XFB real is faster in some games

[Optimis0r]

I have noticed that in most of my games setting XFB to 'Real' is much faster than disabling it or setting it to 'Textures'. I posted this as an issue.

The below screen shots demonstrates this weird behavior:






Other Games
Its not just the games above but these games as well:

Luigi's Mansion
Resident Evil 1 / 4
Zelda WW / TP
Killer 7
The Last Story

I'm not complaining about this behavior... I think its great having more accurate emulation at higher speeds. Perhaps the Performance Guide should be updated and the default Dolphin config should have XFB set to real?

Please could someone else test their games and report? It seems to only happen on D3D11/OGL backends for me. You need to disabled frame limiter and just toggle XFB to/from Real and Virtual whilst the game is running.
[color=#ff0000]
This behavior have been tested on Dolphin-1701 and Dolphin-1722.[/color]

[MayImilae]

Look at the speed percentages. I'll use Xenoblade as the example here.

XFB Real - 242%

XFB Virtual - 244%

Also you are using high compatibility settings and you aren't posting the speed results for XFB disabled. This is a faulty comparison if I ever saw one. The only thing that is demonstrated here is that XFB Virtual is showing half the FPS in it's counter despite running at the same speeds.

[Optimis0r]

Doesn't matter if its DSP HLE / DSP LLE. I have a fast rig and game with DSP LLE and notice the fps drops (with the frame limiter off).

You have a fast rig yourself so perhaps you can help with some screen shot comparisons?
My OP is a sample of two screen shots - would be good to have more from others. But looking at them more closely - it could be a case of wrong fps reading.

Where XFB virtual is slower than real - disabling it provides the same level of fps drops.

Also you are saying this is a faulty comparison and cherry picking your conclusions. What about resident evil 1?

[MayImilae]

Lol. You accuse me of cherry picking as you cherry pick. Resident Evil can easily be explained by the stuff on the tabs you aren't showing here, such as internal resolution (forced at 1x Native in XFB Real, no limits on XFB Virtual). There's also GameINI settings to take into consideration. And again, all of your settings are very high compatibility even in situations where those settings are not needed. And you aren't showing XFB Disabled speeds. You didn't mention any of those in your reply. There is just not enough information here.

Also, didn't you notice that XFB Real was getting double speeds? All of the games you tried are 30fps games. ALL of them. I find that a little convenient, as Real XFB forces 60fps. Hence the double speed stuff you saw.

[Optimis0r]

You are missing the point completely - it doesnt matter the about any other setting - its all about XFB virtual/disabled vs Real.
All things being equal - XFB virtual/disabled is slower in some games than Real.

You are not willing to even do a quick test yourself in a range of games? Basically out of the box without manipulating the ini etc XFB virtual/disabled is slower in SOME games (most the games I play actually) than Real.

BTW - those tests are conducted with enhancements off and native resolution

It doesnt make sense to me either. Thats why I am kindly asking others to test and post results rather dismiss it outright. Don't feel like an argument I am just posting my findings.

My OP states that we should perhaps have XFB to real as the default if more games run faster with it. If we can provide a better experience on Dolphin out of the box then we should. If indeed XFB real is faster in more games than virtual/disabled - then why not have it set to Dolphin's default setting I think everyone wins then.

[delroth]

(08-09-2013, 06:47 PM)MaJoR Wrote: Also, didn't you notice that XFB Real was getting double speeds? All of the games you tried are 30fps games. ALL of them. I find that a little convenient, as Real XFB forces 60fps. Hence the double speed stuff you saw.

Double speed and VPS=FPS. Stop wasting our time please, this is not "development discussion".