Dolphin, the GameCube and Wii emulator - Forums

Yes,Axxer is right,ofc I meant HLE vs LLE with lle on thread enabled.
Sorry for not beeing verbose
I am not sure it should metter what cpu I have,because I am not going to compare results I get from a game running fullspeed when testing things like this.
For example,if in Last story I get 20 fps with HLE and 19 with LLE,I dont see why there should be more then 10% difference even if run on amd athlon x3 or some laptop quad core running at 1.6 ghz
They would just get much lower fps both in HLE and LLE,difference should remain the same
Or for example 48 vs I think 54 on SMG2 ( would need to verify but its around 10-15% difference for me )
Like I said,only gamecube games seems to run a lot of slower in LLE,not wii games.
Also,not all of them

(09-21-2012, 06:38 AM)rpglord Wrote: [ -> ]For example,if in Last story I get 20 fps with HLE and 19 with LLE,I dont see why there should be more then 10% difference even if run on amd athlon x3 or some laptop quad core running at 1.6 ghz

The difference between LLE and HLE audio is going to be pretty significant on lower-end hardware. As some one who played Dolphin on extremely weak machines before getting something decent, I know this from experience. Here's a dumbed down version of what happens in each.

In HLE audio, the DSP receives mail messages from the CPU basically telling the DSP where to find the audio. Dolphin looks directly at that data and processes it for audio. This is very quick (though not always accurate). Once the DSP is done with that, it sends another mail message back to the CPU declaring that it's finished with that chunk of audio, and the CPU should move on to the next, and repeat the process.

LLE audio goes through the same process as far as communication between the DSP and CPU go. But the DSP has to do a great deal more work. It can't just look at the addresses passed to it via mail messages and start working on the audio right away. It has to process microcode instructions to the letter, which involve various calculations that need to be performed. This is expensive in terms of the resources it'll use on your computer, by a lot compared to HLE.

Like I said, our systems are fast enough to the point where this isn't an issue. We can make up for the extra time LLE needs to spend processing through power alone. Lesser systems can't, at least not gracefully, hence HLE performs much better for them.

I am sorry Shonumi,but that doesnt make sense to me.
LLE needs to do same amount of work ( much more then HLE ) on my system same as on 1.6 cpu.
Same on 5.5 ghz ivy bridge.
Since HLE doesnt have to do that,it should be faster.
Just testing with framelimit off for example in games which run fullspeed should produce significatly more fps on my system with HLE then with LLE
But it doesnt.
You could say that for example disabling framelimit wont produce higher fps with HLE then with LLE because cpu or gpu thread is bottenecking..but why then doesnt enabling LLE lower the fps ? Since it does need to do more work then HLE.
It just doesnt make sense any way you look at it.

How much LLE is slower depends only on which games is played,not on which system
I am pretty sure that with Last story,Xenoblade and SMG2 LLE will not be much slower,no metter what system you run it.
I am also sure LLE will be a lot slower in skies of arcadia,because it is a lot slower on my i5 2400 ( 34 fps vs 84 fps hle !!!! )

If LLE only slowdown game by few fps on my system,there is no reason to not lower the speed by few fps on a slower system also.
But,if game is barely playble with HLE then that few fps might be just the thing it makes game completely unplayble.
Thats a different story.

Also,when talking about low end hardware,we might be talking about dual core cpus. This means not enabling "lle on thread". OFC it will be slower then.

Maybe you just played games which are slower a lot in LLE ?
Did you for example play SMG2 and if yes,what were the differences ?

Oh boy...

(09-21-2012, 07:27 AM)rpglord Wrote: [ -> ]I am sorry Shonumi,but that doesnt make sense to me.
LLE needs to do same amount of work ( much more then HLE ) on my system same as on 1.6 cpu.
Same on 5.5 ghz ivy bridge.

If you're performing the same task, it will take the same amount of "work" or processing regardless of the CPU, clock speed, or architecture. What will change is how fast that process gets done, which is the issue we're discussing.

(09-21-2012, 07:27 AM)rpglord Wrote: [ -> ]Since HLE doesnt have to do that,it should be faster.
Just testing with framelimit off for example in games which run fullspeed should produce significatly more fps on my system with HLE then with LLE
But it doesnt.

Games you tried? Settings you've used? Dolphin rev? Visual proof? I'm not just asking this to be antagonistic (sorry if I'm coming off that way) but let's just be clear anyone can say anything on the internet. Fwiw, I've always encountered just the opposite (see what I did there? )

(09-21-2012, 07:27 AM)rpglord Wrote: [ -> ]You could say that for example disabling framelimit wont produce higher fps with HLE then with LLE because cpu or gpu thread is bottenecking..but why then doesnt enabling LLE lower the fps ? Since it does need to do more work then HLE.
It just doesnt make sense any way you look at it.

You're right, your results aren't the logically expected outcome. Either something is off with your methodology or your system really is better than anyone ever expected. Also, beware of anecdotal evidence. Simply because it's true for you doesn't mean it will behave the same for others. Things like this need to be confirmed many times over by other users.

(09-21-2012, 07:27 AM)rpglord Wrote: [ -> ]How much LLE is slower depends only on which games is played,not on which system
I am pretty sure that with Last story,Xenoblade and SMG2 LLE will not be much slower,no metter what system you run it.
I am also sure LLE will be a lot slower in skies of arcadia,because it is a lot slower on my i5 2400 ( 34 fps vs 84 fps hle !!!! )

No, it definitely depends on both. A decent mid-range system can handle SMG and Metroid Prime on HLE; using LLE will absolutely drop the frame rates. I know this first hand with tests I've done (for MP at least, frame drops occur even with LLE on thread), and having dealt with a handful of complaints here on the forums as to why these games run so slow.

(09-21-2012, 07:27 AM)rpglord Wrote: [ -> ]If LLE only slowdown game by few fps on my system,there is no reason to not lower the speed by few fps on a slower system also.
But,if game is barely playble with HLE then that few fps might be just the thing it makes game completely unplayble.
Thats a different story.

Except there are plenty of cases where users report that game speed on HLE has been at or near full-speed, but will tank (I usually see 60~70%, sometimes worse) when using LLE, even with LLE on Thread enabled.

(09-21-2012, 07:27 AM)rpglord Wrote: [ -> ]Also,when talking about low end hardware,we might be talking about dual core cpus. This means not enabling "lle on thread". OFC it will be slower then.

Maybe you just played games which are slower a lot in LLE ?
Did you for example play SMG2 and if yes,what were the differences ?

I have not dumped either of my SMG games, nor any of my Wii games. Currently only dumping GC games because they're a lot smaller to fit on my 250GB external HDD (I need to get a new one, seriously, but I wanna fill this one up as much as possible first) and it takes less time for CleanRip (I assume) to get the job done. I'm going through another batch of games (Sonic Heroes, F-Zero GX, Paper Mario, Animal Crossing) right now. In the winter, I'm going to start dumping my Wii collection, so eventually.

I don't have any systems with specs low enough and with enough cores to test "LLE on Thread" (they're low end, but yes, dual-core) but you only need to look at all of the threads that come from laptop users and their "powerful" i7s that can't handle LLE. They seem to do fine on HLE though.

(09-21-2012, 03:28 PM)Shonumi Wrote: [ -> ]If you're performing the same task, it will take the same amount of "work" or processing regardless of the CPU, clock speed, or architecture. What will change is how fast that process gets done, which is the issue we're discussing.

No,we are discusing your theory that just because lle is faster on faster systems,difference between lle speed and hle speed should just disappear or shrink,which does not make sense.
If lle is faster,then hle will also be faster.
So enabling lle should have same effect on perfomance on any system.
On faster systems you will get more speed ofc,nobody is denying that.

(09-21-2012, 03:28 PM)Shonumi Wrote: [ -> ]Things like this need to be confirmed many times over by other users.

I agree,we would need to start a new thread,but I am not sure this is something a lot of users would be willing to test and discuss.
Maybe I am wrong ?

(09-21-2012, 03:28 PM)Shonumi Wrote: [ -> ]No, it definitely depends on both. A decent mid-range system can handle SMG and Metroid Prime on HLE; using LLE will absolutely drop the frame rates. I know this first hand with tests I've done (for MP at least, frame drops occur even with LLE on thread), and having dealt with a handful of complaints here on the forums as to why these games run so slow.

You are just confirming what I said.
I just tested Metroid Prime and is about 30% slower with LLE which is significant. ( 31 fps LLE vs 40 fps HLE )
So,as you see,there is no reason for LLE to NOT lower the fps just because dolphin is not run on low end hardware.

(09-21-2012, 03:28 PM)Shonumi Wrote: [ -> ]Except there are plenty of cases where users report that game speed on HLE has been at or near full-speed, but will tank (I usually see 60~70%, sometimes worse) when using LLE, even with LLE on Thread enabled.

I never said there aren't any games where LLE is significantly slower.

(09-21-2012, 03:28 PM)Shonumi Wrote: [ -> ]but you only need to look at all of the threads that come from laptop users and their "powerful" i7s that can't handle LLE. They seem to do fine on HLE though.

It's usually because of the skipping sound,not because LLE is actually a lot slower then HLE.


To sum it up,basically all I am saying is that just because someone has fast computer,does not mean he is unable to test how much LLE effects emulation speed generally.
Also,some games will be a lot slower in LLE,some wont.
How many exactly I cant know because I dont really have a lot of games.
But based on the testing of games I do have,certanly atleast 50% if not more of games will not slowdown more then 10-15% when enabling LLE.
And because I think I am not unable to test LLE effects on speed just because I have faster cpu then some other users,then yes I think even users with low end hardware shouldn't see more then 10-15% slowdown in these games,unless they have dual core cpu.
If you dont agree,we will just have to agree to disagree,because like you said,a lot of other users would need to chime in with their experiences to confirm this or deny.
All I could prove with screenshoots/videos is how games perform on my system with HLE and LLE.
I dont have any other system ( low end or not ) to test

You are mod,if you want to move our discussion into a new thread to spark interest/comments, I wouldn't mind

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]No,we are discusing your theory that just because lle is faster on faster systems,difference between lle speed and hle speed should just disappear or shrink,which does not make sense.
If lle is faster,then hle will also be faster.
So enabling lle should have same effect on perfomance on any system.
On faster systems you will get more speed ofc,nobody is denying that.

Look back to where this discussion began. It started when I claimed that HLE is much faster than LLE. That's the issue. You then claimed that HLE wasn't significantly faster than LLE in your experience. One reason I gave for your experience was that our hardware (which are fairly high-end for Dolphin, compared to most laptops and AMD systems) was sufficient enough to run LLE at or near fullspeed. If HLE runs fullspeed, and LLE runs at or near fullspeed, the difference isn't going to be terribly noticeable. That's why Axxer usually leaves it on LLE. A better way to examine the difference, as you noted, is to turn off framelimiting, then run both. Now in my tests, HLE sometimes runs twice as fast as LLE.

I said hardware makes a difference because this is true when talking about the performance of any CPU-intensive task. The lower the hardware becomes, the lower your LLE performance will be. HLE, by comparison is not CPU-intensive at all; it can run fullspeed on Dolphin's emulated CPU thread just fine, LLE oth often needs to run on a separate thread for this. Additionally it shouldn't slow down the emulated CPU thread, but LLE can, even in its own thread, hence it can slow down the rest of Dolphin. HLE does not not require as much of the CPU's resources as LLE, so when scaling down the hardware, it's going to run better, given Dolphin can run properly at all on the system in question.

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]I agree,we would need to start a new thread,but I am not sure this is something a lot of users would be willing to test and discuss.
Maybe I am wrong ?

A new thread sounds fine, but I'm sure users will be willing to participate if it's done right (like SS's WW benchmark). Even if it's just the two of us, it deserves to be done so that users have at least some info on the subject.

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]You are just confirming what I said.
I just tested Metroid Prime and is about 30% slower with LLE which is significant. ( 31 fps LLE vs 40 fps HLE )
So,as you see,there is no reason for LLE to NOT lower the fps just because dolphin is not run on low end hardware.

Consider why I chose to give an example based on my high-end hardware. There is a significant difference (re-tested last night, something like an average of 65~70 FPS on LLE, and 115~120 on HLE). If you can observe this difference even on fast computers, the difference is going to be just as, or even more noticeable when trying it on weaker computers. You won't get those same speeds with weaker hardware, but you will see that gap between HLE and LLE. That gap isn't going to go away. That's the nature of how Dolphin is coded.

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]I never said there aren't any games where LLE is significantly slower.

I never said you said that either. What you did say is that you don't see how there should be a large gap on lower-end hardware. I'm pointing out that plenty of users have experienced just that.

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]It's usually because of the skipping sound,not because LLE is actually a lot slower then HLE.

The skipping sound is a factor, but I've seen speed matter just as much as the sound quality. Many users who get drops (severe or otherwise) in LLE stick with HLE if it runs fullspeed. They'll take "smooth" gameplay and bad audio over slow gameplay and audio that's still not perfect.

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]To sum it up,basically all I am saying is that just because someone has fast computer,does not mean he is unable to test how much LLE effects emulation speed generally.

I agree, that's why I gave my example of MP on my high-end system. Though as you noted, this is best observed when you uncheck the any framelimiting. If a system (usually it's the high-end ones) runs both HLE and LLE at the game's native fullspeed, obviously that's not going to show anyone any difference.

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]Also,some games will be a lot slower in LLE,some wont.
How many exactly I cant know because I dont really have a lot of games.
But based on the testing of games I do have,certanly atleast 50% if not more of games will not slowdown more then 10-15% when enabling LLE.
And because I think I am not unable to test LLE effects on speed just because I have faster cpu then some other users,then yes I think even users with low end hardware shouldn't see more then 10-15% slowdown in these games,unless they have dual core cpu.
If you dont agree,we will just have to agree to disagree,because like you said,a lot of other users would need to chime in with their experiences to confirm this or deny.

Like I said many people have seen a greater gap than just 10~15%, even with quad-core and enabling LLE on Thread. There already is a large body of evidence in the forums (though there's not videos or screenshots) it's still better than nothing.

(09-21-2012, 08:36 PM)rpglord Wrote: [ -> ]You are mod,if you want to move our discussion into a new thread to spark interest/comments, I wouldn't mind

You know, anyone can start a benchmarking thread in the Hardware forum. It doesn't make it any more "official" simply because I'm a mod Though that does sound like an interesting idea. I have a lot on my plate atm, so if I do decide to do something like that, it'd be on the weekend.

(09-22-2012, 01:11 AM)Shonumi Wrote: [ -> ]You won't get those same speeds with weaker hardware, but you will see that gap between HLE and LLE. That gap isn't going to go away. That's the nature of how Dolphin is coded.

I agree, that's why I gave my example of MP on my high-end system. Though as you noted, this is best observed when you uncheck the any framelimiting. If a system (usually it's the high-end ones) runs both HLE and LLE at the game's native fullspeed, obviously that's not going to show anyone any difference.

Maybe again I wasn't verbose enough
That gap not going away is exactly what I was trying to explain to you entire time. You sounded like you think gap should go away just because someone is using fast computer.
I claimed that LLE isn't significantly slower because in big amount of games,most popular ones,it is the case based on my testing.
Then you said my results cant be valid for users with low end hardware,which would mean you think "the gap" you mentioned will not be the same if not run on same computer.

So,basically,my statement about lle not beeing significantly slower does not mean in any game-just some games

And since we agree that "gap" doesnt change,it would mean my results are relevant even for people with low end hardware

(09-22-2012, 03:07 AM)rpglord Wrote: [ -> ]Maybe again I wasn't verbose enough
That gap not going away is exactly what I was trying to explain to you entire time. You sounded like you think gap should go away just because someone is using fast computer.

My earlier post does sound like I'm saying the gap between HLE and LLE would disappear, but that was under the assumption we were talking about running games at fullspeed; I've repeatedly clarified myself more than once since then. If you examine the difference between HLE and LLE when limiting the frames to a game's native fullspeed they should be fairly close on a high end system. The italicized part is a big stipulation, and it is completely different from implying that the gap will altogether vanish, since it doesn't when no framelimiting is applied. If you've been trying to explain that the gap wouldn't go away this entire time, then no, you haven't been verbose at all, especially since we've largely been debating the size of the gap.

(09-22-2012, 03:07 AM)rpglord Wrote: [ -> ]I claimed that LLE isn't significantly slower because in big amount of games,most popular ones,it is the case based on my testing.
Then you said my results cant be valid for users with low end hardware,which would mean you think "the gap" you mentioned will not be the same if not run on same computer.

I said what's true for you might not be true for everyone else. You've only tested your machine, and your conclusion that the difference between HLE and LLE will remain small regardless of hardware conflicts with many of the issues reported here when trying LLE. The gap will vary depending on the computer, just like every other performance aspect of Dolphin.

(09-22-2012, 03:07 AM)rpglord Wrote: [ -> ]So,basically,my statement about lle not beeing significantly slower does not mean in any game-just some games

And I asked you which games have you tried. You mentioned SMG and TLS, but that's about it. Are there any more (particularly GC games so I can test some myself).

(09-22-2012, 03:07 AM)rpglord Wrote: [ -> ]And since we agree that "gap" doesnt change,it would mean my results are relevant even for people with low end hardware

We agreed that the gap doesn't just go away, not that it doesn't change. We're still at odds over what the size of the gap will be. You still seem to think that since the gap is small for you on your system, the same should be true for people with lesser hardware. I still maintain that on weaker hardware, this gap becomes more and more noticeable, at least until the hardware becomes so weak as to be unable to adequately play Dolphin at all.

I'm just popping in here to say that in three years of browsing this forum this is the first time I have ever seen back and forth wall of text posts between two people (other than myself) comparable to my own. Bravo sirs (I'll eventually get around to actually reading this thread).

(09-22-2012, 04:09 AM)Shonumi Wrote: [ -> ]And I asked you which games have you tried. You mentioned SMG and TLS, but that's about it. Are there any more (particularly GC games so I can test some myself).

I have done some testing.
And I made interesting discovery.
To answer your question: no,they aren't any GC games where lle is not significantly slower !!!
And now comes interesting discovery : they aren't any wii games where LLE is significantly slower !!!
I am thinking about writing a thread titled "LLE is not slower in wii games".

These are my results for some of the games:
LLE vs HLE

Gamecube:

Tales of Symphonia:
In town 100 fps vs 200 fps
In field : 35 vs 75
This means LLE is 100% slower in ToS !

Fire Emblem Path of Radiance:
80 vs 110

Metroid Prime:
31 vs 40 ( I dont know why I get this low speed you mentioned much faster speeds )

Skies of Arcadia:
34 vs 84

Zelda TP:
21 vs 28

And now,wii games :

Monster Hunter tri
39 vs 40

Pandora's Tower
26 vs 28

SMG2:
Starship: 60 vs 65
Flufy Bluff : 52 vs 52
Yoshi Star : 64 vs 67

Last story
18 vs 19

Xenoblade:
42 vs 45

These are some pretty interesting results,dont you think ?
All games were run with framelimit off.

(09-22-2012, 04:09 AM)Shonumi Wrote: [ -> ]We agreed that the gap doesn't just go away, not that it doesn't change. We're still at odds over what the size of the gap will be. You still seem to think that since the gap is small for you on your system, the same should be true for people with lesser hardware. I still maintain that on weaker hardware, this gap becomes more and more noticeable, at least until the hardware becomes so weak as to be unable to adequately play Dolphin at all.

Yes,I still think gap should remain the same,either same fps or same % of emulation speed reduced.
I guess we will just have to agree to disagree