diff --git a/plugins/GSdx/GSRendererHW.cpp b/plugins/GSdx/GSRendererHW.cpp
index 890519e344..7d9f30fdf0 100644
--- a/plugins/GSdx/GSRendererHW.cpp
+++ b/plugins/GSdx/GSRendererHW.cpp
@@ -396,16 +396,23 @@ void GSRendererHW::Draw()
 		else if ((m_userhacks_round_sprite_offset == 2) || (m_userhacks_round_sprite_offset == 1 && !m_vt.IsLinear())) {
 //#define DEBUG_U
 //#define DEBUG_V
+			//bool debug = m_vt.IsLinear();
+			const int half = m_vt.IsLinear() ? 8 : 0;
+
 			for(size_t i = 0; i < count; i += 2) {
-				// Compute the coordinate of first texels (in native)
+				// Performance note: if it had any impact on perf, someone would port it to SSE
+
+				// Compute the coordinate of first and last texels (in native with a linear filtering)
 				float ax0 = alpha0(context->XYOFFSET.OFX, v[i].XYZ.X, v[i+1].XYZ.X);
 				float ax1 = alpha1(context->XYOFFSET.OFX, v[i].XYZ.X, v[i+1].XYZ.X);
 				int   tx0 = Interpolate_UV(ax0, v[i].U, v[i+1].U);
 				int   tx1 = Interpolate_UV(ax1, v[i].U, v[i+1].U);
 #ifdef DEBUG_U
-				fprintf(stderr, "u0:%d and u1:%d\n", v[i].U, v[i+1].U);
-				fprintf(stderr, "a0:%f and a1:%f\n", ax0, ax1);
-				fprintf(stderr, "t0:%d and t1:%d\n", tx0, tx1);
+				if (debug) {
+					fprintf(stderr, "u0:%d and u1:%d\n", v[i].U, v[i+1].U);
+					fprintf(stderr, "a0:%f and a1:%f\n", ax0, ax1);
+					fprintf(stderr, "t0:%d and t1:%d\n", tx0, tx1);
+				}
 #endif
 
 				float ay0 = alpha0(context->XYOFFSET.OFY, v[i].XYZ.Y, v[i+1].XYZ.Y);
@@ -413,22 +420,33 @@ void GSRendererHW::Draw()
 				int   ty0 = Interpolate_UV(ay0, v[i].V, v[i+1].V);
 				int   ty1 = Interpolate_UV(ay1, v[i].V, v[i+1].V);
 #ifdef DEBUG_V
-				fprintf(stderr, "v0:%d and v1:%d\n", v[i].V, v[i+1].V);
-				fprintf(stderr, "a0:%f and a1:%f\n", ay0, ay1);
-				fprintf(stderr, "t0:%d and t1:%d\n", ty0, ty1);
+				if (debug) {
+					fprintf(stderr, "v0:%d and v1:%d\n", v[i].V, v[i+1].V);
+					fprintf(stderr, "a0:%f and a1:%f\n", ay0, ay1);
+					fprintf(stderr, "t0:%d and t1:%d\n", ty0, ty1);
+				}
 #endif
 
 #ifdef DEBUG_U
-				fprintf(stderr, "GREP_BEFORE %d => %d\n", v[i].U, v[i+1].U);
+				if (debug)
+					fprintf(stderr, "GREP_BEFORE %d => %d\n", v[i].U, v[i+1].U);
 #endif
 #ifdef DEBUG_V
-				fprintf(stderr, "GREP_BEFORE %d => %d\n", v[i].V, v[i+1].V);
+				if (debug)
+					fprintf(stderr, "GREP_BEFORE %d => %d\n", v[i].V, v[i+1].V);
 #endif
 
 #if 1
+				// Use rounded value of the newly computed texture coordinate. It ensures
+				// that sampling will remains inside texture boundary
+				//
+				// Note for bilinear: in this mode the PS2 add -0.5 offset (aka half) and 4 texels
+				// will be sampled so (t0 - 8) and (t1 - 8 + 16) must be valid.
+				//
+				// Minus half for t1 case might be too much
 				if (tx0 < tx1) {
-					v[i].U   = tx0 + 1;
-					v[i+1].U = tx1 + 1 + 16;
+					v[i].U   = tx0 + half + 1;
+					v[i+1].U = tx1 - half + 1 + 16;
 				} else {
 					v[i].U   = tx0 + 15;
 					v[i+1].U = tx1 + 15 + 16;
@@ -436,8 +454,8 @@ void GSRendererHW::Draw()
 #endif
 #if 1
 				if (ty0 < ty1) {
-					v[i].V   = ty0 + 1;
-					v[i+1].V = ty1 + 1 + 16;
+					v[i].V   = ty0 + half + 1;
+					v[i+1].V = ty1 - half + 1 + 16;
 				} else {
 					v[i].V   = ty0 + 15;
 					v[i+1].V = ty1 + 15 + 16;
@@ -445,10 +463,12 @@ void GSRendererHW::Draw()
 #endif
 
 #ifdef DEBUG_U
-				fprintf(stderr, "GREP_AFTER %d => %d\n\n", v[i].U, v[i+1].U);
+				if (debug)
+					fprintf(stderr, "GREP_AFTER %d => %d\n\n", v[i].U, v[i+1].U);
 #endif
 #ifdef DEBUG_V
-				fprintf(stderr, "GREP_AFTER %d => %d\n\n", v[i].V, v[i+1].V);
+				if (debug)
+					fprintf(stderr, "GREP_AFTER %d => %d\n\n", v[i].V, v[i+1].V);
 #endif
 
 			}