diff --git a/plugins/GSdx/Renderers/DX11/GSDevice11.h b/plugins/GSdx/Renderers/DX11/GSDevice11.h
index aec90d2cac..dee3e6f32d 100644
--- a/plugins/GSdx/Renderers/DX11/GSDevice11.h
+++ b/plugins/GSdx/Renderers/DX11/GSDevice11.h
@@ -99,6 +99,7 @@ public:
 		GSVector4 MinF_TA;
 		GSVector4i MskFix;
 		GSVector4i ChannelShuffle;
+		GSVector4i FbMask;
 
 		GSVector4 TC_OffsetHack;
 
@@ -111,6 +112,7 @@ public:
 			MinF_TA = GSVector4::zero();
 			MskFix = GSVector4i::zero();
 			ChannelShuffle = GSVector4i::zero();
+			FbMask = GSVector4i::zero();
 		}
 
 		__forceinline bool Update(const PSConstantBuffer* cb)
@@ -118,7 +120,7 @@ public:
 			GSVector4i* a = (GSVector4i*)this;
 			GSVector4i* b = (GSVector4i*)cb;
 
-			if(!((a[0] == b[0]) /*& (a[1] == b1)*/ & (a[2] == b[2]) & (a[3] == b[3]) & (a[4] == b[4]) & (a[5] == b[5]) & (a[6] == b[6])).alltrue()) // if WH matches HalfTexel does too
+			if(!((a[0] == b[0]) /*& (a[1] == b1)*/ & (a[2] == b[2]) & (a[3] == b[3]) & (a[4] == b[4]) & (a[5] == b[5]) & (a[6] == b[6]) & (a[7] == b[7])).alltrue()) // if WH matches HalfTexel does too
 			{
 				a[0] = b[0];
 				a[1] = b[1];
@@ -127,6 +129,7 @@ public:
 				a[4] = b[4];
 				a[5] = b[5];
 				a[6] = b[6];
+				a[7] = b[7];
 
 				return true;
 			}
@@ -201,6 +204,7 @@ public:
 				// Shuffle and fbmask effect
 				uint32 shuffle:1;
 				uint32 read_ba:1;
+				uint32 fbmask:1;
 
 				// *** Word 2
 				// Blend and Colclip
@@ -210,7 +214,6 @@ public:
 				uint32 channel:3;
 
 				// Hack
-				uint32 aout:1;
 				uint32 spritehack:1;
 				uint32 tcoffsethack:1;
 				uint32 urban_chaos_hle:1;
diff --git a/plugins/GSdx/Renderers/DX11/GSRendererDX11.cpp b/plugins/GSdx/Renderers/DX11/GSRendererDX11.cpp
index 0e91717444..0a7095f2b9 100644
--- a/plugins/GSdx/Renderers/DX11/GSRendererDX11.cpp
+++ b/plugins/GSdx/Renderers/DX11/GSRendererDX11.cpp
@@ -26,13 +26,9 @@ GSRendererDX11::GSRendererDX11()
 	: GSRendererHW(new GSTextureCache11(this))
 {
 	if (theApp.GetConfigB("UserHacks"))
-	{
 		UserHacks_AlphaStencil = theApp.GetConfigB("UserHacks_AlphaStencil");
-	}
 	else
-	{
 		UserHacks_AlphaStencil = false;
-	}
 
 	ResetStates();
 }
@@ -244,38 +240,71 @@ void GSRendererDX11::EmulateTextureShuffleAndFbmask()
 		if (rg_mask != 0xFF)
 		{
 			if (write_ba)
+			{
+				// fprintf(stderr, "Color shuffle %s => B\n", read_ba ? "B" : "R");
 				m_om_bsel.wb = 1;
+			}
 			else
+			{
+				// fprintf(stderr, "Color shuffle %s => R"\n, read_ba ? "B" : "R");
 				m_om_bsel.wr = 1;
-		}
-		else if ((fbmask & 0xFF) != 0xFF)
-		{
+			}
 #ifdef _DEBUG
-			fprintf(stderr, "Please fix me! wb %u wr %u\n", m_om_bsel.wb, m_om_bsel.wr);
+			if (rg_mask)
+			{
+				fprintf(stderr, "ERROR: FBMASK SW emulated fb_mask:%x on RG tex shuffle not supported\n", fbmask);
+				// ASSERT(0);
+			}
 #endif
-			//ASSERT(0);
 		}
 
 		if (ba_mask != 0xFF)
 		{
 			if (write_ba)
+			{
+				// fprintf(stderr, "Color shuffle %s => A"\n, read_ba ? "A" : "G");
 				m_om_bsel.wa = 1;
+			}
 			else
+			{
+				// fprintf(stderr, "Color shuffle %s => G"\n, read_ba ? "A" : "G");
 				m_om_bsel.wg = 1;
-		}
-		else if ((fbmask & 0xFF) != 0xFF)
-		{
+			}
 #ifdef _DEBUG
-			fprintf(stderr, "Please fix me! wa %u wg %u\n", m_om_bsel.wa, m_om_bsel.wg);
+			if (ba_mask)
+			{
+				fprintf(stderr, "ERROR: FBMASK SW emulated fb_mask:%x on BA tex shuffle not supported\n", fbmask);
+				// ASSERT(0);
+			}
 #endif
-			//ASSERT(0);
 		}
 	}
 	else
 	{
 		m_ps_sel.dfmt = GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt;
 
-		m_om_bsel.wrgba = ~GSVector4i::load((int)m_context->FRAME.FBMSK).eq8(GSVector4i::xffffffff()).mask();
+		GSVector4i fbmask_v = GSVector4i::load((int)m_context->FRAME.FBMSK);
+		int ff_fbmask = fbmask_v.eq8(GSVector4i::xffffffff()).mask();
+		int zero_fbmask = fbmask_v.eq8(GSVector4i::zero()).mask();
+
+		m_om_bsel.wrgba = ~ff_fbmask; // Enable channel if at least 1 bit is 0
+
+		m_ps_sel.fbmask = m_sw_blending && (~ff_fbmask & ~zero_fbmask & 0xF);
+
+		if (m_ps_sel.fbmask)
+		{
+			ps_cb.FbMask = fbmask_v.u8to32();
+			// Only alpha is special here, I think we can take a very unsafe shortcut
+			// Alpha isn't blended on the GS but directly copyied into the RT.
+			//
+			// Behavior is clearly undefined however there is a high probability that
+			// it will work. Masked bit will be constant and normally the same everywhere
+			// RT/FS output/Cached value.
+
+			/*fprintf(stderr, "FBMASK SW emulated alpha fb_mask:%x on %d bits format\n", m_context->FRAME.FBMSK,
+				(GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt == 2) ? 16 : 32);*/
+			m_bind_rtsample = true;
+		}
 	}
 }
 
@@ -621,6 +650,8 @@ void GSRendererDX11::EmulateTextureSampler(const GSTextureCache::Source* tex)
 
 void GSRendererDX11::ResetStates()
 {
+	m_bind_rtsample = false;
+
 	m_vs_sel.key = 0;
 	m_gs_sel.key = 0;
 	m_ps_sel.key = 0;
@@ -912,6 +943,15 @@ void GSRendererDX11::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sou
 		m_ps_sel.tfx = 4;
 	}
 
+	if (m_bind_rtsample)
+	{
+		// Bind the RT.This way special effect can use it.
+		// Do not always bind the rt when it's not needed,
+		// only bind it when effects use it such as fbmask emulation currently
+		// because we copy the frame buffer and it is quite slow.
+		dev->PSSetShaderResource(3, rt);
+	}
+
 	if (m_game.title == CRC::ICO)
 	{
 		GSVertex* v = &m_vertex.buff[0];
diff --git a/plugins/GSdx/Renderers/DX11/GSRendererDX11.h b/plugins/GSdx/Renderers/DX11/GSRendererDX11.h
index ef193a10ec..cb490486cf 100644
--- a/plugins/GSdx/Renderers/DX11/GSRendererDX11.h
+++ b/plugins/GSdx/Renderers/DX11/GSRendererDX11.h
@@ -29,6 +29,7 @@ class GSRendererDX11 final : public GSRendererHW
 {
 private:
 	bool UserHacks_AlphaStencil;
+	bool m_bind_rtsample;
 
 private:
 	inline void ResetStates();
diff --git a/plugins/GSdx/Renderers/DX11/GSTextureFX11.cpp b/plugins/GSdx/Renderers/DX11/GSTextureFX11.cpp
index 6db9ab1f55..43628152b8 100644
--- a/plugins/GSdx/Renderers/DX11/GSTextureFX11.cpp
+++ b/plugins/GSdx/Renderers/DX11/GSTextureFX11.cpp
@@ -220,7 +220,7 @@ void GSDevice11::SetupPS(PSSelector sel, const PSConstantBuffer* cb, PSSamplerSe
 		str[8] = format("%d", sel.fog);
 		str[9] = format("%d", sel.clr1);
 		str[10] = format("%d", sel.fba);
-		str[11] = format("%d", sel.aout);
+		str[11] = format("%d", sel.fbmask);
 		str[12] = format("%d", sel.ltf);
 		str[13] = format("%d", sel.spritehack);
 		str[14] = format("%d", sel.tcoffsethack);
@@ -249,7 +249,7 @@ void GSDevice11::SetupPS(PSSelector sel, const PSConstantBuffer* cb, PSSamplerSe
 			{"PS_FOG", str[8].c_str()},
 			{"PS_CLR1", str[9].c_str()},
 			{"PS_FBA", str[10].c_str()},
-			{"PS_AOUT", str[11].c_str()},
+			{"PS_FBMASK", str[11].c_str()},
 			{"PS_LTF", str[12].c_str()},
 			{"PS_SPRITEHACK", str[13].c_str()},
 			{"PS_TCOFFSETHACK", str[14].c_str()},
diff --git a/plugins/GSdx/Renderers/HW/GSRendererHW.cpp b/plugins/GSdx/Renderers/HW/GSRendererHW.cpp
index 6066348dc2..9f37299dc6 100644
--- a/plugins/GSdx/Renderers/HW/GSRendererHW.cpp
+++ b/plugins/GSdx/Renderers/HW/GSRendererHW.cpp
@@ -40,6 +40,7 @@ GSRendererHW::GSRendererHW(GSTextureCache* tc)
 	m_upscale_multiplier = theApp.GetConfigI("upscale_multiplier");
 	m_large_framebuffer  = theApp.GetConfigB("large_framebuffer");
 	m_accurate_date = theApp.GetConfigI("accurate_date");
+	m_sw_blending   = theApp.GetConfigI("accurate_blending_unit");
 	if (theApp.GetConfigB("UserHacks")) {
 		m_userhacks_enabled_gs_mem_clear = !theApp.GetConfigB("UserHacks_Disable_Safe_Features");
 		m_userHacks_enabled_unscale_ptln = !theApp.GetConfigB("UserHacks_Disable_Safe_Features");
diff --git a/plugins/GSdx/Renderers/HW/GSRendererHW.h b/plugins/GSdx/Renderers/HW/GSRendererHW.h
index 7481cb535a..676ec8c083 100644
--- a/plugins/GSdx/Renderers/HW/GSRendererHW.h
+++ b/plugins/GSdx/Renderers/HW/GSRendererHW.h
@@ -151,6 +151,7 @@ protected:
 	float m_userhacks_tcoffset_y;
 
 	int m_accurate_date;
+	int m_sw_blending;
 
 	bool m_channel_shuffle;
 
diff --git a/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.cpp b/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.cpp
index 0d59aef64e..0327f9325e 100644
--- a/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.cpp
+++ b/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.cpp
@@ -26,7 +26,6 @@
 GSRendererOGL::GSRendererOGL()
 	: GSRendererHW(new GSTextureCacheOGL(this))
 {
-	m_sw_blending = theApp.GetConfigI("accurate_blending_unit");
 	if (theApp.GetConfigB("UserHacks"))
 		UserHacks_tri_filter = static_cast<TriFiltering>(theApp.GetConfigI("UserHacks_TriFilter"));
 	else
diff --git a/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.h b/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.h
index 1563da93af..08a6b56fd5 100644
--- a/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.h
+++ b/plugins/GSdx/Renderers/OpenGL/GSRendererOGL.h
@@ -49,7 +49,6 @@ class GSRendererOGL final : public GSRendererHW
 	};
 
 	private:
-		int m_sw_blending;
 		PRIM_OVERLAP m_prim_overlap;
 		std::vector<size_t> m_drawlist;
 
diff --git a/plugins/GSdx/res/tfx.fx b/plugins/GSdx/res/tfx.fx
index ea3b9cd793..a81848a53f 100644
--- a/plugins/GSdx/res/tfx.fx
+++ b/plugins/GSdx/res/tfx.fx
@@ -29,7 +29,7 @@
 #define PS_FOG 0
 #define PS_CLR1 0
 #define PS_FBA 0
-#define PS_AOUT 0
+#define PS_FBMASK 0
 #define PS_LTF 1
 #define PS_SPRITEHACK 0
 #define PS_TCOFFSETHACK 0
@@ -81,6 +81,7 @@ struct PS_OUTPUT
 
 Texture2D<float4> Texture : register(t0);
 Texture2D<float4> Palette : register(t1);
+Texture2D<float4> RtSample : register(t3);
 Texture2D<float4> RawTexture : register(t4);
 SamplerState TextureSampler : register(s0);
 SamplerState PaletteSampler : register(s1);
@@ -103,6 +104,7 @@ cbuffer cb1
 	float2 TA;
 	uint4 MskFix;
 	int4 ChannelShuffle;
+	uint4 FbMask;
 	float4 TC_OffsetHack;
 };
 
@@ -902,9 +904,10 @@ PS_OUTPUT ps_main(PS_INPUT input)
 		}
 	}
 
-	output.c1 = c.a * 255.0f / 128.0f; // used for alpha blending
+	// Must be done before alpha correction
+	float alpha_blend = c.a * 255.0f / 128.0f;
 
-	if ((PS_DFMT == FMT_16) || PS_AOUT) // 16 bit output
+	if (PS_DFMT == FMT_16) // 16 bit output
 	{
 		float a = 128.0f / 255; // alpha output will be 0x80
 
@@ -915,7 +918,16 @@ PS_OUTPUT ps_main(PS_INPUT input)
 		if (c.a < 128.0f / 255.0f) c.a += 128.0f / 255.0f;
 	}
 
+	if (PS_FBMASK)
+	{
+		float4 rt = RtSample.Load(int3(input.p.xy, 0));
+		uint4 denorm_rt = uint4(rt * 255.0f + 0.5f);
+		uint4 denorm_c = uint4(c * 255.0f + 0.5f);
+		c = float4((denorm_c & ~FbMask) | (denorm_rt & FbMask)) / 255.0f;
+	}
+
 	output.c0 = c;
+	output.c1 = (float4)(alpha_blend);
 
 	return output;
 }