|
D3D and DRAW performance, and bechmarking
#315749 - 10/12/13 08:03 PM
|
|
|
I'd like to ask what is the difference between benchmarking using "mame gamename -bench" and "mame gamename -seconds_to_run". I guess the first one measures merely the cpu performance emulating the game whilest the later shows the real performance. Is this right?
I've been doing some tests with a Core2Duo and a i5, both slightly overclocked. The first one is using MAME 32 bits in Windows XP 32 bits and the i5 is using MAME 64 bits under Windows XP 64 bits. The higher clock of the i5, the most modern technology and the 64 bits are evident in the results (not so sure about 4 cores vs 2). However, when using "seconds_to_run", you can see that the i5 performe worse than the core2duo (¿?¿?). The only reason I can figure out is that the core2duo is using D3D with stretch and the i5 is using ddraw in native resolutions.
How can the differences be so huge? Is really D3D such a big advantage over ddraw? I though it would be the other way around, since in ddraw the resolution is low and there's no need to stretch or use filters than could kill performance.
Finally, I've tried with "-seconds_to_run" using the multithread option and the performance boost was quite important. However, the sound is completelly distorted. Is this normal? Any way to fix the problem?
The benchmarks:
E8500, 3,6ghz 32 bits, D3D
mame pinkswts -bench 120: 312 mame ibarablk -bench 120: 232 mame pinkswts -seconds_to_run 120: 99,43 mame ibarablk -seconds_to_run 120: 99,96
i5, 3,8ghz 64 bits, DDRAW
mame pinkswts -bench 120: 440 mame ibarablk -bench 120: 316 mame pinkswts -seconds_to_run 120: 95,34 mame ibarablk -seconds_to_run 120: 97,01
|
|
|
RATMNL |
Patron Saint of the Totally F*cked
|
|
|
Reged: 02/02/13
|
Posts: 425
|
Loc: 026, NL
|
|
Send PM
|
|
|
Re: D3D and DRAW performance, and bechmarking
[Re: Elaphe]
#315750 - 10/13/13 01:05 AM
|
|
|
Boy do you pick unlucky romsets for the example...
|
"Those voices in his head might not be real, but they have really good ideas!"
|
|
|
Re: D3D and DRAW performance, and bechmarking
[Re: Elaphe]
#315757 - 10/13/13 06:10 AM
|
|
|
|
Stiletto![Administrator Administrator](//www.mameworld.info/ubbthreads/images/adm.gif) |
They're always after me Lucky ROMS!
|
|
|
Reged: 03/07/04
|
Posts: 6472
|
|
|
Send PM
|
|
|
Re: D3D and DRAW performance, and bechmarking
[Re: RATMNL]
#315760 - 10/13/13 07:06 AM
|
|
|
> Boy do you pick unlucky romsets for the example...
Yeah, um.. that's technically a Rule 8 violation. (although to be REALLY technical, that isn't a license violation per se anymore)... http://www.mameworld.info/ubbthreads/rules.php?Cat= But I get the impression you just didn't know.
How about using the MAMEUI benchmark page romsets as an example? http://www.mameui.info/Bench/Bench.htm blitz dolphin gauntleg gradius4 propcycl radikalb scud starsldr
- Stiletto
|
|
|
|
Re: D3D and DRAW performance, and bechmarking
[Re: Stiletto]
#315765 - 10/13/13 09:43 PM
|
|
|
Today I accidently read something about ddraw and d3d. I've always used the d3d option when dealing with stretch, filtering and just one high res mode, but ddraw with low res monitors, individual games inis, etc. because it was the way to go to use native res modes without stretch. I have no idea when this changed.
Today I've tried using d3d in the computer that I have plugged into an arcade monitor and the results on screen are identical: perfect low res, not scaled, not filtered, silk-smooth scrolls and correct sound. HOWEVER, as I already noticed with my benchmarks, the performance is much much better.
I have upgraded my computer to an i5, overclocked it, I've installed Windows 64 and compiled MAME 64 bits... and all these changes don't have an effect in performance as noticiable as simply changing one line in mame.ini. :O
|
|
|