How much memory are you willing to sacrifice?

Speed is nice, as long as it copies 16k from 8000 to 0800 I don't really care how it looks if it's faster.

I need to repeat the copy process three more times (change banks and continue copy), preferably also fit it in the tape buffer. Last time is only leftover 1500 bytes.

this is a non issue, copying 62k (not 64k, because you have to copy to $800) takes 1 second anyway.
You also don't consider that you have to reserve about $100 bytes in bank 0 to actually store the crt init routine + the relocation code, and that you're not actually copying 64k but 62 max ($800-ffff), so every bank need to copy a variable amount of bytes. Not counting that you have to switch roms off when you write, else you will make a mess when copying to $d000. =)
A cart that support 4x16kb banks is Supergames, this is a quick example how to use it. Refer to CRT.txt or your cart documentation on how to switch banks and turn off the cart completely after you're done relocating.
http://iancoog.altervista.org/...es_crt.rar

Code:

; copy from 64kcart rom, 4x16kb banks to $0800-$ffff (62kb usable) ; using Supergames type cart trbuff=$0334 bank=$02 gameinit=$080d     *=$8000     word start     word start     hex c3 c2 cd 38 30 start     jsr $fda3     jsr $fd50     jsr $fd15     jsr $ff5b     cli     ldx #$ff     txs     inx     stx $d011     stx $d020     stx $d021     lda #$c8     sta $d016     lda #$15     sta $d018     lda #$79     sta $d019     lda #$f0     sta $d01a     ldy #$00 lp1     lda reloc,y     sta trbuff,y     iny     cpy #[endreloc-reloc+1]     bne lp1     jmp trbuff     ; this is relocated to $0334 reloc     rorg trbuff     sei     ldx #$0     stx bank     ; 1st bank is NOT copied from $8000 but, let's say from $8100, because     ; you have to put init code and this copy routine at $8000 in bank 0     ; hence from $8000-80ff it's reserved mainloop     stx $df00   ; set cart bank     ldy blocklen,x     ldx #$00 inloop ptr1=*+2     lda $8100,x     inc $01 ptr2=*+2     sta $0800,x     dec $01     inx     bne inloop     inc ptr1     inc ptr2     dey     bne inloop     lda #$80    ; reset ptr1 to $8000, ptr2 is already pointing correctly     sta ptr1     inc bank     ldx bank     cpx #4     bne mainloop exitloop     lda #$0c    ; disable cart (supergames), your cart may vary, check crt.txt     sta $df00     cli     jmp gameinit blocklen byte $3f,$40,$40,$39     rend endreloc     *=$8100     byte 0 ; this is byte $0800, usually game prgs are at $0801     incprg yourgame.prg endofgame     ds [endofcart-endofgame],$ff ; fill rest of the ROM with $ff     *=$18000 endofcart

_________________

-=[]=--- iAN CooG/HVSC & C64Intros ---=[]=-

I'm using an EasyFlash cart and the method I used works fine, just curious if it can be done faster. I'll take a look at that and see if I ever understand it fully...