Inhandra

I believe this is true. The settings for "ultra" can vary. It is more meaningful to state what features are on and to what level (shadows, depth of field, ambient occulsion, AA, etc) as well as draw distance, object quality sliders, etc. Things that CAN be measured point for point to see the results.

Sent from my PG86100 using Tapatalk

__________________

NiranV Dean

well thats right , my Viewer is definetly not the fastest anymore , as already said , my Viewer is jump-into-cold water learning project for me , i have nearly no idea how to optimise the render pipeline code-wise , thats why im basing my Viewer on Shining which had from start the best Performance yet for me (not even Kirstens couldnt beat that) , now after everything mostly set up , a lot of changes to settings and "eyecandy" im at a point where i get about 66% of the Performance of where i started meaning i started with 60 FPS which was my highest yet with Shadows on , and now im having about 40 FPS which i think i still good but could be definetly faster , the problem is you can only really optimise one way in Second Life , either you go the Performance way , or the Quality way , and im definetly going the Quality way , but and thats very important to me , i will always try to let people decide at the end what graphic settings they want , they can always fall back to the fastest way like before (now that you can disable Tofu´s new SSAO and return back to the performance way you can gain additional 16% Performance back , making it up to ~50+ FPS) anyway , my viewer isnt created to be in High or Ultra mode , its specially designed to work with Deferred all the way , i wont care much about non Deferred , for me only Deferred counts (well i would definetly like to see a comparison in Deferred with Pictures in Quality )

also Note that Viewer Performance may drastically change from PC to PC , Driver to Driver , Video Card to Video Card , CPU to CPU , as far as i´ve seen that a Dual Core CPU (no matter if its Intel or AMD) and a GF8600 is nothing you will get that much with in any Mesh enabled Viewer because its simply below the requirements

so yes theoretically seen you cannot compare Viewers with each others , atlast not Viewers that strive to go a whole different way

...i will definetly continue going the Quality way , ...well lets say i dont care about Performance anymore , its time for SL to change and do a huge jump into the feature and not stick along with old shit

one last thing from me: there will hopefully never be a Viewer , NEVER that can do everything , if that happens we wouldnt need to take any other Viewer anymore , that would make each Viewer not special anymore

Niran

__________________

Trinity Dejavu

Just a FYI, This is about what we have found and is without doubt down to some of the recent work we did reducing the amount of frame time the UI consumed. (Kitty and I are on mid range hardware and we drink our own coffee ..)

Catznip does not tinker in the render code and uses the same settings as the V3 official out of the box.

KDU vs Openjpg should make little difference in this kind of test as it's only really a factor in decoding the images. KDU rezzes new stuff faster. Standard OpenJPG vs Kirstens .. I found it hard to tell them apart blind.

Our own test of a viewer is to trot in circles in my sandbox doing double jumps (yay kuso!!)

__________________
Catznip Viewer : www.catznip.com

NiranV Dean

i´ve done tests with my Viewer on a Tree packed SIM , with both Kirstens Openjpeg and LL´s Openjpeg , with both exactly and ultra stable 31.0 FPS , doesnt even slide 0.1 left or right ...
meaning in the final render scene it does 0

Romana

Excuse me if this sounds totally dumb, but I'm code -impaired. Where is this SSAO and how do you set it?

Sent from my SCH-I905 using Tapatalk

Gabriell Anatra

Drat.

It's not looking good for me then. I was hoping it was memory here as I only have 2GB. In my case the V3 ones are no more unstable than the V1 looking ones. I only run Singularity and the Cool VL though so they're all using the 2.X rendering chain. That may be it, but I'd still expect LL's to be less stable if memory were the issue.

Adeon Writer

SSAO are ambient shadows. They are a separate setting from regular shadows, they darken edges of walls and such.

In most viewers it's right in graphic preferences as "Ambient Occlusion"

It tends to be a FPS eater although looks great if you can run it.

Andromeda Rage

I do like the sound of that... Henri Beauchamp has mentioned time and again how "bloated" the V2/3 UI code is.

Andromeda Rage

I've been playing with Catznip some this evening and it's definitely the fastest V3 viewer I've tried. Very nice! Awesome job taming that hungry UI.

EricCastanea

I wish I could try Catzip - it crashes immediately on startup, just like the official viewer.

NiranV Dean

LL´s way to optimise performance is to throw out such features and hope people think theres a difference =D


i´ve set SSAO to max size (Quality) , the FPS diff was so small that i noticed this after the video when i was wondering why its minimally stuttering with 31 FPS , and then i saw omgaw i´ve set SSAO higher lul i forgot that in about 10 seconds after i did so xD

Inhandra

ctrl-alt-d at login screen.
Enter Debug Settings
Set "CurlUseMultipleThreads" to FALSE.
Relog.
Fixed.

It's looking like it'll be fixed next Catznip release, mebbe. :3 For official, 3.4.5 (current beta) it is presumably fixed, or so Bao Linden has said so.

NiranV Dean

meaning i can savely merge all the curl thingys now?

Inhandra

Maybe. Best way is to see if you get curl crashes with 3.2.5 beta. I might give it a shot tomorrow afternoon.

Sent from my PG86100 using Tapatalk

Hitomi Tiponi

I'm sorry to see the animated trees go - what we really needed was better Linden wind.

NiranV Dean

wind actually is very good , with the wind vector debug option you can clearly see that theres a halfway good wind simulation :O

Romana

I ran Catznip for a while last night to see if I got the same FPS boost as others here; I didn't, but it didn't crash either. I get about the same FPS in Catznip as I do in Dolphin, generally in the 20-30 range on medium (I don't like to run low graphics unless I have to). Nirans viewer gives me about 17-21 FPS on medium.

ShrugDangle

I was getting 80+ FPS in Catznip last night on Ultra with everything turned on at some random RP sim. Firestorm had locked up so I logged back in on Catznip and never had even the slightest hiccup the rest of the night.

I'm *this* close to shelving Firestorm for my day to day viewer. If Catznip had a built-in AO and derender it would be a foregone conclusion.

Inhandra

Derender is supposedly coming soon.

Sent from my PG86100 using Tapatalk

Romana

Would someone please share their graphics settings with me so I can figure out why I am not getting this marked FPS increase with Catznip?
I have yet to test Catznip on the family computer, which has a nVidia card, so it may be something about ATI, but I am getting the same FPS in Catznip as I do in Nirans.
Admittedly my ATI cards are not gaming cards, and I well not get the 80 FPS that some are, but it seems to me that since Niran himself says his viewer is (currently) not the fastest one, that I should be getting better FPS in Catznip...

Sent from my SCH-I905 using Tapatalk

Inhandra

Well, you said yourself you get roughly 10 FPS more. 10 can be significant, and mean the difference between choppy and fluid. A friend of mine also got 10FPS more even after switching from mid to high. One individual who said they got 80 have an EXTREMELY good rig, and the other didn't say if they were running shadows or anything else of the sort.

Mileage between viewers ALWAYS varies.

Arrehn

Just want to caution on using a small number of data points... In our beta testers of 100+ datapoints (and this is still far too small a sample size) the experiences are all over the map. Some people have much better performance with Firestorm, some with specific versions of Phoenix, some with LL's stock viewer. Which viewer performs best depends in a huge part on your graphics card, how well you can avoiding memory swapping, your other system hardware, your viewer settings, the specific mix of what's in your live renderscene, and how much of the viewer's capabilities you are using. In particular the lighting and shadow rendering (ie, deferred rendering) system has changed recently and will have a different performance profile than older viewers. It's a lot more capable than at previous points in the past, and has different visual characteristics as well. However, it's not always faster.

Basically, you'll have to experiment on your own to see what works best for you. And also keep in mind that anything much over 60fps will not be visible due to typical LCD hardware limits.

__________________
MetaHarpers, Art and Technology for Everyone

Hitomi Tiponi

I agree entirely Arrehn - mine was just one setup on one day - and as expected others have interesting variations, though most are also getting similar results.

One thing I do get fed up with is the "LL are a bunch of ****** because their viewer only runs at 3 fps while super X viewer runs at 50 fps" crowd. No doubt some people have those problems - but they rarely think that the problem may be on their computer or those sliders they played around with in settings.

NiranV Dean

while we are at Deferred Rendering , Shadows have become an upgrade =D dam dam daaaam , will descripe in my Blogpost today and show some comparisons

Romana

I had some bad crashes a few driver versions back, and haven't messed with *any* sliders, except for the main one (low to high). So I guess what I'm asking is, what adjustments do people make that improve their FPS (without killing quality), with Catznip or any other viewer?

Sent from my SCH-I905 using Tapatalk