transfers."
+  // In section 9.4.2 "Write Gather Pipe Operation" (page 327) it says:
+  // "Software can check WPAR[BNE] to determine if the buffer is empty or not."
+  // On page 327, it also says "The only way for software to flush out a partially full 32 byte
+  // block is to fill up the block with dummy data,."
+  // On page 328, it says: "Before disabling the write gather pipe, the WPAR[BNE] bit should be
+  // tested to insure that all outstanding transfers from the buffer to the bus have completed."
+  //
+  // GXRedirectWriteGatherPipe and GXRestoreWriteGatherPipe (used for display lists) wait for
+  // the bit to be 0 before continuing, so it can't be a case of any data existing in the FIFO;
+  // it might be a case of over 32 bytes being stored pending transfer to memory? For now, always
+  // return false since that prevents hangs in games that use display lists.
+  return false;
 }
 
 void ResetGatherPipe()
diff --git a/Source/Core/Core/HW/GPFifo.h b/Source/Core/Core/HW/GPFifo.h
index b29b1b9fde..4ba023117f 100644
--- a/Source/Core/Core/HW/GPFifo.h
+++ b/Source/Core/Core/HW/GPFifo.h
@@ -26,7 +26,7 @@ void UpdateGatherPipe();
 void CheckGatherPipe();
 void FastCheckGatherPipe();
 
-bool IsEmpty();
+bool IsBNE();
 
 // Write
 void Write8(u8 value);
diff --git a/Source/Core/Core/PowerPC/Interpreter/Interpreter_SystemRegisters.cpp b/Source/Core/Core/PowerPC/Interpreter/Interpreter_SystemRegisters.cpp
index b7da980db3..8699f05b75 100644
--- a/Source/Core/Core/PowerPC/Interpreter/Interpreter_SystemRegisters.cpp
+++ b/Source/Core/Core/PowerPC/Interpreter/Interpreter_SystemRegisters.cpp
@@ -238,11 +238,14 @@ void Interpreter::mfspr(UGeckoInstruction inst)
 
   case SPR_WPAR:
   {
-    // TODO: If wpar_empty ever is false, Paper Mario hangs. Strange.
-    // Maybe WPAR is automatically flushed after a certain amount of time?
-    bool wpar_empty = true;  // GPFifo::IsEmpty();
-    if (!wpar_empty)
-      rSPR(index) |= 1;  // BNE = buffer not empty
+    // The bottom, read-only bit checks if the buffer is not empty.
+    // GXRedirectWriteGatherPipe and GXRestoreWriteGatherPipe (used for display lists) wait for
+    // this bit to be cleared before writing to SPR_WPAR again (with a value of 0x0c00800 (aka
+    // GPFifo::GATHER_PIPE_PHYSICAL_ADDRESS)).
+    // Currently, we always treat the buffer as not empty, as the exact behavior is unclear
+    // (and games that use display lists will hang if the bit doesn't eventually become zero).
+    if (GPFifo::IsBNE())
+      rSPR(index) |= 1;
     else
       rSPR(index) &= ~1;
   }
diff --git a/Source/Core/Core/PowerPC/MMU.cpp b/Source/Core/Core/PowerPC/MMU.cpp
index 3cd9bde984..301ff55a6b 100644
--- a/Source/Core/Core/PowerPC/MMU.cpp
+++ b/Source/Core/Core/PowerPC/MMU.cpp
@@ -282,16 +282,16 @@ static void WriteToHardware(u32 em_address, const u32 data, const u32 size)
     wi = translated_addr.wi;
   }
 
-  // Check for a gather pipe write.
+  // Check for a gather pipe write (which are not implemented through the MMIO system).
   //
   // Note that we must mask the address to correctly emulate certain games; Pac-Man World 3
   // in particular is affected by this. (See https://bugs.dolphin-emu.org/issues/8386)
   //
-  // Note that the PowerPC 750CL manual says (in section 9.4.2 Write Gather Pipe Operation on page
-  // 327): "A noncacheable store to an address with bits 0-26 matching WPAR[GB_ADDR] but with bits
-  // 27-31 not all zero will result in incorrect data in the buffer." So, it's possible that in some
-  // cases writes which do not exactly match the masking behave differently, but Pac-Man World 3's
-  // writes happen to behave correctly.
+  // The PowerPC 750CL manual says (in section 9.4.2 Write Gather Pipe Operation on page 327):
+  // "A noncacheable store to an address with bits 0-26 matching WPAR[GB_ADDR] but with bits 27-31
+  // not all zero will result in incorrect data in the buffer." So, it's possible that in some cases
+  // writes which do not exactly match the masking behave differently, but Pac-Man World 3's writes
+  // happen to behave correctly.
   if (flag == XCheckTLBFlag::Write &&
       (em_address & 0xFFFFF000) == GPFifo::GATHER_PIPE_PHYSICAL_ADDRESS)
   {