Speed Regression in latest revs- some Benchmark

[IceD_X]

So guys..
I ve see a big slowdown in latest revisions...

Revs compared:
6901 vs 7168(XtremeToDamax build) - both x64

r6901 is the 20 to 50% faster compared to r7168.
My benchmark with Auto FPS to OFF,Audio Throttle OFF - 800x600 windowed - DX9 - Accurate Texture Cache to Fast - Format Change OFF - Cpu Access Off - EFB to Texture:
-Resident Evil Rebirth WII (Hall video [Jill-Barry-Wesker]) :r6901 - 146% / 7168 - 111%
-Resident Evil 0 Wii (Start Game - 1st level): r6901 - 163% / r7168 - 165% (only this game take a 2% speedup compared to r6901)
-NSMB Wii (Big Cake video demo) : r6901 - 148% / r7168 - 98~110%
-Zelda Twilight Princess (In Game Start): r6901: 221~225% / r7168: 176%

I love the improvements made to the GFX for games, introduced by NeoBrainX, but unfortunately I can not use them, because of the slowness in latest versions.
For now I'll stay to the 6901 version with multi-folders specified for game.

I try to help as much as possible the tracking of issues.
But the best I can do is test the games
I hope to be a help.

Regards IceD_X

[RaptorZX3]

i think one of the best games to do benchmarks with is F-Zero GX, since it got some issues with tracks, sometimes getting slower due to more polygons being shown (probably the emulator's problem, i think my PC is good enough to handle this game at 100% normally).

This game count on speed most of the time for racing, so stutterings and slowdowns really cut out all the great fun you get while playing this game, so that's why i think it's a very good game to do benchmarks with.

[IceD_X]

I am aware ...
But unfortunately I do not have the game in question ...

Should I buy it but honestly it is not my favorite kind of game.
If anyone can make the test with F-Zero GX I would be very grateful.

[Lordani66]

I have checked on MK armageddon if you are right and hell man! You were right. Almost no fps drops. I get some only depending on arenas I play on. And these are not big. Max i checked was 3-4 fps drop.

[IceD_X]

Lordani try with other games, and carefull about config test:

- Auto FPS : OFF
- Audio Throttle : OFF

this can greatly increase the fps drop.
I started this research, for two reasons:
-Support Developers for possible decrease in performance
-Help users who have a PC not very powerful.

Probably not everyone will notice the decrease in performance, because to play properly, you must enable the audio throttle or the auto fps limit. (60 or 30 fps many times).
At least this happens in normal conditions.
When should I test two versions of dolphin on a fast PC, however, need to see what is the maximum peak performance.

Lordani, try a game that really works perfectly at 30 or 60fps in your PC.
Then try to create a configuration of dolphin, but without FPS limit (Not auto but OFF)and without Audio Throttle.
Configure dolphin in windowed mode, and controls the speed%, generated by the ratio FPS-VPS.
Thanks alot for your feedback!

[Lordani66]

You are telling me auto FPS and Audio throttle should be disabled, and that will give me less fps drops?

[IceD_X]

No. The increase in the FPS drop I will be comparing two versions of dolphin.
In particular, by comparing the peak% of both versions.
Let me explain.

Ex
We have Mortal Kombat Armageddon.

Test 6901 (Audio & Auto Throttle FPS ON):
Arena 1: 100%
Test 6901 (Audio & Auto Throttle FPS OFF):
Arena 1: 115% (here reach the limit of computing power)


Test 7168 (Audio & Auto Throttle FPS ON):
Arena 1: 93%
Test 7168 (Audio & Auto Throttle FPS OFF):
Arena 1: 93%


For example, you notice that you play with the r6901 works 100% and then at 60 fps
With r7168, however, the game goes to 56 fps.
Obviously, you notice a regression of only 4fps and therefore a difference of only 7%
This is because it uses the FPS Limit and Audio Throttle.
These two elements will not let you see the game at full power (69fps - 115%), but you will always see the game at 60fps (100%)
... But if we analyze the maximum peak performance between the r6901 (115) and 7168 (93%) the difference is 22% (13fps)!

Paradoxically, if you do not analyze the peak performance, you may not see any regression between versions of dolphin, if you have a very powerful hardware
This is because you may have with a 180% power r6901 and r7168 with a 120%.
In fact, the speed is there and regression of 60%, but your hardware is so powerful to leave out the regression of r7168, showing the game at 60fps.

I know I suck in explaining things accomplice my bad English ... but I still wanted to try.

[NaturalViolence]

Narrow it down to a specific revision. Test 6970 next. You should also post screenshots of all of your settings, not just list some of them.

Quote:I love the improvements made to the GFX for games, introduced by NeoBrainX, but unfortunately I can not use them, because of the slowness in latest versions.

Which changes?

[RaptorZX3]

so far a lot of peoples say that R6970 is the best revision.

I play Mario Kart Wii with this revision, but sometimes i found that the framerate drop at 98%, not enough to disrupt the game, but if you look at the checkered line on some tracks (Mario raceway, Luigi Raceway...), you can notice it when it drop from 100% to 98%.

F-Zero GX play fine with this revision, except some tracks slow down when "too many polygons" are being shown (Port Town tracks i think). Videos are not fixed, and Sand Ocean track(s) are still slowing down by half framerate because of the "heat" effect.

Any improvements since then?

[NaturalViolence]

Personally I don't see any difference in performance at all between 6970 and 7150. Both give me 70% speed in world 2 mario ship in SMG2 with use fps for limiting on and audio throttle off.