Dolphin, the GameCube and Wii emulator - Forums

NSMB is wii game,and it should use streamed audio,which is supposed to be faster in LLE,or that is audio type where big difference between LLE and HLE is not to be expected.
Sequenced audio is the one which gamecube games use,and the one where LLE is significantly slower.
Your results so far confirm mine-I was getting around 8-10% fps difference in wii games such as Xenoblade or SMG2-difference in gamecube game is significantly higher,from 30 up to 100%.

I also agree that more people should contribute...c'mon people dont be lazy !

BTW, use the "Log FPS to file" feature when benchmarking

I think i should post here instead

Quote:please provide exact name of the galaxy and I will download saves online and test it

I read a thread (which i forgot the name) that mentioned the most demanding area is lava level at Bowser's castle or sth . Don't blame me if it's not

Xenoblade Benchmark (DX9,2x IR , no AA , 16x AF,Widescreen Hack)





Directx 11 , 2.5 IR , no AA , 16 AF , HLE , Widescreen Hack :

Directx 11 , 2.5 IR , no AA , 16 AF , LLE Widescreen Hack :


Since i want the screenshot in small size , i used JPEG format -> kill quality

(10-11-2012, 11:23 PM)rpglord Wrote: [ -> ]NSMB is wii game,and it should use streamed audio,which is supposed to be faster in LLE,or that is audio type where big difference between LLE and HLE is not to be expected.
Sequenced audio is the one which gamecube games use,and the one where LLE is significantly slower.
Your results so far confirm mine-I was getting around 8-10% fps difference in wii games such as Xenoblade or SMG2-difference in gamecube game is significantly higher,from 30 up to 100%.

Let's do some math on the results from your for Wii game tests:


Only once did your tests give you an 8% difference, and you never hit anything greater than 10%. I've gotten an 11.1% difference on NSMB, which is higher than a lot of your own results. Like I said, I'd consider 10% or above relevant to this benchmark, that's just my opinion though. The point is, there's still a noticeable gap between HLE and LLE in my test. Not as great as with GC games, but it's still there.

(10-12-2012, 12:14 AM)delroth Wrote: [ -> ]BTW, use the "Log FPS to file" feature when benchmarking

Impossible on 3.0, which is what we're using Glad someone finally implemented it though. If we decide to use the most recent rev (and stick to it) that would help. On that note...

@admin89 - We're trying to use 3.0 + Starscream's speed settings (changing between HLE and LLE + LLE on Thread). We're trying to minimize the effects the GPU could play on performance. Thanks for participating though, we appreciate it

Thank you for contributing,but you have dual core and its understandable you will get more perfomance hit when using LLE since you cant enable LLE on thread.

But even so,in most of the game you are only getting 3-4 fps difference,you only got more difference in first screenshoots.
I contribute that to not having lle on thread enabled

---

(10-12-2012, 02:15 AM)Shonumi Wrote: [ -> ]Only once did your tests give you an 8% difference, and you never hit anything greater than 10%. I've gotten an 11.1% difference on NSMB, which is higher than a lot of your own results. Like I said, I'd consider 10% or above relevant to this benchmark, that's just my opinion though. The point is, there's still a noticeable gap between HLE and LLE in my test. Not as great as with GC games, but it's still there.

Yeah its higher then lot of my results,but its pretty close to SMG2 which was 8% like you said.
It's expectable that perfomance hit will very a little bit from game to game,since it varies in my results too.
I think best thing would be if you could test same games I did so we can compare results

SMG2 dumped and tested. Here are my results. Note, since I just dumped this game, there aren't a whole lot of levels I can test (haven't dumped my NAND yet, and at any rate, I've only played SMG2 on a real Wii for like, 5 hours). Tested on the Starship, since that's where the greatest difference was recorded according to your results, rpglord. Same settings as NSMB. Again, since this is a vanilla 3.0 build, I took a screenshot of the FPS value it consistent hung at; noticeable variations weren't greater than +/- 1 or 2 FPS at least not for the minute I stared at my screen.

Super Mario Galaxy 2 - Starship - LLE Audio
Result: 72 FPS average


Super Mario Galaxy 2 - Starship - HLE Audio
Result: 82 FPS average


Percentage Difference: +13%

This one has a greater difference than NSMB and still a significant difference. Will test other areas probably tonight/tomorrow for a greater variety.

Getting similar results to Shonumi on my end.
FPS didn't fluctuate higher than +/- 1 FPS.
Used Starscreams fast settings with 3.0.

Super Mario Galaxy 2 HLE


Super Mario Galaxy 2 LLE


Super Mario Galaxy 2 LLE + VBEAM


New Super Mario Bros Wii HLE


New Super Mario Bros Wii LLE

Well,this is good when more people contribute to testing,so we can discover new things.
I just discovered vbeam speeds up LLE in SMG2,while almost does nothing to HLE.
So,I presume Shonumi and Garteal didnt use vbeam.
I also get 10 fps difference in starship with vbeam off (just tested it ),and 5 fps difference with vbeam on !
I didnt use vbeam with other games though,so this is something related to SMG2 only.
Shonumi and Garteal,it would be great if you can also retest with vbeam

Garteal,for you difference in SMG2 is only 10% since you get a lot more fps then me and Shonumi

Shonumi,your test proves that running SMG2 doesnt require 4.4 ghz Sandy Bridge Unless there is some much more demanding area later in the game,which I doubt

I will test that later, but enabling VBeam emulation is known to mess with Dolphin's timings. It'll give a speed boost, but for the purpose of benchmarking real differences between HLE and LLE audio, it's a factor that will skew the results.

I did say in first post I used vbeam for SMG2 and Skies of Arcadia-but you are right probably should have tested it without it for more "proper" results