diff --git a/cachetest.asm b/cachetest.asm
index 1b0b45c..cbd14f2 100644
--- a/cachetest.asm
+++ b/cachetest.asm
@@ -44,8 +44,8 @@ constant MAIN_FROM(0x0000)
 constant MAIN_TO(0x0080)
 constant MAIN_FONT(0x4000)
 
-constant WIDTH(640)
-constant HEIGHT(480)
+constant WIDTH(576) // 640 * 0.9
+constant HEIGHT(432) // 480 * 0.9
 constant DEPTH(2)
 constant VIDEO_BUFFER(0x80400000 - WIDTH * HEIGHT * DEPTH)
 constant VIDEO_MODE(BPP16 | INTERLACE | AA_MODE_2 | DIVOT_EN | PIXEL_ADV_3 | DITHER_FILTER_EN)
@@ -89,7 +89,7 @@ Start:
     li      t1, 0xDEADBEEF
     li      t2, 0xCAFEBABE
     li      t3, 0xABAD1DEA
-    li      t4, 0x12345678
+    li      t4, 0xADD0BEE5
 
     lui     a0, MAIN_BASE
     // spaced out a bit just to see what happens.
@@ -137,8 +137,8 @@ if 0 {
     ori     a0, MAIN_FONT
 
     // show our results on-screen.
-    lli     s0, 64          // s0: X
-    lli     s1, 48          // s1: Y
+    lli     s0, 16          // s0: X
+    lli     s1, 12          // s1: Y
     lli     s2, 0x20 / 4    // s2: number of words to draw
     lui     s3, MAIN_BASE
     ori     s3, MAIN_TO     // s3: start of data to dump
@@ -148,12 +148,12 @@ MainHexDumpLoop:
 
     lw      s4, 0(s3)       // s4: current word being drawn
 
-    addiu   s0, 8 * 8
+    addiu   s0, 8 * FONT_WIDTH
     lli     s5, 8           // s5: inner loop iteration count
 
 MainHexDumpInnerLoop:
     andi    t0, s4, 0x0F
-    subiu   s0, 8
+    subiu   s0, FONT_WIDTH
 
     lui     a0, MAIN_BASE
     ori     a0, MAIN_FONT
@@ -169,10 +169,40 @@ MainHexDumpInnerLoop:
     srl     s4, 4
 
     subiu   s2, 1
-    addiu   s1, 12
+    addiu   s1, FONT_HEIGHT
     bnez    s2, MainHexDumpLoop
     addiu   s3, 4
 
+    la      s3, TEXT
+DrawTextLoop:
+    lbu     a1, 0(s3)
+    addiu   s3, 1
+
+    beqz    a1, DrawTextDone
+
+    lli     t0, 0x0D // delay slot
+    beq     a1, t0, DrawTextLoop
+
+    lli     t1, 0x0A // delay slot
+    bne     a1, t1,+
+    nop
+    addiu   s1, FONT_HEIGHT // next line
+    b       DrawTextLoop
+    lli     s0, 16 // return to base X offset
++
+
+    lui     a0, MAIN_BASE // delay slot
+    ori     a0, MAIN_FONT
+    sll     a2, s1, 16
+    or      a2, s0
+    la      a3, VIDEO_BUFFER
+    jal     DrawChar16
+
+    addiu   s0, FONT_WIDTH // delay slot
+    b       DrawTextLoop
+    nop
+
+DrawTextDone:
     // use our old cache-poking utility for now.
     jal     PokeDataCache
     nop
@@ -180,50 +210,57 @@ MainHexDumpInnerLoop:
     lui     a0, VI_BASE
     li      t1, VIDEO_MODE
     li      t2, VIDEO_BUFFER & ADDR_MASK
-    li      t3, 640 // width in pixels (for the buffer)
-    li      t4, 0 // interrupt on line
+    li      t3, WIDTH // width of the buffer in pixels
+    li      t4, 0 // interrupt on line, 0 to disable (i think?)
     li      t5, 0 // current line; any write clears VI interrupt
     li      t6, 0x03E52239 // timings (split into 4)
-    li      t7, 525 - 1 // lines. subtracting by one enables interlacing.
-    sw      t1, 4 *  0(a0)
-    sw      t2, 4 *  1(a0)
-    sw      t3, 4 *  2(a0)
-    sw      t4, 4 *  3(a0)
-    sw      t5, 4 *  4(a0)
-    sw      t6, 4 *  5(a0)
-    sw      t7, 4 *  6(a0)
+    li      t7, 525 - 1 // lines; subtracting by one enables interlacing
+    sw      t1, VI_STATUS(a0)           // offset 0x00
+    sw      t2, VI_ORIGIN(a0)           // offset 0x04
+    sw      t3, VI_WIDTH(a0)            // offset 0x08
+    sw      t4, VI_V_INTR(a0)           // offset 0x0C
+    sw      t5, VI_V_CURRENT_LINE(a0)   // offset 0x10
+    sw      t6, VI_TIMING(a0)           // offset 0x14
+    sw      t7, VI_V_SYNC(a0)           // offset 0x18
 
     li      t1, 0x00000C15 // divide VI clock to get proper NTSC rate
     li      t2, 0x0C150C15 // likewise (this is only different on PAL)
-    li      t3, 0x006C02EC // 640 pixels per row, starting at 108 units
-    li      t4, 0x00230203 // 480 pixels per column, starting at 35 units
+    li      t3, 0x008C02CC // 576 pixels per row, starting at ... units
+    li      t4, 0x003B01EB // 432 pixels per column, starting at ... units
     li      t5, 0x000E0204 // video burst starts at 14 and lasts for 502 units
     li      t6, 0x00000400 // x offset and x step size (inverse scaling)
-    li      t7, 0x00000800 // y offset and y step size (inverse scaling)
-    sw      t1, 4 *  7(a0)
-    sw      t2, 4 *  8(a0)
-    sw      t3, 4 *  9(a0)
-    sw      t4, 4 * 10(a0)
-    sw      t5, 4 * 11(a0)
-    sw      t6, 4 * 12(a0)
-    sw      t7, 4 * 13(a0)
+    li      t7, 0x02000800 // y offset and y step size (inverse scaling)
+    // setting y offset to 0.5 (it's Q10 fixed point)
+    // reduces interlacing jitter at the cost of a little image sharpness.
+    sw      t1, VI_H_SYNC(a0)           // offset 0x1C
+    sw      t2, VI_H_SYNC_LEAP(a0)      // offset 0x20
+    sw      t3, VI_H_VIDEO(a0)          // offset 0x24
+    sw      t4, VI_V_VIDEO(a0)          // offset 0x28
+    sw      t5, VI_V_BURST(a0)          // offset 0x2C
+    sw      t6, VI_X_SCALE(a0)          // offset 0x30
+    sw      t7, VI_Y_SCALE(a0)          // offset 0x34
 
 VideoLoop:
     lui     a0, VI_BASE
     li      t1, VIDEO_BUFFER & ADDR_MASK
 
 -
+    // wait until we're done displaying the frame.
     lw      t0, VI_V_CURRENT_LINE(a0)
     sltiu   at, t0, 2 + 1
     beqz    at,-
     nop
 
-    andi    t0, 1
-    bnez    t0,+
+    andi    t0, 1   // check if we're on an odd field.
+    li      t2, 0x003B01EB
+    bnez    t0,+    // if we're not, branch.
     nop
-    addiu   t1, WIDTH * DEPTH
+    addiu   t1, WIDTH * DEPTH   // odd field, so offset the image by one row.
+    li      t2, 0x003B01E9 // slightly shorter image so that
+                           // the y offset doesn't cause sampling out of bounds
 +
     sw      t1, VI_ORIGIN(a0)
+    sw      t2, VI_V_VIDEO(a0)
 
     j       VideoLoop
     nop
@@ -250,4 +287,9 @@ PokeDataCache:
     jr      ra
     nop
 
-include "font.asm"
+align(16)
+insert TEXT, "text.txt"
+db 0
+align(4)
+
+include "font.8x16.asm"
diff --git a/dlist.asm b/dlist.asm
index 6baff7c..c27c792 100644
--- a/dlist.asm
+++ b/dlist.asm
@@ -174,14 +174,25 @@ if HIRES {
     gMatrix(view_mat1, G_MTX_NOPUSH | G_MTX_MUL | G_MTX_PROJECTION)
 
     gPipeSync()
-    gSetCombine(15,15,31,4,7,7,7,4, 15,15,31,4,7,7,7,4)
-    gSetOtherMode(G_PM_NPRIMITIVE | G_CYC_1CYCLE | G_TP_NONE | G_TD_CLAMP | G_TL_TILE | G_TT_NONE | G_TF_AVERAGE | G_TC_FILT | G_CK_NONE | G_CD_MAGICSQ | G_AD_PATTERN, G_AC_NONE | G_ZS_PIXEL | Z_CMP | Z_UPD)
+    gSetCombine(0,0,0,4,0,0,0,4, 0,0,0,4,0,0,0,4)
+variable upper(G_PM_NPRIMITIVE | G_CYC_1CYCLE | G_TP_NONE | G_TD_CLAMP | G_TL_TILE | G_TT_NONE | G_TF_AVERAGE | G_TC_FILT | G_CK_NONE | G_CD_MAGICSQ | G_AD_PATTERN)
+variable lower(AA_EN | Z_CMP | Z_UPD | IM_RD | CVG_DST_CLAMP | ZMODE_OPA | ALPHA_CVG_SEL | G_BL_CLR_IN << 30 | G_BL_A_IN << 26 | G_BL_CLR_MEM << 22 | G_BL_A_MEM << 18)
+    gSetOtherMode(upper, lower)
     gGeometryMode(0, G_ZBUFFER | G_SHADE | G_CULL_FRONT | G_SHADING_SMOOTH)
 
     gSetSegment6(model)
     gMatrix(model_mat, G_MTX_NOPUSH | G_MTX_LOAD | G_MTX_MODELVIEW)
     gDisplayList((6 << 24) | MODEL_START)
 
+if 0 {
+    // debug: display coverage values onscreen.
+    gPipeSync()
+    gSetOtherMode(G_CYC_1CYCLE, G_ZS_PRIM | IM_RD | FORCE_BL | G_BL_CLR_IN << 30 | G_BL_0 << 26 | G_BL_CLR_BL << 22 | G_BL_A_MEM << 18)
+    gSetBlendColor(0xFF,0xFF,0xFF,0xFF)
+    gSetPrimDepth(0xFFFF, 0xFFFF)
+    gFillRect(0, 0, WIDTH - 1, HEIGHT - 1)
+}
+
     // finish.
     gFullSync()
     gEndList()
@@ -204,8 +215,8 @@ align(8)
 
 viewport:
     // note that the third parameters here affect the range of Z-buffering.
-    dh WIDTH/2*4, HEIGHT/2*4, 0x3FF, 0 // scale
-    dh WIDTH/2*4, HEIGHT/2*4, 0x000, 0 // translation
+    dh WIDTH/2*4, HEIGHT/2*4, 0x1FF, 0 // scale
+    dh WIDTH/2*4, HEIGHT/2*4, 0x1FF, 0 // translation
 
 view_mat0:
 if FOV90 {
@@ -215,10 +226,10 @@ if FOV90 {
     Mat.X($0001'4C8D, $0000'0000, $0000'0000, $0000'0000)
     Mat.Y($0000'0000, $0001'BB67, $0000'0000, $0000'0000)
 }
-    Mat.Z($0000'0000, $0000'0000, $FFFE'FF00, $FFFF'0000)
-    Mat.W($0000'0000, $0000'0000, $FFFD'FF00, $0000'0000)
+    Mat.Z($0000'0000, $0000'0000, $FFFE'FF9A, $FFFF'0000)
+    Mat.W($0000'0000, $0000'0000, $FFEB'FC00, $0000'0000)
     Mat.rix()
-constant PERSPECTIVE_NORMALIZATION($00FF)
+constant PERSPECTIVE_NORMALIZATION($000A)
 
 view_mat1:
     Mat.X($0001'0000, $0000'0000, $0000'0000, $0000'0000)
@@ -265,6 +276,6 @@ constant MODEL_START(pc() - model)
     gQuadTri(2, 3, 7, 6)
     gEndList()
 } else {
-    constant MODEL_START(0x32B0)
+    constant MODEL_START(0)
     insert model, "res/teapot.F3D"
 }
diff --git a/font.8x16.asm b/font.8x16.asm
new file mode 100644
index 0000000..5b67e53
--- /dev/null
+++ b/font.8x16.asm
@@ -0,0 +1,144 @@
+constant FONT_WIDTH(8)
+constant FONT_HEIGHT(16)
+constant FONT_SIZE16(FONT_WIDTH * FONT_HEIGHT * 2 * 95) // 0x5F00
+
+LoadFont16:
+    // loads a 95-character, 8x16 font
+    // as an RGB5A1 (16-bpp) image to the specified address.
+    // a0: address to load font to (size: 0x5F00)
+    li      t9, FONT_SIZE16
+    addu    a1, a0, t9              // a1: end of output (exclusive)
+    la      a2, FONT                // a2: start of input
+    la      a3, FONT + FONT.size    // a3: end of input (exclusive)
+
+LoadFont16Loop:
+    lhu     t9, 0(a2)
+    addiu   a2, 2
+
+    srl     t1, t9, 12
+    srl     t2, t9, 8
+    srl     t3, t9, 4
+    andi    t4, t9, 0x0F
+    andi    t2, 0x0F
+    andi    t3, 0x0F
+
+if 0 {
+    sll     t1, 2
+    sll     t2, 2
+    sll     t3, 2
+    sll     t4, 2
+} else {
+    // copy lsb to get the full 5-bit range of values.
+    sll     t5, t1, 1
+    sll     t6, t2, 1
+    sll     t7, t3, 1
+    sll     t8, t4, 1
+    andi    t1, 1
+    andi    t2, 1
+    andi    t3, 1
+    andi    t4, 1
+    or      t1, t5
+    or      t2, t6
+    or      t3, t7
+    or      t4, t8
+    sll     t1, 1
+    sll     t2, 1
+    sll     t3, 1
+    sll     t4, 1
+}
+
+    sll     t5, t1, 5
+    sll     t6, t2, 5
+    sll     t7, t3, 5
+    sll     t8, t4, 5
+    or      t1, t5
+    or      t2, t6
+    or      t3, t7
+    or      t4, t8
+
+    sll     t5, t1, 5
+    sll     t6, t2, 5
+    sll     t7, t3, 5
+    sll     t8, t4, 5
+    or      t1, t5
+    or      t2, t6
+    or      t3, t7
+    or      t4, t8
+
+    ori     t1, 1
+    ori     t2, 1
+    ori     t3, 1
+    ori     t4, 1
+
+    sh      t1, 0x0(a0)
+    sh      t2, 0x2(a0)
+    sh      t3, 0x4(a0)
+    sh      t4, 0x6(a0)
+
+    bne     a2, a3, LoadFont16Loop
+    addiu   a0, 0x8
+
+    jr      ra
+    nop
+
+DrawChar16:
+    // draws a 16-bpp character on-screen at the specified coordinates.
+    // a0: font data address (same argument as LoadFont16)
+    // a1: character (range: 32 to 126 inclusive)
+    // a2: X, Y coordinate in pixels: X | Y << 16
+    // a3: output image address
+
+    // exit early if character is outside of valid range.
+    subiu   a1, 0x20
+    bltz    a1, DrawCharDone
+    sltiu   at, a1, 0x80 - 0x20
+    beqz    at, DrawCharDone
+
+    lli     t9, FONT_WIDTH * FONT_HEIGHT * 2 // delay slot
+    multu   a1, t9
+    mflo    t9
+    addu    a0, t9          // a0: character data address
+
+    andi    a1, a2, 0xFFFF  // a1: X
+    srl     a2, 16          // a2: Y
+
+    sll     t0, a1, 1
+    lli     t9, WIDTH * 2
+    multu   a2, t9
+    mflo    t9
+    addu    a3, t0 // offset output by X
+    addu    a3, t9 // offset output by Y
+
+    lli     t9, FONT_HEIGHT // t9: rows remaining
+
+DrawChar16Loop:
+    // character width hardcoded for 8.
+    lhu     t1, 0x0(a0)
+    lhu     t2, 0x2(a0)
+    lhu     t3, 0x4(a0)
+    lhu     t4, 0x6(a0)
+    lhu     t5, 0x8(a0)
+    lhu     t6, 0xA(a0)
+    lhu     t7, 0xC(a0)
+    lhu     t8, 0xE(a0)
+
+    sh      t1, 0x0(a3)
+    sh      t2, 0x2(a3)
+    sh      t3, 0x4(a3)
+    sh      t4, 0x6(a3)
+    sh      t5, 0x8(a3)
+    sh      t6, 0xA(a3)
+    sh      t7, 0xC(a3)
+    sh      t8, 0xE(a3)
+
+    addiu   a0, 0x10
+    subiu   t9, 1
+
+    bnez    t9, DrawChar16Loop
+    addiu   a3, WIDTH * 2
+
+DrawCharDone:
+    jr      ra
+    nop
+
+insert FONT, "res/fonts/NotoSansMono-SemiCondensedMedium.16.i4"
diff --git a/font.asm b/font.asm
deleted file mode 100644
index 05bd3d7..0000000
--- a/font.asm
+++ /dev/null
@@ -1,103 +0,0 @@
-constant FONT_SIZE16(8 * 12 * 2 * 256) // 0xC000
-
-LoadFont16:
-    // loads a 256-character, 8x12 font
-    // as an RGB5A1 (16-bpp) image to the specified address.
-    // a0: address to load font to (size: 0xC000)
-    li      t9, FONT_SIZE16 
-    addu    a1, a0, t9              // a1: end of output (exclusive)
-    la      a2, FONT                // a2: start of input
-    la      a3, FONT + FONT.size    // a3: end of input (exclusive)
-
-LoadFont16Loop:
-    lbu     t9, 0(a2)
-    addiu   a2, 1
-
-    // sign-extend every pixel to get our blacks and whites.
-    sll     t1, t9, 24
-    sll     t2, t9, 25
-    sll     t3, t9, 26
-    sll     t4, t9, 27
-    sll     t5, t9, 28
-    sll     t6, t9, 29
-    sll     t7, t9, 30
-    sll     t8, t9, 31
-    //
-    sra     t1, 31
-    sra     t2, 31
-    sra     t3, 31
-    sra     t4, 31
-    sra     t5, 31
-    sra     t6, 31
-    sra     t7, 31
-    sra     t8, 31
-
-    sh      t1, 0x0(a0)
-    sh      t2, 0x2(a0)
-    sh      t3, 0x4(a0)
-    sh      t4, 0x6(a0)
-    sh      t5, 0x8(a0)
-    sh      t6, 0xA(a0)
-    sh      t7, 0xC(a0)
-    sh      t8, 0xE(a0)
-
-    bne     a2, a3, LoadFont16Loop
-    addiu   a0, 0x10
-
-    jr      ra
-    nop
-
-DrawChar16:
-    // draws a 16-bpp character on-screen at the specified coordinates.
-    // a0: font data address (same argument as LoadFont16)
-    // a1: character (range: 0 to 255 inclusive)
-    // a2: X, Y coordinate in pixels: X | Y << 16
-    // a3: output image address
-
-    lli     t9, 8 * 12 * 2
-    multu   a1, t9
-    mflo    t9
-    addu    a0, t9          // a0: character data address
-
-    andi    a1, a2, 0xFFFF  // a1: X
-    srl     a2, 16          // a2: Y
-
-    sll     t0, a1, 1
-    lli     t9, WIDTH * 2
-    multu   a2, t9
-    mflo    t9
-    addu    a3, t0 // offset output by X
-    addu    a3, t9 // offset output by Y
-
-    lli     t9, 12          // t9: rows remaining (character height)
-
-DrawChar16Loop:
-    // character width hardcoded for 8.
-    lhu     t1, 0x0(a0)
-    lhu     t2, 0x2(a0)
-    lhu     t3, 0x4(a0)
-    lhu     t4, 0x6(a0)
-    lhu     t5, 0x8(a0)
-    lhu     t6, 0xA(a0)
-    lhu     t7, 0xC(a0)
-    lhu     t8, 0xE(a0)
-
-    sh      t1, 0x0(a3)
-    sh      t2, 0x2(a3)
-    sh      t3, 0x4(a3)
-    sh      t4, 0x6(a3)
-    sh      t5, 0x8(a3)
-    sh      t6, 0xA(a3)
-    sh      t7, 0xC(a3)
-    sh      t8, 0xE(a3)
-
-    addiu   a0, 0x10
-    subiu   t9, 1
-
-    bnez    t9, DrawChar16Loop
-    addiu   a3, WIDTH * 2
-
-    jr      ra
-    nop
-
-align(16); insert FONT, "res/dwarf.1bpp"
diff --git a/inc/F3DEX2.inc b/inc/F3DEX2.inc
index 6943c4d..fd06814 100644
--- a/inc/F3DEX2.inc
+++ b/inc/F3DEX2.inc
@@ -159,16 +159,29 @@ constant CVG_DST_WRAP(0x100)
 constant CVG_DST_FULL(0x200)
 constant CVG_DST_SAVE(0x300)
 // more Z stuff
-constant ZMODE_OPA(0)
-constant ZMODE_INTER(0x400)
-constant ZMODE_XLU(0x800)
-constant ZMODE_DEC(0xC00)
+constant ZMODE_OPA(0) // opaque
+constant ZMODE_INTER(0x400) // interpenetrating
+constant ZMODE_XLU(0x800) // translucent
+constant ZMODE_DEC(0xC00) // decal
 // more coverage stuff
 constant CVG_X_ALPHA(0x1000)
 constant ALPHA_CVG_SEL(0x2000)
 // blending
 constant FORCE_BL(0x4000) // no effect?
 
+// blending modes
+constant G_BL_CLR_IN(0)
+constant G_BL_CLR_MEM(1)
+constant G_BL_CLR_BL(2)
+constant G_BL_CLR_FOG(3)
+constant G_BL_1MA(0)
+constant G_BL_A_MEM(1)
+constant G_BL_A_IN(0)
+constant G_BL_A_FOG(1)
+constant G_BL_A_SHADE(2)
+constant G_BL_1(2)
+constant G_BL_0(3)
+
 macro gNoOp() {
     // stalls the RDP (and not the RSP?)
     _g(0x00, 0, 0)
@@ -489,7 +502,7 @@ macro gSetPrimDepth(variable z, variable dz) {
     // fixes the Z value to a constant for the following primitives.
     // useful for 2.5D effects, maybe?
     // TODO: what are the ranges?
-    _g(0xEE, 0, (z << 16) | d)
+    _g(0xEE, 0, (z << 16) | dz)
 }
 
 macro gRdpSetOtherMode(variable omodeH, variable omodeL) {
diff --git a/inc/main.inc b/inc/main.inc
index 8f1ffb0..094e6fa 100644
--- a/inc/main.inc
+++ b/inc/main.inc
@@ -5,7 +5,7 @@ constant HICOLOR(0)
 
 constant MAIN_DECOMP_IMAGE(HIRES & HICOLOR)
 
-constant MAIN_BASE(0x800E)
+constant MAIN_BASE(0x8004)
 constant MAIN_COUNTS(0x0010)
 constant MAIN_SP_TASK(0x0040)
 constant MAIN_XXD(0x0080)
@@ -13,6 +13,8 @@ constant MAIN_DLIST(0x1000)
 constant MAIN_DLIST_SIZE(0xF000)
 constant MAIN_DLIST_JUMPER(MAIN_DLIST - 0xA8)
 
+constant FONT_BASE(0x8005)
+
 if HIRES {
     constant WIDTH(640)
     constant HEIGHT(480)
@@ -21,6 +23,12 @@ if HIRES {
     constant HEIGHT(240)
 }
 
+if HICOLOR {
+    constant DEPTH(4)
+} else {
+    constant DEPTH(2)
+}
+
 if HIRES {
     if HICOLOR {
         // 640x480, 32-bit
@@ -44,7 +52,7 @@ constant TASK_DP_WAIT(0x0002)
 constant TASK_LOADABLE(0x0004)
 constant TASK_SP_ONLY(0x0008)
 
-constant VIDEO_C_IMAGE_SIZE(WIDTH * HEIGHT * (HICOLOR * 2 + 2))
+constant VIDEO_C_IMAGE_SIZE(WIDTH * HEIGHT * DEPTH)
 constant VIDEO_Z_IMAGE_SIZE(WIDTH * HEIGHT * 2)
 constant VIDEO_OUTPUT_SIZE(0x18000) // technically a buffer?
 constant VIDEO_STACK_SIZE(0x8000) // used for dlist calls, pushing matrices, etc?
diff --git a/inc/n64.inc b/inc/n64.inc
index f347fb6..278c266 100644
--- a/inc/n64.inc
+++ b/inc/n64.inc
@@ -15,6 +15,11 @@ include "n64_si.inc"
 include "n64_pif.inc"
 include "n64_util.inc"
 
+// Rates
+constant CLOCK_RATE(93750000)
+constant COUNT_RATE(CLOCK_RATE / 2) // for use with the CP0 Count register.
+constant COP_RATE(CLOCK_RATE * 2 / 3)
+
 // Memory Map
 constant RDRAM_BASE($A3F0)       // $03F00000 RDRAM Base
 constant RDRAM_DEVICE_TYPE($00)  // $03F00000 Device Type
diff --git a/kernel.asm b/kernel.asm
index 8686b35..211921d 100644
--- a/kernel.asm
+++ b/kernel.asm
@@ -408,7 +408,6 @@ K_MI_AI:
     andi    s0, ~MI_INTR_AI
 
 K_MI_VI:
-
     lw      a0, KV_RES(k0)
     lw      a1, KV_MODE(k0)
     jal     SetScreenNTSC
diff --git a/main.asm b/main.asm
index 57824ac..ef7c62f 100644
--- a/main.asm
+++ b/main.asm
@@ -23,9 +23,13 @@ include "kernel.asm"
 
 Main:
 
+    lui     s0, MAIN_BASE
+
+    jal     LoadFont16
+    lui     a0, FONT_BASE
+
 if MAIN_DECOMP_IMAGE {
 DecompImage:
-    lui     s0, MAIN_BASE
 
     nop; nop; nop; nop
     mfc0    t0, CP0_Count
@@ -72,8 +76,7 @@ Start3D:
     mfc0    t0, CP0_Status
     mtc0    t0, CP0_Status
 
-    lui     a0, MAIN_BASE
-    ori     a0, MAIN_DLIST
+    ori     a0, s0, MAIN_DLIST
     jal     WriteDList
     or      a1, s1, r0
     jal     PokeDataCache
@@ -91,9 +94,8 @@ Start3D:
     // only the lowest 12 bits are used, so 00000000 is equivalent to 04001000.
     sw      r0, SP_PC(a0)
 
-    lui     a0, MAIN_BASE
     jal     PushVideoTask
-    ori     a0, MAIN_SP_TASK
+    ori     a0, s0, MAIN_SP_TASK
 
     jal     LoadRSPBoot
     nop
@@ -106,6 +108,8 @@ Start3D:
     EnableInt()
 
 MainLoop:
+    mfc0    s2, CP0_Count
+
     WriteString(S_SP_Wait)
 -
     mfc0    t0, CP0_Status
@@ -117,6 +121,79 @@ MainLoop:
     beqz    t0,-
     nop
 
+    mfc0    s3, CP0_Count
+    subu    s3, s2
+    addiu   s3, COUNT_RATE / (60 * 200) // for rounding
+    li      t9, COUNT_RATE / (60 * 100)
+    divu    s3, t9
+    mflo    s2 // s2: frame budget spent in (integer) percent
+
+    // there are faster ways, but i'll prefer smaller codesize for now.
+    lli     t9, 10
+    divu    s2, t9
+    mfhi    s3 // s3: (RTL) first digit
+    mflo    t0
+    divu    t0, t9
+    mfhi    s4 // s4: (RTL) second digit
+    mflo    t0
+    divu    t0, t9
+    mfhi    s5 // s5: (RTL) third digit
+
+if !HICOLOR {
+
+if HIRES {
+    lli     s6, 64 // s6: X position
+    lli     s7, 48 // s7: Y position
+} else {
+    lli     s6, 32 // s6: X position
+    lli     s7, 24 // s7: Y position
+}
+
+    la      s8, VIDEO_C_IMAGE // s8: output image buffer
+    beqz    s1,+
+    nop
+    la      s8, VIDEO_C_IMAGE_ALT
++
+
+    beqz    s5,+
+    lui     a0, FONT_BASE
+    addiu   a1, s5, '0'
+    sll     a2, s7, 16
+    or      a2, s6
+    move    a3, s8
+    jal     DrawChar16
++
+    addiu   s6, FONT_WIDTH
+
+    or      at, s5, s6
+    beqz    at,+
+    lui     a0, FONT_BASE
+    addiu   a1, s4, '0'
+    sll     a2, s7, 16
+    or      a2, s6
+    move    a3, s8
+    jal     DrawChar16
++
+    addiu   s6, FONT_WIDTH
+
+    lui     a0, FONT_BASE
+    addiu   a1, s3, '0'
+    sll     a2, s7, 16
+    or      a2, s6
+    move    a3, s8
+    jal     DrawChar16
+    addiu   s6, FONT_WIDTH
+
+    lui     a0, FONT_BASE
+    addiu   a1, r0, '%'
+    sll     a2, s7, 16
+    or      a2, s6
+    move    a3, s8
+    jal     DrawChar16
+    addiu   s6, FONT_WIDTH
+
+}
+
     // queue buffers to swap
     lui     a0, K_BASE
     beqz    s1, SwapToMain
@@ -197,6 +274,7 @@ if MAIN_DECOMP_IMAGE {
 }
 include "dlist.asm"
 include "task.asm"
+include "font.8x16.asm"
 
 if pc() > (MAIN_BASE << 16) {
     error "ran out of memory for code and data"
diff --git a/res/fonts/NotoSansMono-SemiCondensedMedium.16.i4 b/res/fonts/NotoSansMono-SemiCondensedMedium.16.i4
new file mode 100644
index 0000000..34c77aa
Binary files /dev/null and b/res/fonts/NotoSansMono-SemiCondensedMedium.16.i4 differ
diff --git a/res/teapot.F3D b/res/teapot.F3D
index c52a5e4..4908765 100644
Binary files a/res/teapot.F3D and b/res/teapot.F3D differ
diff --git a/task.asm b/task.asm
index 26fc43a..edf382c 100644
--- a/task.asm
+++ b/task.asm
@@ -5,10 +5,15 @@ PushVideoTask:
 
     lli     t0, 1 // mode: video
     lli     t1, TASK_DP_WAIT // flags
-    li      t2, UCODE_BOOT // does not need masking (not actually used?)
+    li      t2, UCODE_BOOT & ADDR_MASK // not actually used?
     li      t3, UCODE_BOOT.size
     li      t4, F3DZEX_IMEM & ADDR_MASK
+constant IMEM_SIZE(F3DZEX_IMEM.size) // weird bass quirk
+if IMEM_SIZE > 0xFC0 {
+    li      t5, 0xFC0
+} else {
     li      t5, F3DZEX_IMEM.size
+}
     li      t6, F3DZEX_DMEM & ADDR_MASK
     li      t7, F3DZEX_DMEM.size
     sw      t0, 0x00(a0)
@@ -25,7 +30,7 @@ PushVideoTask:
     li      t2, (VIDEO_OUTPUT & ADDR_MASK) | UNCACHED
     // most commercial games re-use the yield pointer, so i assume it's fine:
     li      t3, VIDEO_YIELD & ADDR_MASK // stores output buffer size
-    li      t4, ((MAIN_BASE << 16) | MAIN_DLIST_JUMPER) & ADDR_MASK // initial DList
+    li      t4, ((MAIN_BASE << 16) | MAIN_DLIST_JUMPER) // initial DList
     lli     t5, 8 // size of one jump command. this is ignored and 0xA8 is used instead
     li      t6, VIDEO_YIELD & ADDR_MASK
     li      t7, VIDEO_YIELD_SIZE
@@ -38,6 +43,8 @@ PushVideoTask:
     sw      t6, 0x38(a0)
     sw      t7, 0x3C(a0)
 
+    nop; nop; nop; nop
+
     // tell data cache to write itself out
     cache   0x19, 0x00(a0)
     cache   0x19, 0x10(a0)
diff --git a/text.txt b/text.txt
new file mode 100644
index 0000000..fa310e3
--- /dev/null
+++ b/text.txt
@@ -0,0 +1,12 @@
+Done with the battles he once
+waged across time, he embarked
+on a journey. A secret and
+personal journey...
+
+A journey in search of a 
+beloved and invaluable friend...
+
+A friend with whom he parted
+ways when he finally fulfilled his
+heroic destiny and took his place
+among legends...