diff --git a/mandel.s b/mandel.s
index 31097bf..b5c5953 100644
--- a/mandel.s
+++ b/mandel.s
@@ -1,17 +1,21 @@
 ; Our zero-page vars
-sx    = $80     ; 8 bits: screen pixel x
-sy    = $81     ; 8 bits: screen pixel y
-cx    = $82     ; 16 bits fixed point
-cy    = $84     ; 16 bits fixed point
-zx    = $86     ; 16 bits fixed point
-zy    = $88     ; 16 bits fixed point
-zx_2  = $8a     ; 32 bits fixed point
-zy_2  = $8e     ; 32 bits fixed point
-zx_zy = $92     ; 32 bits fixed point
-dist  = $96     ; 32 bits fixed point  
-iter  = $9a     ; 8 bits iteration count
+sx    = $80     ; i16: screen pixel x
+sy    = $82     ; i16: screen pixel y
+ox    = $84     ; fixed3.13: center point x
+oy    = $86     ; fixed3.13: center point y
+cx    = $84     ; fixed3.13: c_x
+cy    = $86     ; fixed3.13: c_y
+zx    = $88     ; fixed3.13: z_x
+zy    = $8a     ; fixed3.13: z_y
 
-temp  = $a0     ; debug temp area
+zx_2  = $90     ; fixed6.26: z_x^2
+zy_2  = $94     ; fixed6.26: z_y^2
+zx_zy = $98     ; fixed6.26: z_x * z_y
+dist  = $9c     ; fixed6.26: z_x^2 + z_y^2
+
+iter  = $a0     ; u8: iteration count
+zoom  = $a1     ; u8: zoom shift level
+temp  = $a2     ; u16
 
 ; FP registers in zero page
 FR0 = $d4
@@ -19,6 +23,77 @@ FRE = $da
 FR1 = $e0
 FR2 = $e6
 
+; High data
+framebuffer_top = $8000
+framebuffer_bottom = $9000
+textbuffer = $9f00
+framebuffer_end = $a000
+
+height = 184
+half_height = height >> 1
+width = 160
+half_width = 160 >> 1
+width_ratio_3_13 = (5 << 11) ; 5/4
+height_ratio_3_13 = (3 << 11) ; 5/4
+
+DMACTL = $D400
+DLISTL = $D402
+DLISTH = $D403
+
+.data
+
+aspect:
+    ; aspect ratio!
+    ; pixels at 320w are 5:6 (narrow)
+    ; pixels at 160w are 5:3 (wide)
+    ;
+    ; cy = (sy << (8 - zoom)) * (96 / 128 = 3 / 4)
+    ; cx = (sx << (8 - zoom)) * ((3 / 4) * (5 / 3) = 5 / 4)
+    ;
+    ; so vertical range -92 .. 91.9 is -2.15625 .. 2.15624
+    ; &horizontal range -80 .. 79.9 is -3.125 .. 3.124
+    ;
+    ; 184h is the equiv of 220.8h at square pixels
+    ; 320 / 220.8 = 1.45 display aspect ratio
+aspect_x:
+    .word 5 << (13 - 2)
+
+aspect_y:
+    .word 3 << (13 - 2)
+
+
+bit_masks:
+    .byte 3
+    .byte 3 << 2
+    .byte 3 << 4
+    .byte 3 << 6
+
+display_list:
+    ; 48 lines overscan
+    .repeat 5
+        .byte $70 ; 8 blank lines
+    .endrep
+
+    ; 184 lines graphics
+    ; ANTIC mode e (160px 2bpp, 1 scan line per line)
+    .byte $4e
+    .addr framebuffer_top
+    .repeat half_height - 1
+        .byte $0e
+    .endrep
+    .byte $4e
+    .addr framebuffer_bottom
+    .repeat half_height - 1
+        .byte $0e
+    .endrep
+
+    ; 1 line 40-column text
+    .byte $42
+    .addr textbuffer
+
+    .byte $41 ; jump and blank
+    .addr display_list
+
 .code
 
 .export start
@@ -181,10 +256,18 @@ positive:
 .macro imul16 dest, arg1, arg2
     copy16 FR0, arg1  ; 12 cyc
     copy16 FR1, arg2  ; 12 cyc
-    jsr imul16_func   ; 470-780
+    jsr imul16_func   ; 470-780 cyc
     copy32 dest, FR2  ; 24 cyc
 .endmacro
 
+.macro imul16_round dest, arg1, arg2
+    copy16 FR0, arg1  ; 12 cyc
+    copy16 FR1, arg2  ; 12 cyc
+    jsr imul16_func   ; 470-780 cyc
+    round16 FR2       ; 5-28 cyc
+    copy16 dest, FR2 + 2  ; 12 cyc
+.endmacro
+
 ; min 470 cycles
 ; max 780 cycles
 .proc imul16_func
@@ -269,8 +352,6 @@ next:
 
 .endmacro
 
-
-
 .proc mandelbrot
     ; input:
     ; cx: position scaled to 4.12 fixed point - -8..+7.9
@@ -355,29 +436,128 @@ still_in:
 
 .endproc
 
+.macro zoom_factor dest, src, zoom, aspect
+    .local cont
+    .local enough
+
+    ; cx = (sx << (8 - zoom))
+    copy16 dest, src
+    ldx zoom
+cont:
+    cpx #8
+    beq enough
+    shl16 dest
+    inx
+    jmp cont
+enough:
+
+    ; cy = cy * (3 / 4)
+    ; cx = cx * (5 / 4)
+    imul16_round dest, dest, aspect
+.endmacro
+
+.proc pset
+    ; screen coords in signed sx,sy
+    ; iter holds the target to use
+    ; @todo implement
+    rts
+.endproc
+
 .proc start
 
-looplong:
-    ; cx = -0.5
-    lda #$f7
-    sta cx
-    lda #$ff
-    sta cx + 1
+    ; ox = 0; oy = 0; zoom = 0
+    lda #0
+    sta ox
+    sta ox + 1
+    sta oy
+    sta oy + 1
+    sta zoom
 
-    ; cy = 1
-    lda #$10
-    sta cy
+    ; Disable display DMA
     lda #$00
-    sta cy + 1
+    sta DMACTL
 
-    jsr mandelbrot
-    ; should have 32-bit -15 in FR2
+    ; Set up the display list
+    lda #.lobyte(display_list)
+    sta DLISTL
+    lda #.hibyte(display_list)
+    sta DLISTH
 
-    ; save the completed iter count for debugging
-    lda iter
+    ; zero the range from framebuffer_top to framebuffer_end
+    lda #.lobyte(framebuffer_top)
     sta temp
+    lda #.hibyte(framebuffer_top)
+    sta temp + 1
+
+zero_page_loop:
+    lda #0
+    ldy #0
+zero_byte_loop:
+    sta (temp),y
+    iny
+    bne zero_byte_loop
+
+    inc temp + 1
+    lda temp + 1
+    cmp #.hibyte(framebuffer_end)
+    bne zero_page_loop
+
+    ; Re-enable display DMA
+    lda #$22
+    sta DMACTL
+
+main_loop:
+    ; sy = -92 .. 91
+    lda #(256-half_height)
+    sta sy
+    lda #(256-1)
+    sta sy + 1
+
+loop_sy:
+    ; sx = -80 .. 79
+    lda #(256-half_width)
+    sta sx
+    lda #(256-1)
+    sta sx + 1
+
+loop_sx:
+    zoom_factor cx, sx, zoom, aspect_x
+    zoom_factor cy, sy, zoom, aspect_y
+    jsr mandelbrot
+    jsr pset
+
+
+    lda sx
+    clc
+    adc #1
+    sta sx
+
+    cmp #half_width
+    beq loop_sx_done
+
+    lda sx + 1
+    adc #0
+    sta sx + 1
+    jmp loop_sx
+
+loop_sx_done:
+
+    lda sy
+    clc
+    adc #1
+    sta sy
+
+    cmp #half_height
+    beq loop_sy_done
+
+    lda sy + 1
+    adc #0
+    sta sy + 1
+    jmp loop_sy
+
+loop_sy_done:
 
 loop:
-    ; keep looping over so we can work in the debugger
-    jmp looplong
+    ; finished
+    jmp loop
 .endproc