diff --git a/Core/BaseVideoFilter.cpp b/Core/BaseVideoFilter.cpp
index 8d6c75b09..95ab58812 100644
--- a/Core/BaseVideoFilter.cpp
+++ b/Core/BaseVideoFilter.cpp
@@ -56,7 +56,7 @@ void BaseVideoFilter::SendFrame(uint16_t *ppuOutputBuffer, uint32_t frameNumber)
 {
 	_frameLock.Acquire();
 	_overscan = _console->GetSettings()->GetOverscanDimensions();
-	_isOddFrame = frameNumber % 2;
+	_isOddFrame = frameNumber & 0x01;
 	UpdateBufferSize();
 	OnBeforeApplyFilter();
 	ApplyFilter(ppuOutputBuffer);
diff --git a/Core/BaseVideoFilter.h b/Core/BaseVideoFilter.h
index ed9ad630b..add010d91 100644
--- a/Core/BaseVideoFilter.h
+++ b/Core/BaseVideoFilter.h
@@ -23,13 +23,13 @@ class BaseVideoFilter
 
 	virtual void ApplyFilter(uint16_t *ppuOutputBuffer) = 0;
 	virtual void OnBeforeApplyFilter();
-	bool IsOddFrame();
 
 public:
 	BaseVideoFilter(shared_ptr<Console> console);
 	virtual ~BaseVideoFilter();
 
 	uint32_t* GetOutputBuffer();
+	bool IsOddFrame();
 	void SendFrame(uint16_t *ppuOutputBuffer, uint32_t frameNumber);
 	void TakeScreenshot(string romName, VideoFilterType filterType);
 	void TakeScreenshot(VideoFilterType filterType, string filename, std::stringstream *stream = nullptr, bool rawScreenshot = false);
diff --git a/Core/BisqwitNtscFilter.cpp b/Core/BisqwitNtscFilter.cpp
index b77a6cfe1..bd2baa037 100644
--- a/Core/BisqwitNtscFilter.cpp
+++ b/Core/BisqwitNtscFilter.cpp
@@ -50,8 +50,7 @@ BisqwitNtscFilter::BisqwitNtscFilter(shared_ptr<Console> console, int resDivider
 			} else {
 				outputBuffer += GetOverscan().GetScreenWidth() * 64 / _resDivider / _resDivider * (120 - GetOverscan().Top);
 			}
-
-			DecodeFrame(120, 239 - GetOverscan().Bottom, _ppuOutputBuffer, outputBuffer, (IsOddFrame() ? 8 : 0) + 327360); 
+			DecodeFrame(120, 239 - GetOverscan().Bottom, _ppuOutputBuffer, outputBuffer, ((_console->GetModel() == NesModel::NTSC ? _console->GetStartingPhase() : 0) * 4) + 327360);
 
 			_workDone = true;
 		}
@@ -71,7 +70,7 @@ void BisqwitNtscFilter::ApplyFilter(uint16_t *ppuOutputBuffer)
 
 	_workDone = false;
 	_waitWork.Signal();
-	DecodeFrame(GetOverscan().Top, 120, ppuOutputBuffer, GetOutputBuffer(), (IsOddFrame() ? 8 : 0) + GetOverscan().Top*341*8);
+	DecodeFrame(GetOverscan().Top, 120, ppuOutputBuffer, GetOutputBuffer(), ((_console->GetModel() == NesModel::NTSC ? _console->GetStartingPhase() : 0) * 4) + GetOverscan().Top * 341 * 8);
 	while(!_workDone) {}
 }
 
diff --git a/Core/Console.cpp b/Core/Console.cpp
index 9b03835c5..ad609cc87 100644
--- a/Core/Console.cpp
+++ b/Core/Console.cpp
@@ -568,6 +568,11 @@ uint32_t Console::GetFrameCount()
 	return _ppu ? _ppu->GetFrameCount() : 0;
 }
 
+uint8_t Console::GetStartingPhase()
+{
+	return _ppu ? _ppu->GetStartingPhase() : 0;
+}
+
 NesModel Console::GetModel()
 {
 	return _model;
diff --git a/Core/Console.h b/Core/Console.h
index 96fb7786c..d714e4f08 100644
--- a/Core/Console.h
+++ b/Core/Console.h
@@ -204,6 +204,8 @@ class Console : public std::enable_shared_from_this<Console>
 	VirtualFile GetPatchFile();
 	RomInfo GetRomInfo();
 	uint32_t GetFrameCount();
+	// https://forums.nesdev.org/viewtopic.php?p=30625#p30625
+	uint8_t GetStartingPhase();
 	NesModel GetModel();
 
 	uint32_t GetLagCounter();
diff --git a/Core/NtscFilter.cpp b/Core/NtscFilter.cpp
index 33cac76c6..9fe0a1830 100644
--- a/Core/NtscFilter.cpp
+++ b/Core/NtscFilter.cpp
@@ -9,7 +9,7 @@ NtscFilter::NtscFilter(shared_ptr<Console> console) : BaseVideoFilter(console)
 	memset(_palette, 0, sizeof(_palette));
 	memset(&_ntscData, 0, sizeof(_ntscData));
 	_ntscSetup = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
-	_ntscBuffer = new uint32_t[NES_NTSC_OUT_WIDTH(256) * 240];
+	_ntscBuffer = new uint32_t[NES_NTSC_OUT_WIDTH(256) * 240]();
 }
 
 FrameInfo NtscFilter::GetFrameInfo()
@@ -55,14 +55,7 @@ void NtscFilter::OnBeforeApplyFilter()
 		}
 	}
 
-	if (_console->GetModel() == NesModel::NTSC) {
-		// BG color borders on NTSC machines
-		_ntsc_border = _console->GetPpu()->GetCurrentBgColor();
-	}
-	else {
-		// black borders on other machines
-		_ntsc_border = 15;
-	}
+	_ntscBorder = (_console->GetModel() == NesModel::NTSC);
 
 	PictureSettings pictureSettings = _console->GetSettings()->GetPictureSettings();
 	NtscFilterSettings ntscSettings = _console->GetSettings()->GetNtscFilterSettings();
@@ -132,7 +125,22 @@ void NtscFilter::OnBeforeApplyFilter()
 
 void NtscFilter::ApplyFilter(uint16_t *ppuOutputBuffer)
 {
-	nes_ntsc_blit(&_ntscData, ppuOutputBuffer, _ntsc_border, PPU::ScreenWidth, IsOddFrame() ? 0 : 1, PPU::ScreenWidth, 240, _ntscBuffer, NES_NTSC_OUT_WIDTH(PPU::ScreenWidth)*4);
+	uint8_t phase = _console->GetModel() == NesModel::NTSC ? _console->GetStartingPhase() : 0;
+	for (int i = 0; i < 240; i++) {
+		nes_ntsc_blit(&_ntscData,
+			// input += in_row_width;
+			ppuOutputBuffer + PPU::ScreenWidth * i,
+			_ntscBorder ? _console->GetPpu()->GetCurrentBgColor() : 0x0F,
+			PPU::ScreenWidth,
+			phase,
+			PPU::ScreenWidth,
+			1,
+			// rgb_out = (char*) rgb_out + out_pitch;
+			reinterpret_cast<char*>(_ntscBuffer) + (NES_NTSC_OUT_WIDTH(PPU::ScreenWidth) * 4 * i),
+			NES_NTSC_OUT_WIDTH(PPU::ScreenWidth) * 4);
+
+		phase = (phase + 1) % 3;
+	}
 	GenerateArgbFrame(_ntscBuffer);
 }
 
diff --git a/Core/NtscFilter.h b/Core/NtscFilter.h
index b304511f1..071448633 100644
--- a/Core/NtscFilter.h
+++ b/Core/NtscFilter.h
@@ -14,7 +14,7 @@ class NtscFilter : public BaseVideoFilter
 	bool _useExternalPalette = true;
 	uint8_t _palette[512 * 3];
 	uint32_t* _ntscBuffer;
-	uint16_t _ntsc_border;
+	bool _ntscBorder = true;
 
 	void GenerateArgbFrame(uint32_t *outputBuffer);
 
diff --git a/Core/PPU.cpp b/Core/PPU.cpp
index d15416485..fee1169b7 100644
--- a/Core/PPU.cpp
+++ b/Core/PPU.cpp
@@ -105,6 +105,8 @@ void PPU::Reset()
 	memset(_oamDecayCycles, 0, sizeof(_oamDecayCycles));
 	_enableOamDecay = _settings->CheckFlag(EmulationFlags::EnableOamDecay);
 
+	_startingPhase = 0;
+
 	UpdateMinimumDrawCycles();
 }
 
@@ -112,30 +114,37 @@ void PPU::SetNesModel(NesModel model)
 {
 	_nesModel = model;
 
+	// https://www.nesdev.org/wiki/Cycle_reference_chart
 	switch(_nesModel) {
 		case NesModel::Auto:
 			//Should never be Auto
 			break;
 
 		case NesModel::NTSC:
-			_nmiScanline = 241;
-			_vblankEnd = 260;
-			_standardNmiScanline = 241;
-			_standardVblankEnd = 260;
+			// picture height + postrender blanking lines
+			_nmiScanline = 240 + 1;
+			// picture height + (postrender blanking lines - 1) + vblank length
+			_vblankEnd = 240 + 20;
+			_standardNmiScanline = _nmiScanline;
+			_standardVblankEnd = _vblankEnd;
 			_masterClockDivider = 4;
 			break;
 		case NesModel::PAL:
-			_nmiScanline = 241;
-			_vblankEnd = 310;
-			_standardNmiScanline = 241;
-			_standardVblankEnd = 310;
+			// (picture height + border line) + postrender blanking lines
+			_nmiScanline = 240 + 1;
+			// (picture height + border line) + (postrender blanking lines - 1) + vblank length + 50
+			_vblankEnd = 240 + 20 + 50;
+			_standardNmiScanline = _nmiScanline;
+			_standardVblankEnd = _vblankEnd;
 			_masterClockDivider = 5;
 			break;
 		case NesModel::Dendy:
-			_nmiScanline = 291;
-			_vblankEnd = 310;
-			_standardNmiScanline = 291;
-			_standardVblankEnd = 310;
+			// (picture height + border line) + postrender blanking lines + 50
+			_nmiScanline = 240 + 1 + 50;
+			// (picture height + border line) + (postrender blanking lines - 1) + 50 + vblank length
+			_vblankEnd = 240 + 50 + 20;
+			_standardNmiScanline = _nmiScanline;
+			_standardVblankEnd = _vblankEnd;
 			_masterClockDivider = 5;
 			break;
 	}
@@ -999,10 +1008,9 @@ void PPU::ProcessScanline()
 			LoadTileInfo();
 		}
 	} else if(_cycle == 337 || _cycle == 339) {
-		if(IsRenderingEnabled()) {
+		if (IsRenderingEnabled()) {
 			ReadVram(GetNameTableAddr());
-
-			if(_scanline == -1 && _cycle == 339 && (_frameCount & 0x01) && _nesModel == NesModel::NTSC && _settings->GetPpuModel() == PpuModel::Ppu2C02) {
+			if (_scanline == -1 && _cycle == 339 && (_frameCount & 0x01) && _nesModel == NesModel::NTSC && _settings->GetPpuModel() == PpuModel::Ppu2C02) {
 				//This behavior is NTSC-specific - PAL frames are always the same number of cycles
 				//"With rendering enabled, each odd PPU frame is one PPU clock shorter than normal" (skip from 339 to 0, going over 340)
 				_cycle = 340;
@@ -1347,6 +1355,8 @@ void PPU::Exec()
 	if(_needStateUpdate) {
 		UpdateState();
 	}
+
+	_startingPhase = _cycle % 3;
 }
 
 void PPU::UpdateState()
diff --git a/Core/PPU.h b/Core/PPU.h
index 755041865..9007d03f9 100644
--- a/Core/PPU.h
+++ b/Core/PPU.h
@@ -109,6 +109,9 @@ class PPU : public IMemoryHandler, public Snapshotable
 		bool _enableOamDecay;
 		bool _corruptOamRow[32];
 
+		// https://forums.nesdev.org/viewtopic.php?p=30625#p30625
+		uint8_t _startingPhase;
+
 		void UpdateStatusFlag();
 
 		void SetControlRegister(uint8_t value);
@@ -217,6 +220,11 @@ class PPU : public IMemoryHandler, public Snapshotable
 			return _frameCount;
 		}
 
+		uint8_t GetStartingPhase()
+		{
+			return _startingPhase;
+		}
+
 		uint32_t GetFrameCycle()
 		{
 			return ((_scanline + 1) * 341) + _cycle;