diff --git a/plugins/GSdx/GSCapture.cpp b/plugins/GSdx/GSCapture.cpp
index 8cec5a02db..53feaf6878 100644
--- a/plugins/GSdx/GSCapture.cpp
+++ b/plugins/GSdx/GSCapture.cpp
@@ -249,10 +249,15 @@ public:
 			GSVector4 ys(0.257f, 0.504f, 0.098f, 0.0f);
 			GSVector4 us(-0.148f / 2, -0.291f / 2, 0.439f / 2, 0.0f);
 			GSVector4 vs(0.439f / 2, -0.368f / 2, -0.071f / 2, 0.0f);
-			const GSVector4 offset(16, 128, 16, 128);
 
-			if (!rgba)
-				ys = ys.zyxw(), us = us.zyxw(), vs = vs.zyxw();
+			if(!rgba)
+			{
+				ys = ys.zyxw();
+				us = us.zyxw();
+				vs = vs.zyxw();
+			}
+
+			const GSVector4 offset(16, 128, 16, 128);
 
 			for(int j = 0; j < h; j++, dst += dstpitch, src += srcpitch)
 			{
@@ -261,8 +266,8 @@ public:
 
 				for(int i = 0; i < w; i += 2)
 				{
-					GSVector4 c0 = GSVector4(s[i + 0]);
-					GSVector4 c1 = GSVector4(s[i + 1]);
+					GSVector4 c0 = GSVector4::rgba32(s[i + 0]);
+					GSVector4 c1 = GSVector4::rgba32(s[i + 1]);
 					GSVector4 c2 = c0 + c1;
 
 					GSVector4 lo = (c0 * ys).hadd(c2 * us);
diff --git a/plugins/GSdx/GSClut.cpp b/plugins/GSdx/GSClut.cpp
index e223b72cb6..8481ed7b1e 100644
--- a/plugins/GSdx/GSClut.cpp
+++ b/plugins/GSdx/GSClut.cpp
@@ -109,7 +109,7 @@ bool GSClut::WriteTest(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT)
 	case 4: if(m_CBP[0] == TEX0.CBP) return false; m_CBP[0] = TEX0.CBP; break;
 	case 5: if(m_CBP[1] == TEX0.CBP) return false; m_CBP[1] = TEX0.CBP; break;
 	case 6: ASSERT(0); return false; // ffx2 menu
-	case 7: ASSERT(0); return false;
+	case 7: ASSERT(0); return false; // ford mustang racing
 	default: __assume(0);
 	}
 
diff --git a/plugins/GSdx/GSDevice11.cpp b/plugins/GSdx/GSDevice11.cpp
index 9c9e7866f1..c080398e29 100644
--- a/plugins/GSdx/GSDevice11.cpp
+++ b/plugins/GSdx/GSDevice11.cpp
@@ -350,7 +350,7 @@ void GSDevice11::ClearRenderTarget(GSTexture* t, const GSVector4& c)
 
 void GSDevice11::ClearRenderTarget(GSTexture* t, uint32 c)
 {
-	GSVector4 color = GSVector4(c) * (1.0f / 255);
+	GSVector4 color = GSVector4::rgba32(c) * (1.0f / 255);
 
 	m_ctx->ClearRenderTargetView(*(GSTexture11*)t, color.v);
 }
diff --git a/plugins/GSdx/GSDrawScanlineCodeGenerator.x64.avx.cpp b/plugins/GSdx/GSDrawScanlineCodeGenerator.x64.avx.cpp
index 069abae945..36797c7f5e 100644
--- a/plugins/GSdx/GSDrawScanlineCodeGenerator.x64.avx.cpp
+++ b/plugins/GSdx/GSDrawScanlineCodeGenerator.x64.avx.cpp
@@ -25,7 +25,7 @@
 
 #if _M_SSE >= 0x500 && (defined(_M_AMD64) || defined(_WIN64))
 
-#error TODO: this is still bogus somewhere
+#error TODO
 
 void GSDrawScanlineCodeGenerator::Generate()
 {
@@ -38,17 +38,13 @@ void GSDrawScanlineCodeGenerator::Generate()
 	push(r12);
 	push(r13);
 
-	enter(10 * 16, true);
+	sub(rsp, 8 + 10 * 16);
 	
 	for(int i = 6; i < 16; i++)
 	{
 		vmovdqa(ptr[rsp + (i - 6) * 16], Xmm(i));
 	}
 
-	movsxd(rcx, ecx); // right
-	movsxd(rdx, edx); // left
-	movsxd(r8, r8d); // top
-
 	mov(r10, (size_t)&m_test[0]);
 	mov(r11, (size_t)&m_local);
 	mov(r12, (size_t)m_local.gd);
@@ -84,7 +80,14 @@ L("loop");
 
 	// ebp = za
 
-	SampleTexture();
+	if(m_sel.mmin)
+	{
+		SampleTextureLOD();
+	}
+	else
+	{
+		SampleTexture();
+	}
 
 	// ebp = za
 	// xmm2 = rb
@@ -201,7 +204,7 @@ L("exit");
 		vmovdqa(Xmm(i), ptr[rsp + (i - 6) * 16]);
 	}
 
-	leave();
+	add(rsp, 8 + 10 * 16);
 
 	pop(r13);
 	pop(r12);
@@ -237,10 +240,9 @@ void GSDrawScanlineCodeGenerator::Init()
 	mov(rax, rcx);
 	sar(rax, 63);
 	and(rax, rcx);
-	add(rax, 7);
 	shl(rax, 4);
 
-	vpor(xmm15, ptr[rax + r10]);
+	vpor(xmm15, ptr[rax + r10 + 7 * 16]);
 
 	// GSVector2i* fza_base = &m_local.gd->fzbr[top];
 
@@ -256,8 +258,7 @@ void GSDrawScanlineCodeGenerator::Init()
 	{
 		// edx = &m_local.d[skip]
 
-		shl(rdx, 3);
-		lea(rdx, ptr[rdx + r11 + offsetof(GSScanlineLocalData, d)]);
+		lea(rdx, ptr[rdx * 8 + r11 + offsetof(GSScanlineLocalData, d)]);
 	}
 
 	if(!m_sel.sprite)
@@ -325,7 +326,7 @@ void GSDrawScanlineCodeGenerator::Init()
 
 				vpaddd(xmm10, ptr[rdx + offsetof(GSScanlineLocalData::skip, s)]);
 
-				if(!m_sel.sprite)
+				if(!m_sel.sprite || m_sel.mmin)
 				{
 					vpaddd(xmm11, ptr[rdx + offsetof(GSScanlineLocalData::skip, t)]);
 				}
@@ -338,12 +339,6 @@ void GSDrawScanlineCodeGenerator::Init()
 						vpsrlw(xmm6, 1);
 					}
 				}
-
-				if(m_sel.mipmap && !m_sel.lcm)
-				{
-					vshufps(xmm12, xmm0, xmm0, _MM_SHUFFLE(2, 2, 2, 2));
-					vaddps(xmm12, ptr[rdx + offsetof(GSScanlineLocalData::skip, q)]);
-				}
 			}
 			else
 			{
@@ -441,17 +436,11 @@ void GSDrawScanlineCodeGenerator::Step()
 				vpshufd(xmm1, xmm0, _MM_SHUFFLE(0, 0, 0, 0));
 				vpaddd(xmm10, xmm1);
 
-				if(!m_sel.sprite)
+				if(!m_sel.sprite || m_sel.mmin)
 				{
 					vpshufd(xmm1, xmm0, _MM_SHUFFLE(1, 1, 1, 1));
 					vpaddd(xmm11, xmm1);
 				}
-
-				if(m_sel.mipmap && !m_sel.lcm)
-				{
-					vshufps(xmm3, xmm0, xmm0, _MM_SHUFFLE(2, 2, 2, 2));
-					vaddps(xmm12, xmm3);
-				}
 			}
 			else
 			{
@@ -510,10 +499,9 @@ void GSDrawScanlineCodeGenerator::Step()
 	mov(rdx, rcx);
 	sar(rdx, 63);
 	and(rdx, rcx);
-	add(rdx, 7);
 	shl(rdx, 4);
 
-	vmovdqa(xmm15, ptr[rdx + r10]);
+	vmovdqa(xmm15, ptr[rdx + r10 + 7 * 16]);
 }
 
 void GSDrawScanlineCodeGenerator::TestZ(const Xmm& temp1, const Xmm& temp2)
@@ -589,11 +577,9 @@ void GSDrawScanlineCodeGenerator::TestZ(const Xmm& temp1, const Xmm& temp2)
 			vpslld(xmm2, 31);
 
 			// GSVector4i zso = zs - o;
-
-			vpsubd(xmm0, xmm2);
-
 			// GSVector4i zdo = zd - o;
 
+			vpsubd(xmm0, xmm2);
 			vpsubd(xmm1, xmm2);
 		}
 
@@ -629,11 +615,6 @@ void GSDrawScanlineCodeGenerator::SampleTexture()
 
 	// ebx = tex
 
-	if(m_sel.mipmap && !m_sel.lcm)
-	{
-
-	}
-
 	if(!m_sel.fst)
 	{
 		vrcpps(xmm0, xmm12);
@@ -766,10 +747,7 @@ void GSDrawScanlineCodeGenerator::SampleTexture()
 		// c10 = addr10.gather32_32((const uint32/uint8*)tex[, clut]);
 		// c11 = addr11.gather32_32((const uint32/uint8*)tex[, clut]);
 		
-		ReadTexel(xmm0, xmm0, xmm4, xmm5);
-		ReadTexel(xmm1, xmm1, xmm4, xmm5);
-		ReadTexel(xmm2, xmm2, xmm4, xmm5);
-		ReadTexel(xmm3, xmm3, xmm4, xmm5);
+		ReadTexel(4, 0);
 
 		// xmm0 = c00
 		// xmm1 = c01
@@ -863,7 +841,7 @@ void GSDrawScanlineCodeGenerator::SampleTexture()
 
 		// c00 = addr00.gather32_32((const uint32/uint8*)tex[, clut]);
 
-		ReadTexel(xmm2, xmm3, xmm0, xmm1);
+		ReadTexel(1, 0);
 
 		// GSVector4i mask = GSVector4i::x00ff();
 
@@ -1032,6 +1010,18 @@ void GSDrawScanlineCodeGenerator::Wrap(const Xmm& uv0, const Xmm& uv1)
 	}
 }
 
+void GSDrawScanlineCodeGenerator::SampleTextureLOD()
+{
+}
+
+void GSDrawScanlineCodeGenerator::WrapLOD(const Xmm& uv)
+{
+}
+
+void GSDrawScanlineCodeGenerator::WrapLOD(const Xmm& uv0, const Xmm& uv1)
+{
+}
+
 void GSDrawScanlineCodeGenerator::AlphaTFX()
 {
 	if(!m_sel.fb)
@@ -1046,6 +1036,7 @@ void GSDrawScanlineCodeGenerator::AlphaTFX()
 		// gat = gat.modulate16<1>(ga).clamp8();
 
 		modulate16(xmm3, xmm14, 1);
+
 		clamp16(xmm3, xmm0);
 
 		// if(!tcc) gat = gat.mix16(ga.srl16(7));
@@ -1053,6 +1044,7 @@ void GSDrawScanlineCodeGenerator::AlphaTFX()
 		if(!m_sel.tcc)
 		{
 			vpsrlw(xmm1, xmm14, 7);
+
 			mix16(xmm3, xmm1, xmm0);
 		}
 
@@ -1065,6 +1057,7 @@ void GSDrawScanlineCodeGenerator::AlphaTFX()
 		if(!m_sel.tcc)
 		{
 			vpsrlw(xmm1, xmm14, 7);
+
 			mix16(xmm3, xmm1, xmm0);
 		}
 
@@ -1075,7 +1068,12 @@ void GSDrawScanlineCodeGenerator::AlphaTFX()
 		// gat = gat.mix16(!tcc ? ga.srl16(7) : gat.addus8(ga.srl16(7)));
 
 		vpsrlw(xmm1, xmm14, 7);
-		if(m_sel.tcc) vpaddusb(xmm1, xmm3);
+
+		if(m_sel.tcc) 
+		{
+			vpaddusb(xmm1, xmm3);
+		}
+
 		mix16(xmm3, xmm1, xmm0);
 
 		break;
@@ -1087,6 +1085,7 @@ void GSDrawScanlineCodeGenerator::AlphaTFX()
 		if(!m_sel.tcc)
 		{
 			vpsrlw(xmm1, xmm14, 7);
+
 			mix16(xmm3, xmm1, xmm0);
 		}
 
@@ -1103,6 +1102,8 @@ void GSDrawScanlineCodeGenerator::AlphaTFX()
 
 		break;
 	}
+
+	// TODO: aa1
 }
 
 void GSDrawScanlineCodeGenerator::ReadMask()
@@ -1218,6 +1219,7 @@ void GSDrawScanlineCodeGenerator::ColorTFX()
 		// rbt = rbt.modulate16<1>(rb).clamp8();
 
 		modulate16(xmm2, xmm13, 1);
+
 		clamp16(xmm2, xmm0);
 
 		break;
@@ -1229,22 +1231,28 @@ void GSDrawScanlineCodeGenerator::ColorTFX()
 	case TFX_HIGHLIGHT:
 	case TFX_HIGHLIGHT2:
 
+		// gat = gat.modulate16<1>(ga).add16(af).clamp8().mix16(gat);
+
+		vmovdqa(xmm1, xmm3);
+
+		modulate16(xmm3, xmm14, 1);
+
 		vpshuflw(xmm6, xmm14, _MM_SHUFFLE(3, 3, 1, 1));
 		vpshufhw(xmm6, xmm6, _MM_SHUFFLE(3, 3, 1, 1));
 		vpsrlw(xmm6, 7);
 
-		// gat = gat.modulate16<1>(ga).add16(af).clamp8().mix16(gat);
-
-		vmovdqa(xmm1, xmm3);
-		modulate16(xmm3, xmm14, 1);
 		vpaddw(xmm3, xmm6);
+
 		clamp16(xmm3, xmm0);
+		
 		mix16(xmm3, xmm1, xmm0);
 
 		// rbt = rbt.modulate16<1>(rb).add16(af).clamp8();
 
 		modulate16(xmm2, xmm13, 1);
+
 		vpaddw(xmm2, xmm6);
+		
 		clamp16(xmm2, xmm0);
 
 		break;
@@ -1797,25 +1805,22 @@ void GSDrawScanlineCodeGenerator::WritePixel(const Xmm& src, const Reg64& addr,
 	}
 }
 
-void GSDrawScanlineCodeGenerator::ReadTexel(const Xmm& dst, const Xmm& addr, const Xmm& temp1, const Xmm& temp2)
+void GSDrawScanlineCodeGenerator::ReadTexel(int pixels, int mip_offset)
 {
-	ReadTexel(dst, addr, 0);
-	ReadTexel(dst, addr, 1);
-	ReadTexel(dst, addr, 2);
-	ReadTexel(dst, addr, 3);
+	// TODO
 }
 
 void GSDrawScanlineCodeGenerator::ReadTexel(const Xmm& dst, const Xmm& addr, uint8 i)
 {
 	const Address& src = m_sel.tlu ? ptr[r12 + rax * 4 + offsetof(GSScanlineGlobalData, clut)] : ptr[rbx + rax * 4];
 
-	vpextrd(eax, addr, i);
-
-	movsxd(rax, eax);
+	if(i == 0) vmovd(eax, addr);
+	else vpextrd(eax, addr, i);
 
 	if(m_sel.tlu) movzx(rax, byte[rbx + rax]);
 
-	vpinsrd(dst, src, i);
+	if(i == 0) vmovd(dst, src);
+	else vpinsrd(dst, src, i);
 }
 
 #endif
\ No newline at end of file
diff --git a/plugins/GSdx/GSDrawScanlineCodeGenerator.x86.avx.cpp b/plugins/GSdx/GSDrawScanlineCodeGenerator.x86.avx.cpp
index c5a57dc86a..8a8dbb9d66 100644
--- a/plugins/GSdx/GSDrawScanlineCodeGenerator.x86.avx.cpp
+++ b/plugins/GSdx/GSDrawScanlineCodeGenerator.x86.avx.cpp
@@ -19,9 +19,6 @@
  *
  */
 
-// TODO: x64 (use the extra regs to avoid spills of zs, zd, uf, vf, rb, ga and keep a few constants in the last two like aref or afix)
-// TODO: for edges doing 4 pixels is wasteful (needed memory access * 4)
-
 #include "stdafx.h"
 #include "GSDrawScanlineCodeGenerator.h"
 #include "GSVertexSW.h"
diff --git a/plugins/GSdx/GSFunctionMap.cpp b/plugins/GSdx/GSFunctionMap.cpp
index 098617fda0..8478648819 100644
--- a/plugins/GSdx/GSFunctionMap.cpp
+++ b/plugins/GSdx/GSFunctionMap.cpp
@@ -21,37 +21,3 @@
 
 #include "stdafx.h"
 #include "GSFunctionMap.h"
-
-void GSCodeGenerator::enter(uint32 size, bool align)
-{
-	#ifdef _M_AMD64
-
-	push(r15);
-	mov(r15, rsp);
-	if(size > 0) sub(rsp, size);
-	if(align) and(rsp, 0xfffffffffffffff0);
-
-	#else
-
-	push(ebp);
-	mov(ebp, esp);
-	if(size > 0) sub(esp, size);
-	if(align) and(esp, 0xfffffff0);
-
-	#endif
-}
-
-void GSCodeGenerator::leave()
-{
-	#ifdef _M_AMD64
-
-	mov(rsp, r15);
-	pop(r15);
-
-	#else
-
-	mov(esp, ebp);
-	pop(ebp);
-
-	#endif
-}
diff --git a/plugins/GSdx/GSFunctionMap.h b/plugins/GSdx/GSFunctionMap.h
index d692c0a9c3..55cb474f54 100644
--- a/plugins/GSdx/GSFunctionMap.h
+++ b/plugins/GSdx/GSFunctionMap.h
@@ -161,9 +161,6 @@ class GSCodeGenerator : public Xbyak::CodeGenerator
 protected:
 	Xbyak::util::Cpu m_cpu;
 
-	void enter(uint32 size, bool align);
-	void leave();
-
 public:
 	GSCodeGenerator(void* code, size_t maxsize)
 		: Xbyak::CodeGenerator(maxsize, code)
diff --git a/plugins/GSdx/GSRendererDX.h b/plugins/GSdx/GSRendererDX.h
index 9534570ae3..2e0a4221f5 100644
--- a/plugins/GSdx/GSRendererDX.h
+++ b/plugins/GSdx/GSRendererDX.h
@@ -263,7 +263,7 @@ public:
 		{
 			ps_sel.fog = 1;
 
-			ps_cb.FogColor_AREF = GSVector4(env.FOGCOL.u32[0]) / 255;
+			ps_cb.FogColor_AREF = GSVector4::rgba32(env.FOGCOL.u32[0]) / 255;
 		}
 
 		if(context->TEST.ATE)
diff --git a/plugins/GSdx/GSRendererSW.cpp b/plugins/GSdx/GSRendererSW.cpp
index 0477b20116..bae4da6616 100644
--- a/plugins/GSdx/GSRendererSW.cpp
+++ b/plugins/GSdx/GSRendererSW.cpp
@@ -384,7 +384,7 @@ bool GSRendererSW::GetScanlineGlobalData(GSScanlineGlobalData& gd)
 
 				if((int)m_vt.m_lod.x >= (int)context->TEX1.MXL)
 				{
-					k = (int)m_vt.m_lod.x << 16; // set lod to max
+					k = (int)m_vt.m_lod.x << 16; // set lod to max level
 
 					gd.sel.lcm = 1; // lod is constant
 					gd.sel.mmin = 1; // tri-linear is meaningless
@@ -432,7 +432,11 @@ bool GSRendererSW::GetScanlineGlobalData(GSScanlineGlobalData& gd)
 
 				static int s_counter = 0;
 
-				//t->Save(format("c:/temp1/%08d_%05x_0.bmp", s_counter, context->TEX0.TBP0));
+				if(0)
+				//if(context->TEX0.TH > context->TEX0.TW)
+				//if(s_n >= s_saven && s_n < s_saven + 3)
+				//if(context->TEX0.TBP0 >= 0x2b80 && context->TEX0.TBW == 2 && context->TEX0.PSM == PSM_PSMT4)
+				t->Save(format("c:/temp1/%08d_%05x_0.bmp", s_counter, context->TEX0.TBP0));
 
 				for(int i = 1, j = std::min<int>((int)context->TEX1.MXL, 6); i <= j; i++)
 				{
@@ -487,7 +491,28 @@ bool GSRendererSW::GetScanlineGlobalData(GSScanlineGlobalData& gd)
 
 					gd.tex[i] = t->m_buff;
 
-					// t->Save(format("c:/temp1/%08d_%05x_%d.bmp", s_counter, context->TEX0.TBP0, i));
+					if(0)
+					//if(context->TEX0.TH > context->TEX0.TW)
+					//if(s_n >= s_saven && s_n < s_saven + 3)
+					//if(context->TEX0.TBP0 >= 0x2b80 && context->TEX0.TBW == 2 && context->TEX0.PSM == PSM_PSMT4)
+					{
+						t->Save(format("c:/temp1/%08d_%05x_%d.bmp", s_counter, context->TEX0.TBP0, i));
+						/*
+						GIFRegTEX0 TEX0 = MIP_TEX0;
+						TEX0.TBP0 = context->TEX0.TBP0;
+						do
+						{
+							TEX0.TBP0++;
+							const GSTextureCacheSW::Texture* t = m_tc->Lookup(TEX0, env.TEXA, r, gd.sel.tw + 3);
+							if(t == NULL) {ASSERT(0); return false;}
+							t->Save(format("c:/temp1/%08d_%05x_%d.bmp", s_counter, TEX0.TBP0, i));
+						}
+						while(TEX0.TBP0 < 0x3fff);
+						*/
+
+						int i = 0;
+					}
+
 				}
 
 				s_counter++;
@@ -701,38 +726,31 @@ void GSRendererSW::VertexKick(bool skip)
 {
 	const GSDrawingContext* context = m_context;
 
-	GSVector4i xy = GSVector4i::load((int)m_v.XYZ.u32[0]);
+	GSVertexSW& dst = m_vl.AddTail();
 
-	xy = xy.insert16<3>(m_v.FOG.F);
-	xy = xy.upl16();
-	xy -= context->XYOFFSET;
+	GSVector4i xy = GSVector4i::load((int)m_v.XYZ.u32[0]).upl16() - context->XYOFFSET;
+	GSVector4i zf = GSVector4i((int)std::min<uint32>(m_v.XYZ.Z, 0xffffff00), m_v.FOG.F);
 
-	GSVertexSW v;
-
-	v.p = GSVector4(xy) * g_pos_scale;
-
-	v.c = GSVector4(GSVector4i::load((int)m_v.RGBAQ.u32[0]).u8to32() << 7);
+	dst.p = GSVector4(xy).xyxy(GSVector4(zf) + (GSVector4::m_x4f800000 & GSVector4::cast(zf.sra32(31)))) * g_pos_scale;
 
 	if(tme)
 	{
+		GSVector4 t;
+
 		if(fst)
 		{
-			v.t = GSVector4(((GSVector4i)m_v.UV).upl16() << (16 - 4));
+			t = GSVector4(((GSVector4i)m_v.UV).upl16() << (16 - 4));
 		}
 		else
 		{
-			v.t = GSVector4(m_v.ST.S, m_v.ST.T);
-			v.t *= GSVector4(0x10000 << context->TEX0.TW, 0x10000 << context->TEX0.TH);
+			t = GSVector4(m_v.ST.S, m_v.ST.T) * GSVector4(0x10000 << context->TEX0.TW, 0x10000 << context->TEX0.TH);
+			t = t.xyxy(GSVector4::load(m_v.RGBAQ.Q));
 		}
 
-		v.t = v.t.xyxy(GSVector4::load(m_v.RGBAQ.Q));
+		dst.t = t;
 	}
 
-	GSVertexSW& dst = m_vl.AddTail();
-
-	dst = v;
-
-	dst.p.z = (float)min(m_v.XYZ.Z, 0xffffff00); // max value which can survive the uint32 => float => uint32 conversion
+	dst.c = GSVector4::rgba32(m_v.RGBAQ.u32[0], 7);
 
 	int count = 0;
 
diff --git a/plugins/GSdx/GSSetupPrimCodeGenerator.x64.avx.cpp b/plugins/GSdx/GSSetupPrimCodeGenerator.x64.avx.cpp
index 7c9c7fd224..06c99c945b 100644
--- a/plugins/GSdx/GSSetupPrimCodeGenerator.x64.avx.cpp
+++ b/plugins/GSdx/GSSetupPrimCodeGenerator.x64.avx.cpp
@@ -29,7 +29,7 @@ using namespace Xbyak;
 
 void GSSetupPrimCodeGenerator::Generate()
 {
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	vmovdqa(ptr[rsp + 0], xmm6);
 	vmovdqa(ptr[rsp + 16], xmm7);
@@ -55,7 +55,7 @@ void GSSetupPrimCodeGenerator::Generate()
 	vmovdqa(xmm6, ptr[rsp + 0]);
 	vmovdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }
@@ -186,16 +186,11 @@ void GSSetupPrimCodeGenerator::Texture()
 
 	if(m_sel.fst)
 	{
-		// m_local.d4.st = GSVector4i(t * 4.0f);
-
-		if(m_sel.mipmap && !m_sel.lcm)
-		{
-			vmovhps(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq.z)], xmm1);
-		}
+		// m_local.d4.stq = GSVector4i(t * 4.0f);
 
 		vcvttps2dq(xmm1, xmm1);
 
-		vmovq(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq)], xmm1);
+		vmovdqa(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq)], xmm1);
 	}
 	else
 	{
@@ -204,7 +199,7 @@ void GSSetupPrimCodeGenerator::Texture()
 		vmovaps(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq)], xmm1);
 	}
 
-	for(int j = 0, k = m_sel.fst && !(m_sel.mipmap && !m_sel.lcm) ? 2 : 3; j < k; j++)
+	for(int j = 0, k = m_sel.fst ? 2 : 3; j < k; j++)
 	{
 		// GSVector4 ds = t.xxxx();
 		// GSVector4 dt = t.yyyy();
@@ -218,7 +213,7 @@ void GSSetupPrimCodeGenerator::Texture()
 
 			vmulps(xmm2, xmm1, Xmm(4 + i));
 
-			if(m_sel.fst && !(m_sel.mipmap && !m_sel.lcm))
+			if(m_sel.fst)
 			{
 				// m_local.d[i].s/t = GSVector4i(v);
 
diff --git a/plugins/GSdx/GSSetupPrimCodeGenerator.x64.cpp b/plugins/GSdx/GSSetupPrimCodeGenerator.x64.cpp
index 4492c8e837..e599c97f71 100644
--- a/plugins/GSdx/GSSetupPrimCodeGenerator.x64.cpp
+++ b/plugins/GSdx/GSSetupPrimCodeGenerator.x64.cpp
@@ -29,7 +29,7 @@ using namespace Xbyak;
 
 void GSSetupPrimCodeGenerator::Generate()
 {
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	vmovdqa(ptr[rsp + 0], xmm6);
 	vmovdqa(ptr[rsp + 16], xmm7);
@@ -53,7 +53,7 @@ void GSSetupPrimCodeGenerator::Generate()
 	vmovdqa(xmm6, ptr[rsp + 0]);
 	vmovdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }
@@ -191,16 +191,11 @@ void GSSetupPrimCodeGenerator::Texture()
 
 	if(m_sel.fst)
 	{
-		// m_local.d4.st = GSVector4i(t * 4.0f);
-
-		if(m_sel.mipmap && !m_sel.lcm)
-		{
-			movhps(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq.z)], xmm1);
-		}
+		// m_local.d4.stq = GSVector4i(t * 4.0f);
 
 		cvttps2dq(xmm1, xmm1);
 
-		movq(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq)], xmm1);
+		movdqa(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq)], xmm1);
 	}
 	else
 	{
@@ -209,7 +204,7 @@ void GSSetupPrimCodeGenerator::Texture()
 		movaps(ptr[r8 + offsetof(GSScanlineLocalData, d4.stq)], xmm1);
 	}
 
-	for(int j = 0, k = m_sel.fst && !(m_sel.mipmap && !m_sel.lcm) ? 2 : 3; j < k; j++)
+	for(int j = 0, k = m_sel.fst ? 2 : 3; j < k; j++)
 	{
 		// GSVector4 ds = t.xxxx();
 		// GSVector4 dt = t.yyyy();
@@ -225,7 +220,7 @@ void GSSetupPrimCodeGenerator::Texture()
 			movaps(xmm2, xmm1);
 			mulps(xmm2, Xmm(4 + i));
 
-			if(m_sel.fst && !(m_sel.mipmap && !m_sel.lcm))
+			if(m_sel.fst)
 			{
 				// m_local.d[i].s/t = GSVector4i(v);
 
diff --git a/plugins/GSdx/GSSetupPrimCodeGenerator.x86.cpp b/plugins/GSdx/GSSetupPrimCodeGenerator.x86.cpp
index 3ca05cd23a..67be7a7b36 100644
--- a/plugins/GSdx/GSSetupPrimCodeGenerator.x86.cpp
+++ b/plugins/GSdx/GSSetupPrimCodeGenerator.x86.cpp
@@ -322,8 +322,7 @@ void GSSetupPrimCodeGenerator::Color()
 	{
 		// GSVector4i c = GSVector4i(vertices[0].c);
 
-		movaps(xmm0, ptr[ecx + offsetof(GSVertexSW, c)]);
-		cvttps2dq(xmm0, xmm0);
+		cvttps2dq(xmm0, ptr[ecx + offsetof(GSVertexSW, c)]);
 
 		// c = c.upl16(c.zwxy());
 
diff --git a/plugins/GSdx/GSState.cpp b/plugins/GSdx/GSState.cpp
index b6cf2f19c7..cf6caf6baa 100644
--- a/plugins/GSdx/GSState.cpp
+++ b/plugins/GSdx/GSState.cpp
@@ -590,41 +590,41 @@ template<int i> void GSState::GIFRegHandlerTEX0(const GIFReg* r)
 
 	if(m_env.CTXT[i].TEX1.MTBA)
 	{
+		// NOTE 1: TEX1.MXL must not be automatically set to 3 here.
+		// NOTE 2: Mipmap levels are tightly packed, if (tbw << 6) > (1 << tw) then the left-over space to the right is used. (common for PSM_PSMT4)
+		// NOTE 3: Non-rectangular textures are treated as rectangular when calculating the occupied space (height is extended, not sure about width)
+
+		uint32 bp = TEX0.TBP0;
+		uint32 bw = TEX0.TBW;
+		uint32 w = 1u << TEX0.TW;
+		uint32 h = 1u << TEX0.TH;
 		uint32 bpp = GSLocalMemory::m_psm[TEX0.PSM].bpp;
 
-		uint32 tbp = TEX0.TBP0;
-		uint32 tbw = TEX0.TBW;
-		uint32 th = TEX0.TH;
+		if(h < w) h = w;
 
-		if(th >= 3)
-		{
-			tbp += (((tbw << 6) * (1 << th) * bpp >> 3) + 255) >> 8;
-			tbw = std::max<uint32>(tbw >> 1, 1);
-			th--;
+		bp += ((w * h * bpp >> 3) + 255) >> 8;
+		bw = std::max<uint32>(bw >> 1, 1);
+		w = std::max<uint32>(w >> 1, 1);
+		h = std::max<uint32>(h >> 1, 1);
 
-			m_env.CTXT[i].MIPTBP1.TBP1 = tbp;
-			m_env.CTXT[i].MIPTBP1.TBW1 = tbw;
+		m_env.CTXT[i].MIPTBP1.TBP1 = bp;
+		m_env.CTXT[i].MIPTBP1.TBW1 = bw;
 
-			tbp += (((tbw << 6) * (1 << th) * bpp >> 3) + 255) >> 8;
-			tbw = std::max<uint32>(tbw >> 1, 1);
-			th--;
+		bp += ((w * h * bpp >> 3) + 255) >> 8;
+		bw = std::max<uint32>(bw >> 1, 1);
+		w = std::max<uint32>(w >> 1, 1);
+		h = std::max<uint32>(h >> 1, 1);
 
-			m_env.CTXT[i].MIPTBP1.TBP2 = tbp;
-			m_env.CTXT[i].MIPTBP1.TBW2 = tbw;
+		m_env.CTXT[i].MIPTBP1.TBP2 = bp;
+		m_env.CTXT[i].MIPTBP1.TBW2 = bw;
 
-			tbp += (((tbw << 6) * (1 << th) * bpp >> 3) + 255) >> 8;
-			tbw = std::max<uint32>(tbw >> 1, 1);
-			th--;
+		bp += ((w * h * bpp >> 3) + 255) >> 8;
+		bw = std::max<uint32>(bw >> 1, 1);
+		w = std::max<uint32>(w >> 1, 1);
+		h = std::max<uint32>(h >> 1, 1);
 
-			m_env.CTXT[i].MIPTBP1.TBP3 = tbp;
-			m_env.CTXT[i].MIPTBP1.TBW3 = tbw;
-
-			// NOTE: TEX1.MXL must not be automatically set to 3 here
-		}
-		else
-		{
-			ASSERT(0);
-		}
+		m_env.CTXT[i].MIPTBP1.TBP3 = bp;
+		m_env.CTXT[i].MIPTBP1.TBW3 = bw;
 
 		// printf("MTBA\n");
 	}
diff --git a/plugins/GSdx/GSVector.cpp b/plugins/GSdx/GSVector.cpp
index eb2769744a..47e724966c 100644
--- a/plugins/GSdx/GSVector.cpp
+++ b/plugins/GSdx/GSVector.cpp
@@ -29,6 +29,7 @@ const GSVector4 GSVector4::m_one(1.0f);
 const GSVector4 GSVector4::m_two(2.0f);
 const GSVector4 GSVector4::m_four(4.0f);
 const GSVector4 GSVector4::m_x4b000000(_mm_castsi128_ps(_mm_set1_epi32(0x4b000000)));
+const GSVector4 GSVector4::m_x4f800000(_mm_castsi128_ps(_mm_set1_epi32(0x4f800000)));
 
 GSVector4i GSVector4i::fit(int arx, int ary) const
 {
diff --git a/plugins/GSdx/GSVector.h b/plugins/GSdx/GSVector.h
index 3b37743a3a..8d14d091e8 100644
--- a/plugins/GSdx/GSVector.h
+++ b/plugins/GSdx/GSVector.h
@@ -2333,6 +2333,7 @@ public:
 	static const GSVector4 m_two;
 	static const GSVector4 m_four;
 	static const GSVector4 m_x4b000000;
+	static const GSVector4 m_x4f800000;
 
 	__forceinline GSVector4()
 	{
@@ -2385,9 +2386,18 @@ public:
 		this->m = m;
 	}
 
-	__forceinline explicit GSVector4(uint32 u32)
+	__forceinline explicit GSVector4(int i)
 	{
-		*this = GSVector4(GSVector4i::load((int)u32).u8to32());
+		GSVector4i v((int)i);
+
+		*this = GSVector4(v);
+	}
+
+	__forceinline explicit GSVector4(uint32 u)
+	{
+		GSVector4i v((int)u);
+
+		*this = GSVector4(v) + (m_x4f800000 & GSVector4::cast(v.sra32(31)));
 	}
 
 	__forceinline explicit GSVector4(const GSVector4i& v);
@@ -2407,11 +2417,6 @@ public:
 		this->m = m;
 	}
 
-	__forceinline void operator = (uint32 u32)
-	{
-		*this = GSVector4(GSVector4i::load((int)u32).u8to32());
-	}
-
 	__forceinline operator __m128() const
 	{
 		return m;
@@ -2422,6 +2427,16 @@ public:
 		return GSVector4i(*this).rgba32();
 	}
 
+	__forceinline static GSVector4 rgba32(uint32 rgba)
+	{
+		return GSVector4(GSVector4i::load((int)rgba).u8to32());
+	}
+
+	__forceinline static GSVector4 rgba32(uint32 rgba, int shift)
+	{
+		return GSVector4(GSVector4i::load((int)rgba).u8to32() << shift);
+	}
+
 	__forceinline static GSVector4 cast(const GSVector4i& v);
 
 	__forceinline GSVector4 abs() const
@@ -2840,6 +2855,13 @@ public:
 		return GSVector4(_mm_load_ss(&f));
 	}
 
+	__forceinline static GSVector4 load(uint32 u)
+	{
+		GSVector4i v = GSVector4i::load((int)u);
+
+		return GSVector4(v) + (m_x4f800000 & GSVector4::cast(v.sra32(31)));
+	}
+
 	template<bool aligned> __forceinline static GSVector4 load(const void* p)
 	{
 		return GSVector4(aligned ? _mm_load_ps((const float*)p) : _mm_loadu_ps((const float*)p));
diff --git a/plugins/GSdx/GSVertexTrace.x64.avx.cpp b/plugins/GSdx/GSVertexTrace.x64.avx.cpp
index 3624ed5b6b..880e5644e4 100644
--- a/plugins/GSdx/GSVertexTrace.x64.avx.cpp
+++ b/plugins/GSdx/GSVertexTrace.x64.avx.cpp
@@ -51,7 +51,7 @@ GSVertexTrace::CGSW::CGSW(const void* param, uint32 key, void* code, size_t maxs
 		break;
 	}
 
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	vmovdqa(ptr[rsp + 0], xmm6);
 	vmovdqa(ptr[rsp + 16], xmm7);
@@ -168,7 +168,7 @@ GSVertexTrace::CGSW::CGSW(const void* param, uint32 key, void* code, size_t maxs
 	vmovdqa(xmm6, ptr[rsp + 0]);
 	vmovdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }
@@ -200,7 +200,7 @@ GSVertexTrace::CGHW9::CGHW9(const void* param, uint32 key, void* code, size_t ma
 		break;
 	}
 
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	vmovdqa(ptr[rsp + 0], xmm6);
 	vmovdqa(ptr[rsp + 16], xmm7);
@@ -334,7 +334,7 @@ GSVertexTrace::CGHW9::CGHW9(const void* param, uint32 key, void* code, size_t ma
 	vmovdqa(xmm6, ptr[rsp + 0]);
 	vmovdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }
@@ -364,7 +364,7 @@ GSVertexTrace::CGHW11::CGHW11(const void* param, uint32 key, void* code, size_t
 		break;
 	}
 
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	vmovdqa(ptr[rsp + 0], xmm6);
 	vmovdqa(ptr[rsp + 16], xmm7);
@@ -488,7 +488,7 @@ GSVertexTrace::CGHW11::CGHW11(const void* param, uint32 key, void* code, size_t
 	vmovdqa(xmm6, ptr[rsp + 0]);
 	vmovdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }
diff --git a/plugins/GSdx/GSVertexTrace.x64.cpp b/plugins/GSdx/GSVertexTrace.x64.cpp
index e7e9a0a0f2..8dfc6db296 100644
--- a/plugins/GSdx/GSVertexTrace.x64.cpp
+++ b/plugins/GSdx/GSVertexTrace.x64.cpp
@@ -51,7 +51,7 @@ GSVertexTrace::CGSW::CGSW(const void* param, uint32 key, void* code, size_t maxs
 		break;
 	}
 
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	movdqa(ptr[rsp + 0], xmm6);
 	movdqa(ptr[rsp + 16], xmm7);
@@ -172,7 +172,7 @@ GSVertexTrace::CGSW::CGSW(const void* param, uint32 key, void* code, size_t maxs
 	movdqa(xmm6, ptr[rsp + 0]);
 	movdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }
@@ -204,7 +204,7 @@ GSVertexTrace::CGHW9::CGHW9(const void* param, uint32 key, void* code, size_t ma
 		break;
 	}
 
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	movdqa(ptr[rsp + 0], xmm6);
 	movdqa(ptr[rsp + 16], xmm7);
@@ -355,7 +355,7 @@ GSVertexTrace::CGHW9::CGHW9(const void* param, uint32 key, void* code, size_t ma
 	movdqa(xmm6, ptr[rsp + 0]);
 	movdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }
@@ -385,7 +385,7 @@ GSVertexTrace::CGHW11::CGHW11(const void* param, uint32 key, void* code, size_t
 		break;
 	}
 
-	enter(32, true);
+	sub(rsp, 8 + 2 * 16);
 
 	movdqa(ptr[rsp + 0], xmm6);
 	movdqa(ptr[rsp + 16], xmm7);
@@ -535,7 +535,7 @@ GSVertexTrace::CGHW11::CGHW11(const void* param, uint32 key, void* code, size_t
 	movdqa(xmm6, ptr[rsp + 0]);
 	movdqa(xmm7, ptr[rsp + 16]);
 
-	leave();
+	add(rsp, 8 + 2 * 16);
 
 	ret();
 }