diff --git a/GPU/Software/Clipper.cpp b/GPU/Software/Clipper.cpp
index 1204e12fe4..b7cef6ccac 100644
--- a/GPU/Software/Clipper.cpp
+++ b/GPU/Software/Clipper.cpp
@@ -162,13 +162,16 @@ void ProcessQuad(VertexData* data)
 	verts[1].drawpos.x = data[1].drawpos.x;
 	verts[4].drawpos.x = data[0].drawpos.x;
 
-	// Color values of second vertex are used for the whole rectangle
+	// Color and depth values of second vertex are used for the whole rectangle
 	verts[0].color0 = verts[1].color0;
 	verts[1].color0 = verts[1].color0;
 	verts[5].color0 = verts[1].color0;
 	verts[0].color1 = verts[1].color1;
 	verts[1].color1 = verts[1].color1;
 	verts[5].color1 = verts[1].color1;
+	verts[0].drawpos.z = verts[1].drawpos.z;
+	verts[1].drawpos.z = verts[1].drawpos.z;
+	verts[5].drawpos.z = verts[1].drawpos.z;
 
 	Rasterizer::DrawTriangle(verts);
 	Rasterizer::DrawTriangle(verts+3);
diff --git a/GPU/Software/Rasterizer.cpp b/GPU/Software/Rasterizer.cpp
index cba4186f18..9c3c0ee6f4 100644
--- a/GPU/Software/Rasterizer.cpp
+++ b/GPU/Software/Rasterizer.cpp
@@ -21,6 +21,7 @@
 #include "Rasterizer.h"
 
 extern u8* fb;
+extern u8* depthbuf;
 
 namespace Rasterizer {
 
@@ -100,7 +101,7 @@ void DrawTriangle(VertexData vertexdata[3])
 	minY = std::max(minY, gstate.getScissorY1());
 	maxY = std::min(maxY, gstate.getScissorY2());
 
-	DrawingCoords p(minX, minY);
+	DrawingCoords p(minX, minY, 0);
 	for (p.y = minY; p.y <= maxY; ++p.y)
 	{
 		for (p.x = minX; p.x <= maxX; ++p.x)
@@ -114,6 +115,57 @@ void DrawTriangle(VertexData vertexdata[3])
 			if (w0 >=0 && w1 >= 0 && w2 >= 0)
 			{
 				float den = 1.0f/vertexdata[0].clippos.w * w0 + 1.0f/vertexdata[1].clippos.w * w1 + 1.0f/vertexdata[2].clippos.w * w2;
+
+				// TODO: Depth range test
+
+				// TODO: Is it safe to ignore gstate.isDepthTestEnabled() when clear mode is enabled?
+				if ((gstate.isDepthTestEnabled() && !gstate.isModeThrough()) || gstate.isModeClear()) {
+					u16 z = (u16)((vertexdata[0].drawpos.z * w0 / vertexdata[0].clippos.w + vertexdata[1].drawpos.z * w1 / vertexdata[1].clippos.w + vertexdata[2].drawpos.z * w2 / vertexdata[2].clippos.w) / den);
+					u16 reference_z = *(u16*)&depthbuf[p.x*2+p.y*(gstate.zbwidth&0x7C0)*2];
+					bool pass = true;
+
+					switch (gstate.getDepthTestFunc()) {
+					case GE_COMP_NEVER:
+						pass = false;
+						break;
+
+					case GE_COMP_ALWAYS:
+						pass = true;
+						break;
+
+					case GE_COMP_EQUAL:
+						pass = (z == reference_z);
+						break;
+
+					case GE_COMP_NOTEQUAL:
+						pass = (z != reference_z);
+						break;
+
+					case GE_COMP_LESS:
+						pass = (z < reference_z);
+						break;
+
+					case GE_COMP_LEQUAL:
+						pass = (z <= reference_z);
+						break;
+
+					case GE_COMP_GREATER:
+						pass = (z > reference_z);
+						break;
+
+					case GE_COMP_GEQUAL:
+						pass = (z >= reference_z);
+						break;
+					}
+
+					// Clear mode forces depth test func to be ALWAYS
+					if (!pass && !gstate.isModeClear())
+						continue;
+
+					if (gstate.isDepthWriteEnabled() || (gstate.clearmode&0x40)) // TODO: Correct to enable depth writing in the clearmode case?
+						*(u16*)&depthbuf[p.x*2+p.y*(gstate.zbwidth&0x7C0)*2] = z;
+				}
+
 				float s = (vertexdata[0].texturecoords.s() * w0 / vertexdata[0].clippos.w + vertexdata[1].texturecoords.s() * w1 / vertexdata[1].clippos.w + vertexdata[2].texturecoords.s() * w2 / vertexdata[2].clippos.w) / den;
 				float t = (vertexdata[0].texturecoords.t() * w0 / vertexdata[0].clippos.w + vertexdata[1].texturecoords.t() * w1 / vertexdata[1].clippos.w + vertexdata[2].texturecoords.t() * w2 / vertexdata[2].clippos.w) / den;
 				u32 color = 0;
@@ -129,7 +181,7 @@ void DrawTriangle(VertexData vertexdata[3])
 				if (gstate.isTextureMapEnabled())
 					color |= /*TextureDecoder::*/SampleNearest(0, s, t);
 
-				*(u32*)&fb[p.x*4+p.y*(gstate.fbwidth&0x3C0)*4] = color;
+				*(u32*)&fb[p.x*4+p.y*(gstate.fbwidth&0x7C0)*4] = color;
 			}
 		}
 	}
diff --git a/GPU/Software/SoftGpu.cpp b/GPU/Software/SoftGpu.cpp
index 9751eb27f8..f1d7c5457c 100644
--- a/GPU/Software/SoftGpu.cpp
+++ b/GPU/Software/SoftGpu.cpp
@@ -34,7 +34,8 @@ static GLint uni_tex = -1;
 static GLuint program;
 
 const int FB_HEIGHT = 272;
-u8* fb = NULL; // TODO: Default address?
+u8* fb = NULL;
+u8* depthbuf = NULL;
 
 GLuint OpenGL_CompileProgram(const char* vertexShader, const char* fragmentShader)
 {
@@ -135,7 +136,8 @@ SoftGPU::SoftGPU()
 	attr_pos = glGetAttribLocation(program, "pos");
 	attr_tex = glGetAttribLocation(program, "TexCoordIn");
 
-	fb = Memory::GetPointer(0x44000000);
+	fb = Memory::GetPointer(0x44000000); // TODO: correct default address?
+	depthbuf = Memory::GetPointer(0x44000000); // TODO: correct default address?
 }
 
 SoftGPU::~SoftGPU()
@@ -531,6 +533,7 @@ void SoftGPU::ExecuteOp(u32 op, u32 diff)
 	case GE_CMD_ZBUFPTR:
 		{
 			u32 ptr = op & 0xFFE000;
+			depthbuf = Memory::GetPointer(0x44000000 | (gstate.fbptr & 0xFFE000) | ((gstate.fbwidth & 0xFF0000) << 8));
 			DEBUG_LOG(G3D,"Zbuf Ptr: %06x", ptr);
 		}
 		break;
@@ -538,6 +541,7 @@ void SoftGPU::ExecuteOp(u32 op, u32 diff)
 	case GE_CMD_ZBUFWIDTH:
 		{
 			u32 w = data & 0xFFFFFF;
+			depthbuf = Memory::GetPointer(0x44000000 | (gstate.fbptr & 0xFFE000) | ((gstate.fbwidth & 0xFF0000) << 8));
 			DEBUG_LOG(G3D,"Zbuf Width: %i", w);
 		}
 		break;
diff --git a/GPU/Software/TransformUnit.cpp b/GPU/Software/TransformUnit.cpp
index a2906bb91c..f51fa48619 100644
--- a/GPU/Software/TransformUnit.cpp
+++ b/GPU/Software/TransformUnit.cpp
@@ -63,6 +63,7 @@ DrawingCoords TransformUnit::ScreenToDrawing(const ScreenCoords& coords)
 	// TODO: What to do when offset > coord?
 	ret.x = (((u32)coords.x - (gstate.offsetx&0xffff))/16) & 0x3ff;
 	ret.y = (((u32)coords.y - (gstate.offsety&0xffff))/16) & 0x3ff;
+	ret.z = coords.z;
 	return ret;
 }
 
@@ -152,6 +153,7 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, u32 prim_type
 			} else {
 				data[i].drawpos.x = pos[0];
 				data[i].drawpos.y = pos[1];
+				data[i].drawpos.z = 0; // TODO: Not sure if that's what we should do here
 			}
 		}
 
diff --git a/GPU/Software/TransformUnit.h b/GPU/Software/TransformUnit.h
index 1a4844702c..b6f48dd5f4 100644
--- a/GPU/Software/TransformUnit.h
+++ b/GPU/Software/TransformUnit.h
@@ -35,7 +35,15 @@ struct ScreenCoords
 	u16 z;
 };
 
-typedef Vec2<u10> DrawingCoords; // TODO: Keep z component?
+struct DrawingCoords
+{
+	DrawingCoords() {}
+	DrawingCoords(u10 x, u10 y, u16 z) : x(x), y(y), z(z) {}
+
+	u10 x;
+	u10 y;
+	u16 z;
+};
 
 struct VertexData
 {
@@ -51,8 +59,10 @@ struct VertexData
 		clippos.z = LINTERP(t, a.clippos.z, b.clippos.z);
 		clippos.w = LINTERP(t, a.clippos.w, b.clippos.w);
 
+		// TODO: Should use a LINTERP_INT, too
 		drawpos.x = LINTERP(t, a.drawpos.x, b.drawpos.x);
 		drawpos.y = LINTERP(t, a.drawpos.y, b.drawpos.y);
+		drawpos.z = LINTERP(t, a.drawpos.z, b.drawpos.z);
 
 		texturecoords.x = LINTERP(t, a.texturecoords.x, b.texturecoords.x);
 		texturecoords.y = LINTERP(t, a.texturecoords.y, b.texturecoords.y);