I found this topic and go testing 'cause I own SuperCPU128 - I tested only Basic program written at topic start. Results are affected like Robert (our Hydrophilic) wrote - by Basic 7.0 and its IRQs and all features operancy - Basic 2.0 is more, more easily. So here's table of my results (machine: C128DCR PAL, SCPU128 w/16MB RAM): 128 mode without SCPU: VIC slow 84 VDC slow 86 VDC fast 43 128 mode with SCPU at 1MHz: VIC slow 84 VDC slow 85 VDC fast 44 128 mode with SCPU at 20MHz: VIC slow 4 VDC slow 5 VDC fast 5 64 mode without SCPU: slow 74 fast 36 64 mode with SCPU at 1MHz: slow 73 fast 69 64 mode with SCPU at 20MHz: slow 3 fast 3 (64 mode slow is standard, fast is done by POKE 53296,253 - simple like in 128 mode, but VIC is turned on, yes I could turn it off - VIC can't serve screen at 2MHz) Look at results - in all SCPU in speed is best at 20MHz and at 1MHz makes computer bit slower.
BTW: Robert sent me long time ago diskimage with Basic 7.80 and Mandelbrot graph - his C128 it writes in ~1/2 day, my C128 with SCPU at 20MHz in ~30 minutes - 1/2 hour...
Thank your for all your technical info... I would like to compile a table, or 'comparison chart' to publish on my website (free for everyone).
But what unit is your measurement... for example, you say 64 mode (SCPU @ 1MHz) = 73 (slow) or 69 (fast). Let's just talk about 'slow' okay?
So is that seconds (73s), or JIFFYs (73s/60 = 1.22s) or milliseconds (73s/1000 = 0.073) or microseconds (73s/1000000 = 0.000073s) or something else?
Most interesting result is C64+SCPU in 'slow' mode (value 73) is almost the same as C64+SCPU in 'fast' mode (value 69)...
Anyway, your statistics seem very accurate... for example, standard C128 has 84~86 with slow speed but 43 with fast speed!
P.S. Thanks for remembering Mandelbrot with BASIC 7.80!! Yes, it is very, very slow with stock (1 or 2 MHz) CPU... glad it works *reasonably fast* (1/2 hour) with Super-CPU... (most people today can NOT wait 30 minutes... even with SCPU )
Just to see what would happen I used the x[up arrow]2 instead of multiplying by x*x... much slower if just squaring. If we go to 3rd or 4th power we get more equal speed, then above that the power function is faster than just multiplying using x*x(*x*x ad infinitum). There are some odd differences depending on what number we are raising to a power too. Which raises a question: are some functions and calculations faster on different CBM 8bits? I think the power function has some initial overhead that slows it down with simpler raising of powers of small integers.
Let us not talk falsely now The hour is getting late
Agreed... X^2 is much slower than X*X... I think (but can not prove) that X^Q is faster than Q iterations of X*X*X*X... at least when Q is 3 or more... Ahh... program optimization... as relevant today (with GHZ-speed 64+bit processors) as it was back in the day (1MHz 8-bit processor). Don't you love life??! Many things change, but some things remain eternal ? <3
Hello Hydro, I am new here and have only used a 128 in the vice emulator, but I can assure the z80 is running at an effective clock rate of 1 MHz (at least with CP/M 3) because I used to have a CP/M machine in the early 80's (a used Altair 8800 no less ) that had a 2 MHz z80 in it and everything I've tried on the c128 vice emulator runs about half speed. According the manual from Commodore, the c128 is supposed to have a 4 MHz z80 but runs at an effective 2 MHz because of sharing memory with the graphics chip. My way around it, is to set the 'Maximum Speed' (under Options) to 200% if I'm running CP/M on it.