(03-10-2010, 07:56 PM)ImbaMe Wrote: [ -> ]I would like to add a few comparative shots to this thread i can agree that globally 2.0-RC1 is slower in most cases but it seems that all games i tested run better on some revisions and worse on others,
so far the best revision that worked for 85% of my games is R4977 with 95% of the time full speeds.
anything above R5000 is slower and slower for me so i stick with R4977 for now.
I also must add that the revisions post R5000 are very Jumpy in fps, in the same scene where i get 50 fps stable on R4977, it jumps around between 40 to 50 fps on the other revisions with no particular reason.
(PS: i made a zip file because the forum w'ont let me add more than 5 attachments if you want to see the other screenshots DL the Zip)
Dolphin Comparative Shots.zip
I just want to add that The Same exact settings were used on all revisions.
can you compare 5159 with 5163 with you game ?
I suspect the change around 5160 to 5163 can slowdown some games.
(03-10-2010, 08:18 PM)spellforce Wrote: [ -> ]can you compare 5159 with 5163 with you game ?
I suspect the change around 5160 to 5163 can slowdown some games.
Looking on the google code page there was no significant change between r5160 and 5163
R5160:
Rewrote a object-comparison code using a standard function. It's a tiny
refactoring.
R5161:
Reverted the comparison function to r5159 and rewrote map to hash_map. I
confirmed that the code can be compiled and run under Visual Studio 2008. And I
confirmed the code can be compiled under gcc 4.4.0 (mingw). I could not confirm
that the code run under linux.
R5162:
Shut down the video software plugin without closing dolphin in Windows. Fixed a
memory leak.
R5163:
Fixed the crazy code in r5161
but i will test them and see if there is any noticable difference, will report back later.
actually compare 5159 with 5163 is enough,thank you.
I use r5160 x64 and speed is everywhere ok, wind waker right now seems like running full speed: safest texture cache, SSAAx2.25, Vsync and even copy EFB to RAM. For me problems are more about stability and emulation:
- stopping emu still randomly crash, normally not when stopping after few seconds but when after hour(s) of gameplay(wind waker)
- sound in WII games is rather fine but in GC games is skipping(Dsound, OpenAL is even worse) - OpenAL in about ALL games in both GC and WII that I tried. I know that for example in menu in Metroid 1 it used to play fine in some older revs. at least on Dsound.
- Maybe it supposed to be so but copy EFB to RAM cause for example wind waker look low-rez from close distance onwards - in buildings its fine. copy EFB to RAM is needed to make pic-box not crash emu. - copy EFB to texture make crash.
- save states are unreliable, in GC games its better but can still screw gameplay onward(wind waker crashed later after load). You can also lost sound on load.
- games run at 30fps instead of 60 even when my rig should have enough power for many of them - framelimiter is broken(this is know bug)
- music in zelda WW on HLE is messing
- SSAA under D3D seems to not work(I would swear it did in r5135)
- Single core not working - make black screen in many games although not freeze
- OpenGL Win. mouse cursor won't hide for example in games like House of dead overkill
- Loading games like Fire Emblem: PoR may cause load to black screen especially in later chapters(around 20)
- Im not sure if Vsync in D3D is working
- "Apply" button in D3D doesn't "fade" after click in first tab - does in others
This is on r5160 x64 JIT, I hope its not too off, changes to RC1 were minor.
Speed is least problem for me and I play on safest texture with idle skipping off. I could play on copy EFB to RAM fine(Wind Waker same speed) too but some games look worse then.
2.0 RC 1 was the build around 5164 - 5167 and I see the pic 5150 running at full speed. so the speed drop in RC 2.0 maybe cause by some build after 5150.then I suspect 5163 that's all
So i did some digging with R5159 and 63
but there was almost no difference between the two.
Example for MP they both ran at about the same speed
but for TvC r5163 ran better (by 5% stable)
But all in all they both ran slower than pre r5000 builds
R5159 and 63 ran both extremely slow on metroid prime compared to r5150 they ran about 20fps behind
then It's mean some svns after 5150 and before 5159 cause the slowdown ?
This is most likely because new safe texture cache implementation. Some people reported big slowdowns while others(me included) won't see much difference. But that thing is good and fixed many games.
Quote:- SSAA under D3D seems to not work(I would swear it did in r5135)
This
DOES work in r5160. I've tested this in multiple games. If you checked my post in the other thread about Anti-aliasing you'd see I posted comparison screenshots for Tales of Symphonia, where you can clearly see a difference.
uh, you are right, without it wind waker look EVEN more jaggy
