Dolphin, the GameCube and Wii emulator - Forums

(10-12-2012, 07:12 AM)rpglord Wrote: [ -> ]I found many times gpu backends affect things they shouldn't,things which are related to cpu speed !

The point was, I don't think we should spend a whole lot of time fussing over the video backend. Yes, changing it between OGL, DX9 and DX11 affects performance, but so will different GPUs and their drivers. Even if two members used the same backend on the same OS with the same GPU, the drivers alone could potentially alter the results for this benchmark. It isn't quite feasible to take all of the factors into account. If you want, we can all agree to stick with OpenGL (or the Software Renderer :p) so that all platforms are using the same video backend, but keep in mind there are other factors besides just this when it comes to GPU related things. Judging from how different your experience and Garteal's was on OpenGL, it appears as if drivers and the GPUs themselves do have an influence. For now, I'll just make a note that my tests are with OpenGL.

Bleh, said we shouldn't spend a lot of time fussing over the video backend, did just that >.<

Moving on, here are those results I took earlier. Wait until you see the analysis :3

New Super Mario Bros. Wii - LLE + VBeam
OpenGL Video Backend
Result: 164 FPS average


New Super Mario Bros. Wii - HLE + VBeam
OpenGL Video Backend
Result: 197 FPS average


Percentage Difference: + 18.2%

Note LLE + VBeam versus HLE (with no VBeam, from my first test). That's a difference of -1.8%, that means LLE + VBeam isn't just at the same performance level as HLE, it's actually above it. Totally unexpected. However, note as well the differences between LLE + VBeam and HLE + VBeam. When they're on equal footing, the gap returns, and this time it jumps to 18%, greater than the 11% from my first test with NSMB. Almost midnight here. Today's been interesting to say the least.

Personally I think it's ridiculous to even suggest that LLE is not going to be much slower with most games over HLE. I for one can play a decent amount of games with my laptop with HLE, if I were to use LLE, the game running at full speed or even close to full speed would be absolutely out of the question.

It doesn't matter if it sounds ridiculous. We now have the data to prove it, so it is true.

(10-12-2012, 03:20 PM)NaturalViolence Wrote: [ -> ]It doesn't matter if it sounds ridiculous. We now have the data to prove it, so it is true.

Wait, which "it" are we talking about? That LLE isn't significantly slower or that it is significantly slower?

(10-12-2012, 03:20 PM)NaturalViolence Wrote: [ -> ]It doesn't matter if it sounds ridiculous. We now have the data to prove it, so it is true.

Your so-called data is wrong. My game speed is drastically reduced with LLE over HLE. The first problem is that all of the tests being done are with high-end CPU's. The second problem is that you haven't actually tired this yourself before commenting.

WHICH GAME?

We have multiple people confirming the same results now. It's very clear that games that use sequenced audio have a high performance hit from using LLE audio while games that use streaming audio have a low performance hit from using HLE.

The performance of your cpu isn't going to affect the performance hit as a differential percent.

(10-12-2012, 03:36 PM)NaturalViolence Wrote: [ -> ]WHICH GAME?

Any game I test. LLE is an order or magnitude slower all the time.

Quote:The performance of your cpu isn't going to affect the performance hit as a differential percent.

The speed percentage is going to be much lower with a much more capable CPU over one that can be considered, not as capable. Once again, basing you opinion on personal experience is much better than guessing.

Quote:Any game I test. LLE is an order or magnitude slower all the time.

Well then do some testing, list the games tested, and post your results. Because so far it's science and credible data against.....nothing.

Quote:Once again, basing you opinion on personal experience is much better than guessing.

My "opinion" as you call it is based on real world data. I will always follow the evidence, where ever it leads me, to do anything else is utterly stupid.

(10-12-2012, 03:55 PM)NaturalViolence Wrote: [ -> ]When then do some testing, list the games tested, and post your results. Because so far it's science and credible data against.....nothing.

I'm not interested in doing any more testing just because you say I should. You'll just have to take my word for it, or not.

Quote:My "opinion" as you call it is based on real world data. I will always follow the evidence, where ever it leads me, to do anything else is utterly stupid.

What is stupid is to not find out for yourself. The idea that '"I read it somewhere so it must be true" is not a valid argument.

The difference between sequenced and streaming audio was indeed a very notable discovery, but by itself hasn't settled what rpglord and I were originally talking about: the differences between LLE audio and HLE audio. Using Wii games with streaming audio significantly reduces the gap between LLE and HLE audio compared to GC games that use sequenced audio (ToS was nearly +83% in my unposted tests, far greater than my two Wii tests), but that doesn't necessarily mean the gap is reduced to an insignificant size when strictly looking at Wii games with streaming audio. My tests show that there's still a difference greater than 10% (I've not been able to reproduce some of rpglord's results where the percent difference was extremely low). Even rpglord's most recent test (OGL, HLE & LLE + VBeam) showed a difference of 12%. I consider those to still be noteworthy differences between LLE and HLE audio. It's debatable what one can define as "significant", but it's not low enough to pass off imo.

@SS - Care to participate in our testing? I mean atm, we really only have rpglord, Garteal, admin89, and myself (LordVador, Axxer, SnakeBoss, they never showed up )
EDIT - You're such a ninja. If you're not interested, that's understandable. Wish I had a spare system or something though.