diff --git a/bin/resources/shaders/dx11/tfx.fx b/bin/resources/shaders/dx11/tfx.fx
index d6ee87eaea..532e1186cd 100644
--- a/bin/resources/shaders/dx11/tfx.fx
+++ b/bin/resources/shaders/dx11/tfx.fx
@@ -1177,7 +1177,15 @@ void ps_blend(inout float4 Color, inout float4 As_rgba, float2 pos_xy)
 			As_rgba.rgb = (float3)1.0f;
 		}
 
-		float4 RT = SW_BLEND_NEEDS_RT ? RtLoad(int2(pos_xy)) : (float4)0.0f;
+		float4 RT = (float4)0.0f;
+		if (SW_BLEND_NEEDS_RT)
+		{
+			RT = RtLoad(int2(pos_xy));
+			float color_multi = PS_COLCLIP_HW ? 65535.0f : 255.0f;
+			float alpha_multi = PS_RTA_CORRECTION ? 128.0f : 255.0f;
+			RT.rgb = trunc(RT.rgb * color_multi + 0.1f);
+			RT.a = trunc(RT.a * alpha_multi + 0.1f);
+		}
 
 		if (PS_SHUFFLE && SW_BLEND_NEEDS_RT)
 		{
@@ -1198,9 +1206,8 @@ void ps_blend(inout float4 Color, inout float4 As_rgba, float2 pos_xy)
 			}
 		}
 		
-		float Ad = PS_RTA_CORRECTION ? trunc(RT.a * 128.0f + 0.1f) / 128.0f : trunc(RT.a * 255.0f + 0.1f) / 128.0f;
-		float color_multi = PS_COLCLIP_HW ? 65535.0f : 255.0f;
-		float3 Cd = trunc(RT.rgb * color_multi + 0.1f);
+		float Ad = RT.a / 128.0f;
+		float3 Cd = RT.rgb;
 		float3 Cs = Color.rgb;
 
 		float3 A = (PS_BLEND_A == 0) ? Cs : ((PS_BLEND_A == 1) ? Cd : (float3)0.0f);
diff --git a/bin/resources/shaders/opengl/tfx_fs.glsl b/bin/resources/shaders/opengl/tfx_fs.glsl
index 0beee27060..a119f855d2 100644
--- a/bin/resources/shaders/opengl/tfx_fs.glsl
+++ b/bin/resources/shaders/opengl/tfx_fs.glsl
@@ -1026,38 +1026,40 @@ float As = As_rgba.a;
 
 #if SW_BLEND_NEEDS_RT
 	vec4 RT = sample_from_rt();
+	#if PS_COLCLIP_HW
+		float color_multi = 65535.0f;
+	#else
+		float color_multi = 255.0f;
+	#endif
+	#if PS_RTA_CORRECTION
+		float alpha_multi = 128.0f;
+	#else
+		float alpha_multi = 255.0f;
+	#endif
+	RT.rgb = trunc(RT.rgb * color_multi + 0.1f);
+	RT.a = trunc(RT.a * alpha_multi + 0.1f);
 #else
 	// Not used, but we define it to make the selection below simpler.
 	vec4 RT = vec4(0.0f);
 #endif
 
-	#if PS_RTA_CORRECTION
-		float Ad = trunc(RT.a * 128.0f + 0.1f) / 128.0f;
+#if PS_SHUFFLE && SW_BLEND_NEEDS_RT
+	uvec4 denorm_rt = uvec4(RT);
+	#if (PS_PROCESS_BA & SHUFFLE_WRITE)
+		RT.r = float((denorm_rt.b << 3) & 0xF8u);
+		RT.g = float(((denorm_rt.b >> 2) & 0x38u) | ((denorm_rt.a << 6) & 0xC0u));
+		RT.b = float((denorm_rt.a << 1) & 0xF8u);
+		RT.a = float(denorm_rt.a & 0x80u);
 	#else
-		float Ad = trunc(RT.a * 255.0f + 0.1f) / 128.0f;
+		RT.r = float((denorm_rt.r << 3) & 0xF8u);
+		RT.g = float(((denorm_rt.r >> 2) & 0x38u) | ((denorm_rt.g << 6) & 0xC0u));
+		RT.b = float((denorm_rt.g << 1) & 0xF8u);
+		RT.a = float(denorm_rt.g & 0x80u);
 	#endif
+#endif
 
-	#if PS_SHUFFLE && SW_BLEND_NEEDS_RT
-		uvec4 denorm_rt = uvec4(RT);
-		#if (PS_PROCESS_BA & SHUFFLE_WRITE)
-			RT.r = float((denorm_rt.b << 3) & 0xF8u);
-			RT.g = float(((denorm_rt.b >> 2) & 0x38u) | ((denorm_rt.a << 6) & 0xC0u));
-			RT.b = float((denorm_rt.a << 1) & 0xF8u);
-			RT.a = float(denorm_rt.a & 0x80u);
-		#else
-			RT.r = float((denorm_rt.r << 3) & 0xF8u);
-			RT.g = float(((denorm_rt.r >> 2) & 0x38u) | ((denorm_rt.g << 6) & 0xC0u));
-			RT.b = float((denorm_rt.g << 1) & 0xF8u);
-			RT.a = float(denorm_rt.g & 0x80u);
-		#endif
-	#endif
-
-	// Let the compiler do its jobs !
-	#if PS_COLCLIP_HW == 1
-		vec3 Cd = trunc(RT.rgb * 65535.0f);
-	#else
-		vec3 Cd = trunc(RT.rgb * 255.0f + 0.1f);
-	#endif
+	float Ad = RT.a / 128.0f;
+	vec3 Cd = RT.rgb;
 	vec3 Cs = Color.rgb;
 
 #if PS_BLEND_A == 0
diff --git a/bin/resources/shaders/vulkan/tfx.glsl b/bin/resources/shaders/vulkan/tfx.glsl
index 264a472a94..30f94479a5 100644
--- a/bin/resources/shaders/vulkan/tfx.glsl
+++ b/bin/resources/shaders/vulkan/tfx.glsl
@@ -1544,20 +1544,26 @@ void ps_blend(inout vec4 Color, inout vec4 As_rgba)
 			As_rgba.rgb = vec3(1.0f);
 		#endif
 
-		#if PS_FEEDBACK_LOOP_IS_NEEDED_RT
+		#if SW_BLEND_NEEDS_RT
 			vec4 RT = sample_from_rt();
+			#if PS_COLCLIP_HW
+				float color_multi = 65535.0f;
+			#else
+				float color_multi = 255.0f;
+			#endif
+			#if PS_RTA_CORRECTION
+				float alpha_multi = 128.0f;
+			#else
+				float alpha_multi = 255.0f;
+			#endif
+			RT.rgb = trunc(RT.rgb * color_multi + 0.1f);
+			RT.a = trunc(RT.a * alpha_multi + 0.1f);
 		#else
 			// Not used, but we define it to make the selection below simpler.
 			vec4 RT = vec4(0.0f);
 		#endif
 
-		#if PS_RTA_CORRECTION
-			float Ad = trunc(RT.a * 128.0f + 0.1f) / 128.0f;
-		#else
-			float Ad = trunc(RT.a * 255.0f + 0.1f) / 128.0f;
-		#endif
-
-		#if PS_SHUFFLE && PS_FEEDBACK_LOOP_IS_NEEDED_RT
+		#if PS_SHUFFLE && SW_BLEND_NEEDS_RT
 			uvec4 denorm_rt = uvec4(RT);
 			#if (PS_PROCESS_BA & SHUFFLE_WRITE)
 				RT.r = float((denorm_rt.b << 3) & 0xF8u);
@@ -1572,13 +1578,9 @@ void ps_blend(inout vec4 Color, inout vec4 As_rgba)
 			#endif
 		#endif
 
-			// Let the compiler do its jobs !
-			#if PS_COLCLIP_HW == 1
-			vec3 Cd = trunc(RT.rgb * 65535.0f);
-			#else
-			vec3 Cd = trunc(RT.rgb * 255.0f + 0.1f);
-			#endif
-			vec3 Cs = Color.rgb;
+		float Ad = RT.a / 128.0f;
+		vec3 Cd = RT.rgb;
+		vec3 Cs = Color.rgb;
 
 		#if PS_BLEND_A == 0
 			vec3 A = Cs;
diff --git a/pcsx2/GS/Renderers/Metal/tfx.metal b/pcsx2/GS/Renderers/Metal/tfx.metal
index 687f25e76d..d4d16a7964 100644
--- a/pcsx2/GS/Renderers/Metal/tfx.metal
+++ b/pcsx2/GS/Renderers/Metal/tfx.metal
@@ -1393,28 +1393,37 @@ struct PSMain
 				As_rgba.rgb = float3(1.f);
 			}
 
-			float Ad = PS_RTA_CORRECTION ? trunc(current_color.a * 128.1f) / 128.f : trunc(current_color.a * 255.1f) / 128.f;
-
-			if (PS_SHUFFLE && NEEDS_RT)
+			float4 RT = float4(0.f);
+			if (NEEDS_RT_FOR_BLEND)
 			{
-				uint4 denorm_rt = uint4(current_color);
+				RT = current_color;
+				float color_multi = PS_COLCLIP_HW ? 65535.f : 255.f;
+				float alpha_multi = PS_RTA_CORRECTION ? 128.f : 255.f;
+				RT.rgb = trunc(RT.rgb * color_multi + 0.1f);
+				RT.a = trunc(RT.a * alpha_multi + 0.1f);
+			}
+
+			if (PS_SHUFFLE && NEEDS_RT_FOR_BLEND)
+			{
+				uint4 denorm_rt = uint4(RT);
 				if (PS_PROCESS_BA & SHUFFLE_WRITE)
 				{
-					current_color.r = float((denorm_rt.b << 3) & 0xF8);
-					current_color.g = float(((denorm_rt.b >> 2) & 0x38) | ((denorm_rt.a << 6) & 0xC0));
-					current_color.b = float((denorm_rt.a << 1) & 0xF8);
-					current_color.a = float(denorm_rt.a & 0x80);
+					RT.r = float((denorm_rt.b << 3) & 0xF8);
+					RT.g = float(((denorm_rt.b >> 2) & 0x38) | ((denorm_rt.a << 6) & 0xC0));
+					RT.b = float((denorm_rt.a << 1) & 0xF8);
+					RT.a = float(denorm_rt.a & 0x80);
 				}
 				else
 				{
-					current_color.r = float((denorm_rt.r << 3) & 0xF8);
-					current_color.g = float(((denorm_rt.r >> 2) & 0x38) | ((denorm_rt.g << 6) & 0xC0));
-					current_color.b = float((denorm_rt.g << 1) & 0xF8);
-					current_color.a = float(denorm_rt.g & 0x80);
+					RT.r = float((denorm_rt.r << 3) & 0xF8);
+					RT.g = float(((denorm_rt.r >> 2) & 0x38) | ((denorm_rt.g << 6) & 0xC0));
+					RT.b = float((denorm_rt.g << 1) & 0xF8);
+					RT.a = float(denorm_rt.g & 0x80);
 				}
 			}
-			float multi = PS_COLCLIP_HW ? 65535.0 : 255.5;
-			float3 Cd = trunc(current_color.rgb * multi);
+
+			float Ad = RT.a / 128.f;
+			float3 Cd = RT.rgb;
 			float3 Cs = Color.rgb;
 
 			float3 A = pick(PS_BLEND_A, Cs, Cd, float3(0.f));
diff --git a/pcsx2/ShaderCacheVersion.h b/pcsx2/ShaderCacheVersion.h
index b296dab1fa..e732834bff 100644
--- a/pcsx2/ShaderCacheVersion.h
+++ b/pcsx2/ShaderCacheVersion.h
@@ -3,4 +3,4 @@
 
 /// Version number for GS and other shaders. Increment whenever any of the contents of the
 /// shaders change, to invalidate the cache.
-static constexpr u32 SHADER_CACHE_VERSION = 117; // Last changed in PR 15005
+static constexpr u32 SHADER_CACHE_VERSION = 118; // Last changed in PR 14987