dosbox cpu speed target
category: offtopic [glöplog]
What's a good cpu speed to target your demo for in dosbox, 40000 cycles?
MAX CYCLES 100%!!!!!
depends on the demo/intro because some are faster and some are slower. so you have to try and search for smooth running.
sensenstahl: I think sigflup is asking from a dev point of view: "good cpu speed to target your demo" :)
Quote:
Estimate the CPU power wanted by the game in megahertz. Multiply by 1000, and set it in the cycles= setting in dosbox.conf. For example, if it requires 3 MHz, set cycles=3000. If it requires 120 MHz, set cycles=120000. Use the largest possible value that you think makes the game run more fluently, but not larger. If uncertain, use 40000. This corresponds to an average 486.
http://tasvideos.org/DOSBox.html
Dosbox is nowhere near cycle-accurate, so if you intend to make some kind of retro demo, don't use Dosbox as a target.
A 'cycle' is an instruction in Dosbox, so you get single-cycle div/mul and super-fast FPU operations as well, where real retro platforms can't even run most instructions in a single cycle, and some of these instructions can take over 100s of cycles.
Use PCem as a more realistic target.
A 'cycle' is an instruction in Dosbox, so you get single-cycle div/mul and super-fast FPU operations as well, where real retro platforms can't even run most instructions in a single cycle, and some of these instructions can take over 100s of cycles.
Use PCem as a more realistic target.
What Scali said.
And also Mhz=cycles didn't work for me. Around 8000-10000 I had something similar to 386 with 33mhz. 16000 maybe a 486dx33 or 66. I think I did some comparisons with Norton SI between a real oldschool PC and dosbox. Also, videoram bandwidth, as fast as hell on dosbox.
I would still consider dosbox for easy testing and PCem for more accurate later tests.
And also Mhz=cycles didn't work for me. Around 8000-10000 I had something similar to 386 with 33mhz. 16000 maybe a 486dx33 or 66. I think I did some comparisons with Norton SI between a real oldschool PC and dosbox. Also, videoram bandwidth, as fast as hell on dosbox.
I would still consider dosbox for easy testing and PCem for more accurate later tests.
But maybe the question was what's a good number of cycles that modern and not so modern PCs wouldn't be too slow to emulate in dosbox? With my crappy asus eeepc netbook I could still run at 16000 perfectly and sometimes 30000-40000 without sound crackles, depending on the demo (and the setting cpu=dynamic which doesn't work with few demos). So,. 40000 is a very safe estimate, even 65536 (I know, I like to use such numbers for my cycles).
@ruairi :
cycles <-> mhz conversion is not that simple (or at least it is not linear), this is because cpu architecture change a lot from one generation to another (for eg, 286 to 386). generally speaking, newer cpus can "crunch" more instructions per cycle.
from this page (http://pandorawiki.org/DOSBox) : 386 33 mhz => approx 5000 cpu cycles.
while a emulating speed of low end 486 (25-50mhz) would require at minimum 20000.
cycles <-> mhz conversion is not that simple (or at least it is not linear), this is because cpu architecture change a lot from one generation to another (for eg, 286 to 386). generally speaking, newer cpus can "crunch" more instructions per cycle.
from this page (http://pandorawiki.org/DOSBox) : 386 33 mhz => approx 5000 cpu cycles.
while a emulating speed of low end 486 (25-50mhz) would require at minimum 20000.
Tigrou: Cool, thanks for the info. I've found that cycles=max works in most situations for me - but you and Bugo the Cat are absolutely right.
What would people frown on? Anything past 200mhz?
DosBox doesn't do MMX and the fastest non-MMX Intel processor was 200 MHz. So I'd consider 200 MHz the absolute maximum.
