Dolphin, the GameCube and Wii emulator - Forums

Tested pokemon XD so far...still slower in LLE then any wii game I tried...



Will get back with baten kaitos results soon.

(09-26-2012, 08:28 PM)rpglord Wrote: [ -> ]We are in luck: I have pretty much every rpg game for gc and wii,so I can test pokemom and baten kaitos.
Will post results a little bit later today.

---

(09-26-2012, 07:45 AM)Axxer Wrote: [ -> ]So why don't we group together all of the games we said we specifically wanted this benchmark to be done on. I'm in; ill test whatever I have.

If you have TP for wii,you should certanly test that one,to confirm its indeed slower in LLE.
You can also test whatever game you want. For example,test SMG2 so we can see if you get same results as me,or Zelda Skyward Sword,I didnt test that because I don't have wii motion +
Lets see if Zelda SS is also only a few fps slower in LLE ?

I'll test those three.

Axxer,ok thanks for contributing

I tested baten kaitos,exactly same fps and fps difference as in pokemon xd




So, 6 fps difference between hle and lle. These 2 games certanly have lowest perfomance impact when enabling lle of all gamecube games I tested so far

---

Here for example are zelda tp gamecube version results ( 7 fps difference I mentioned in first post was in hyrule field-this is when you start a new game):



That is 14 fps..I would call that significant difference

I am looking forward to more tests from other users,but I think Azure Knight is probably right,and what kind of audio type is beeing used effects how LLE will perform.

What about Twin Snakes?

I have the wii version of TP, ill try it out tonight or tomorrow. No vids, you want screenshots right?

Yes,screenshoots are fine.
SnakeBoss,can you test twin snakes ?

I wouldn't mine but that will have to be when I get more free time. Give me a run-down of the criteria to run these benchmarks.

Use dolphin 3.0 and these settings http://forums.dolphin-emu.org/showthread.php?tid=23618
Test with hle and lle make a screnshoots one with hle another one with lle.
You can upload them at imgur.com

Will try to get it done this weekend.

So, finally done some testing. Finally stopped being lazy and told CleanRip to start ripping my Wii discs. Figured I'd start with NSMB because the graphical requirements for this game on Dolphin are already quite low, so it should be at least a decent judge of CPU-bound performance. Using the agreed upon settings (e.g. the speed settings, plus changes for HLE and LLE audio with LLE on Thread). As far as I can tell, the GPU has been pretty marginalized here: IR set to 1x (Window is auto-shrunk to that size too) no AA and AF set to 1x. I even turned off my desktop effects (added about 15 FPS to both HLE and LLE audio).

Nothing special, just World 1-1 after to kill the first Goomba. Haven't played this on Dolphin yet, so I don't know where any intense areas would be to test, so it's just World 1-1 for now. You can click to see larger images of the thumbnails. The averages are the values that Dolphin consistently hung at or near (+/- 1 or 2 frames every now and then).

New Super Mario Bros. Wii - World 1-1 - LLE Audio
Result: 144 FPS average


New Super Mario Bros. Wii - World 1-1 - HLE Audio
Result: 161 FPS average


Both speeds are still well over 2x the normal game speed (for NTSC anyway, nearly 3x if it were PAL :p) but HLE audio performs on average 17 FPS better. That's a little over of an 11% difference in favor of HLE audio. In the context of actually playing the game, it doesn't matter since the speeds are quite ridiculous anyway, but for this benchmark it seems significant enough. Further tests are underway, but it looks like the gap increases with recent revisions (not entirely surprising I suppose).

EDIT: Had a little look-see at the filesystem of NSMB. Looks like it is indeed using streamed audio (BRSTMs under the Sound directory) rather than sequenced audio, if that's worth anything to this discussion.

But this is only examining one admittedly "light" game; many more need to be tested. C'mon people, add your benchmarks. Doesn't matter if your system is strong or weak, we're only interested in the results