Pozwalam sobie przekleić zalinkowany przez krashana test OS4 vs MOS wykonany na Pegasosie2 przez Faba http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=28068&forum=33&start=140&viewmode=flat&order=0#478205
a tu to samo (niestety bez wykresu):
Machine: PegasosII G4@1GHz, 512MB RAM, Radeon 9250 (64bits version).
OS setup and state:
- OS4.1, vanilla setup, idle (about 2-3% cpu usage), composition enabled, INTERRUPT=YES in Radeon monitor.
- MorphOS2.2, vanilla setup, idle (about 2-3% cpu usage), 3d layers enabled.
Tests scenarii:
1. MPlayer tested with: http://fabportnawak.free.fr/benchmark.avi (DDE3 Quake video, xvid, 1024x576)
2. Quake3/OpenArena, tested on demo four, with "normal" presets.
3. MAME/XMAME tested with "Street Fighter Alpha 2".
4. Snes9x/WarpSnes tested with "Super Mario 4".
1. MPlayer
----------
Conditions:
MorphOS uses "MPlayer r28114 (15.12.2008)" available at: http://aminet.net/gfx/show/MPlayer-1.0-20081215.lha
OS4.1 uses "MPlayer Leitmotiv" available at: http://www.amigasoft.net/downloads/MPlayer.lha.lzh
Input file:
http://fabportnawak.free.fr/benchmark.avi stored in RAM: (XVID, 1024x576, 100s)
Command line:
stack 1000000 (only needed for OS4 test)
MPlayer -quiet -nosound -benchmark ram:benchmark.avi
Results:
OS4.1:
BENCHMARKs: VC: 126.960s VO: 0.124s A: 0.000s Sys: 9.558s = 136.641s
BENCHMARK%: VC: 92.9147% VO: 0.0904% A: 0.0000% Sys: 6.9949% = 100.0000%
Needs 136s to play this 100s clip (with sound enabled, it's completely unsynchronized).
MorphOS:
BENCHMARKs: VC: 78.309s VO: 0.100s A: 0.000s Sys: 1.895s = 80.304s
BENCHMARK%: VC: 97.5156% VO: 0.1250% A: 0.0000% Sys: 2.3594% = 100.0000%
80s needed to play this 100s clip.
Conclusion:
MorphOS MPlayer is 70% faster than OS4 MPlayer on this file. That said, MPlayer for OS4 should really be updated to make a more fair comparison.
Also to be noted latest DVPlayer demo version was tested, but it actually proved to be quite slower than MPlayer OS4 on this file, so it wasn't used in this test.
2. Quake 3
----------
Conditions:
- MorphOS uses this version of Quake3: http://bigfoot.morphos-team.net/files/morphosquake3r2.lha
- OS4.1 uses this version of Quake3: http://www.os4depot.net/share/game/fps/quake3.lha
- OS4.1 uses this version of OpenArena: http://capehill.kapsi.fi/quake3/openarena.lha
- "Normal" preset selected (640x480, ...)
- vsync off in both (easily checkable with the constant tearing in menus).
- OpenArena has been tested on OS4 because it's said to be a bit faster than original Quake3.
Testing method:
in console, type:
timedemo 1
demo four
Results:
OS4.1:
Quake3: 24.7fps
OpenArena: 26.1fps
MorphOS:
Quake3: 49.3fps
OpenArena: untested, but should give similar results.
Conclusion:
MorphOS Quake3 is 99% faster than OS4 Quake3 and 89% faster than OS4 OpenArena.
3. MAME
-------
Conditions:
- MorphOS uses 0.104 and 0.120 versions available at: http://fabportnawak.free.fr/mame/
- OS4.1 uses XMAME.SDL 0.97 available at: http://os4depot.net/share/emulation/gamesystem/xmame.lha
- Both tested in fullscreen mode.
Tested ROM:
sfa2 (Street Fighter Alpha 2)
Testing Method:
Display FPS counter (F11).
Measure first value after inserting credits ('5').
Measure second value after pressing start ('1') (i.e when choosing the character).
Command line:
OS4.1:
stack 1000000
xmame.SDL -nothrottle -effect 0 -nodoublebuf sfa2
MorphOS: mame -nowindow -nothrottle -effect 0 sfa2
Results:
OS4.1:
First value: 88fps
Second value: 68fps
MorphOS:
0.104
First value: 100fps
Second value: 75fps
0.120
First value: 95fps
Second value: 74fps
[EDIT]
0.97 MorphOS version was actually testable too.. Here are the (not very different) results:
First value: 103fps
Second value: 78fps
[EDIT2]
Also, on a game like 1943, the blitting optimisations are much more visible.
After inserting credits:
OS4.1: 200fps
MorphOS: 250fps
Conclusion:
If we compared the closest MAME versions (0.104 to 0.97), MorphOS MAME is between 10 and 13% faster.
To be noted that the most recent a MAME version is, the slowest it is, so OS4.1 0.97 version had an advantage there. Also, on a game like SFA2, the optimisations in blitting are less important, since the most time is spent in CPU emulation, which is basicly the very same code in both versions.
[EDIT]
When comparing the same version, using 1943, MorphOS MAME is actually 25% faster.
4. Snes9x
---------
Conditions:
- MorphOS uses snes9x 1.43 available at: http://fabportnawak.free.fr/snes/snes9x-1.43.lha
- OS4.1 uses warpsnes 1.43 available at: http://os4depot.net/share/emulation/gamesystem/warpsnes.lha
Tested ROM:
super mario world 4.
Testing Method:
Since FPS isn't exactly precise, CPU usage is measured during intro screen (when mario runs).
Command line:
OS4.1: Default options in WarpSnes, and speedhacks disabled (increased stack in warpsnes icon, else it just crashes).
MorphOS: snes9x -ov -nospeedhacks (overlay output)
snes9x -nospeedhacks (no overlay output)
Results:
OS4.1: 65% cpu usage
MorphOS: 50% cpu usage with overlay, 56% cpu usage without overlay
Conclusion:
MorphOS Snes9x is 30% faster (1/(50/65) - 1) when using overlay output and 16% faster otherwise.