Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
svn performance differences?
#1
Hey guys, started playing my copy of valkyrie profile 2 and was astounded at the dfference betwwen version r4962 and r5085 playing this game.

In the opening sequence of the harbour there was roughly 20FPS difference, which is no exaggeration. It was the difference between 60 and 40 in my case.

Has anyone got their favourite pcsx2 version for certain games?

goes without saying, but it was exactly the same settings for both
Reply

Sponsored links

#2
pc specs???
keep in mind that game is a "Megathurst" demanding and known graphic bug in forest dungeon
MAIN RIG:i5-7600(3.5ghz)GTX 1050Ti 4GB/128bit/gddr5,Win 10 Pro(x64),ASUS ROG B250G,16GB Hyper Fury ddr4 RAM,Samsung M.2-SSD 500GB
SUB RIG:i5-4670(3.4Ghz),ATi Radeon HD7770(GDDR5+128-bit+1GB),Win 10 SL(x64),ASUS H8M-E,8GB DDR3 RAM
Reply
#3
an i5 460m @2.5
5650m 1gb DDR3
4gbDDR3 RAM


speedhacks;
EE cyclerate 2/VU cycle stealing 1
everything checked except mVU blockhack
clamping - none for both EE/IOP & VU

gsdx;
2X SCALING
DX11
texture filtering checked
alpha hack and offset hack checked

pretty mid end specs and it runs comfortable at least in the harbour sequence at the start, I know its not a thorough testing but on two different versions with exactly the same sequence a difference of 20 FPS is huge
Reply
#4
GSdx under developement = rising gpu demands, maybe not always, but at least should be expected - that's why people with weak gpu's praise ancient gsdx versions.
Reply
#5
r4962 was released Nov 11' Even in computer terms surely thats still recent?

VP2 seems relatively balanced CPU/GPU wise compared to say rogue galaxy which seems GPU centered. I OC'd my card 150mhz core and 270mhz memory and it gave 3-5 FPS boost at best.

To dismiss it as having a weak GPU - which I don't dispute - its still a little harsh. and all I was asking was if anyone has a favourite version for certain games.

Every emulator I've used is the same; the further it progresses the more things get broken that worked in previous versions.

Dismissing a 20 FPS difference is strange, especially when the game performs and looks just as good in r4962 as it does in r5085, I don't understand where your coming from. Are the GPU demands rising for a reason or just for laughs and giggles?

Surely if a game works better in one version than it does in another by such a huge margin then thats helpful for people on lower end machines and may stop beginners posting if they're intel GMA can play 4x scaling without crashing?

Personally hearing configs that have been tested on a rig that has an i7@4ghz with a 3gb GDDR5 GPU is pointless to pretty much everyone including the person who made the video.

What would be far more useful is a version database of what version works best with what games like the PSP popsloader database which did a very good job of documenting what games worked best with what versions?
Reply
#6
Have you compared to what you get using the last stable release, 0.9.8 r4600?
Reply
#7
r4600 is better than 5085 but dropped to 43FPS at one point where 4962 dropped to 49FPS for a split second, it was on average between 55-60FPS for the majority of the opening scenes.

Close upas are fine its when it shows the environment r4600 drops drammatically, r4962 doesn't have as much of a drammatic drop, very much smoother.
Reply
#8
You didn't get me;P I didn't said your gpu is absolute crap I have hd5670 1gb ddr5 myself which isn't soo much better, I simply said GSdx developement often mean higher requirements of gpu. Last XX revisions in svn concentrated over GSdx rewrites and it probably doesn't stop where it is now. Opposite to pcsx2 core most past GSdx fixes/rewrites were slowing it down, that's why I simply say it's to be expected and don't expect much future with same gpu unless something dramatically stands on it's head.

20fps isn't really soo strange from mayor rewrites, ie. comparing game like Soul Calibur on 0.9.8 and any newer which has improved texture cache and it'll crawl on new versions. GSdx isn't written under one game which you play now, even if current changes are optimalisations and some games could gain a speedup, one or two could get the opposite without any noticeable change in graphics, that's also to be expected;P, simply use version which is nice for you and be happy that's called emulation and I'm not trying to be an ass saying it. I use one of 49xx currently too couse started playing a few games at once and I'm awfully lazy to update when there are soo many changes going.


Edit:
Oh and noticing the thing about list of versions "best" if some people want it, they can do it, but it has little of sense, pcsx2 works often with huge differences on potentially similar hardware;P, especially with loads of newbies(trying to use nice word) thinking there is some "best" config/version for everyone and spamming it around everywhere they can, it could give more chaos and confuse people using it that they could gain.

Yet Another Edit - just noticed this post turned into an awfull stream of thoughts, sorry if that turns unreadable, doing few things at once now and I'm lazy to reformat it;P.
Reply
#9
Sorry, I didn't mean to sound like an ass.

In the grand scheme of GPU's my card is in the lower end of the spectrum but still capable - I've been thinking games these days can't keep up with how GPU'S are shooting ahead in performance. Its been shown that even over thousands of different configurations of PC hardware PCSX2 versions show similar results on differing hardware.

I understand there will be driver issues and other factors that can arise but, on a basic level I'm sure the performance difference would still be there wouldn't it?

But I know where your coming from, My friend has a desktop with a q6600 quad@2.4ghz and a 9600gt which can't manage okami on 3xscaling whereas I can run it happily @60FPS on a laptop GPU with half the bandwidth and half the CPU cores Tongue

And the multiple threads speedhack, makes a big difference to my FPS on most games, but r4600 doesn't have this, it doesn't seem fair comparing the two svn's to the stable , as it would probably produce similiar results if it had this feature
Reply
#10
I think more usefull is simple knowledge of major rewrites couse they're mainly changing the speed. Maybe you find those usefull:

- For example already mentioned texture cache which fixed loads of game and completely destroyed the speed of a few(like described here which was implemented around 4600(through not included in that one).

- Gif transfer rewrite which "happened" at r4821 it decreased cpu load/speeded up a bit quite a few of games for me and seem to not cause any issues meaning it's worth checking.
- For any VU/EE heavy game or generally a multi core cpu users r4865 brings on MTvu speedhack which everyone loves;P.
- Latest GSdx versions also change alot of stuff giving different results like you see with VP2, they also fix quite a bit of stuff, but it's mainly software mode.

Most other stuff unless I forgot any other important change affecting speed, is pretty much oriented over small fixes and is worth checking only if you experience some bugs in your game, you can easily search through repository with google:
Code:
site:http://code.google.com/p/pcsx2/source/list "game name here"
For just playing, ie not interested soo much over testing new bugs(I mean features;P), I don't see much point of hunting for best pcsx2 version for each game. It's damn hard to notice small fps increase/decrease in pcsx2 anyway, as even with frame limiter on and playing on full speed it jumps around by 2-3 fps constantly.

Oh and I didn't found your post offending;P, just didn't want my own to be taken like that either. Actually if you're testing and see your game slowed down radically or got bugged, make sure to find exactly which revision slowed it down. Posting that info at google code have biggest chance for eventual in-deepth explanation or even a fix, through by general rule which applies most of the time, speed drop in svn versions of pcsx2 is not considered a bug.
Reply




Users browsing this thread: 1 Guest(s)