View Full Version : AR3 Full GI Droping Frames
paulselhi 09-05-2008, 10:14 PM I am rendering out a test on AR3 and am getting some odd GI calculations best shown in these screen grabs, the last shows the full view and the first 2 show the problems. I have 3 objects that are keyframe baked falling to the floor with TP emitting particles on impcat. The issue only occures just after impact when the tp starts though the first frames with TP are OK it is juts a few odd ones at the early stages
any idea what is causing this ?
|
|
Per-Anders
09-05-2008, 10:17 PM
Are you using TP? Otherwise see what the final render result is, the prepass only renders what it needs and doesn't alays indicate what the final result is going to be.
paulselhi
09-05-2008, 10:21 PM
yse i am using TP and will probably haver a final render in a few days !!
Per-Anders
09-05-2008, 10:52 PM
There are certain limitations with TP in particular and GI full animation mode, however if the problems do manifest, can you post the scene file with the missing frames problem showing?
paulselhi
09-06-2008, 12:21 AM
That sounds scary.."limitations with TP" But it would not suprise me to find out that AR3 has bugs..the hype seems to good to be true
I will post the file if the render finishes before my demo period expires
ThePriest
09-07-2008, 01:30 AM
--remove--
Per-Anders
09-07-2008, 02:55 AM
If you can use MoGraph to do your cloning/generation of particles rather than the TP Geometry (which you'd normally only use if you wanted to use Xpresso to control the particle shapes or were using things like the TP Shatter) then it should work fine.
paulselhi
09-07-2008, 07:45 PM
Thanks Per but if that means that TP does not play well with AR3 and Mograph does not come free with TP then there is a problem
I am half way through the render ( GI took 30 hours to compute 1000 frames at 640 x 380 on a dual core 2.4 ghz 2GB RAM , the GI weighs in at 4.5 GB on disk and I am at render frame 495 after 47 hours total- (OK knock off a couple hours for surfing etc)
From the stills there ARE poblems with those frames i showed earlier, a quick "page flip" using irfanview shows a lot of flicker on the ground maybe due to AA but more worryingly some GI flicker
I will post the final animation probably tommorrow
paulselhi
09-08-2008, 01:14 PM
OK i stopped the render after about 72 hours..blimey this WAS only a test !!
http://www.black-and-white-to-color.com/stuff/blockar3.mov
There is definately an issue with TP though it seems to be intermittent, perhaps only at the start of the emission ? however look at the later frames. This is also compounded by an issue with the sun shadows. It seems that , in this scene at least, the hard and area sun shadows are unstable. I have done a test with a variety of lights ( infinite spot etc) and it seems that the problem is not just sky, soft shadows are OK but sun in sky does not have soft shadows. Perhaps a scene scale issue ?
The flicker on the floor is not due to bump as i have no bump !! so it must be an AA issue, however ncreasing the AA would lead to even more dire render times as would increasing the GI ( there are gi artifacts as well as shadow issues)
Overall i was pretty dissapointed considering the crazy pre calc times..it is a poor test result !!
Any way I am sure others could do better so here is the scene, note sub poly disp is off in the render options
http://www.black-and-white-to-color.com/stuff/fallingblock.rar
Simon Wicker
09-08-2008, 02:56 PM
paul, just trying to do anything with that scene brought my computer grinding to a complete halt (and this is even before playing around with GI settings).
my first thoughts would be that you would get a better result using the old smb + very low stochastic (or QMC as it is known now) sampling but even then i tried rendering frame 200 to see what it would look like and i eventually force quit after it spent ten minutes and had only reached the third sub-frame.
is it the tp causing it to be so slow?
cheers, simon w.
paulselhi
09-08-2008, 03:51 PM
The TP itself was not an issue with me. note this scene was originally 10.5 saved as r11
Simon Wicker
09-08-2008, 05:10 PM
the problem is that if even the bare scene is not playable then it will be difficult to iterate through the various settings for the GI to see what could be done to eliminate any problem areas.
cheers, simon w.
paulselhi
09-08-2008, 07:42 PM
The scene is fine for me on a pc, it uses about 1 GB page file and 1 GB RAM when rendering
I am testing with vray and am at frame 274 after 5 hours use pure brute force ( pri and secondary) at a larger resolution in r 10.5 800 x 340. So far, touch wood- no pun intended !!, the results look very good
lllab
09-08-2008, 08:03 PM
dear Paulsheli,
glad it works for you,
you will be even faster if you use brute force + lightchache method (instead of only brute force) in vray, this is the most advanced method for these things.
you can always email me for settings or go to our forum for specific settings help hot to setup LC+BR.
best greetings
Stefan
vrayforc4d
paulselhi
09-08-2008, 11:21 PM
vray does make a good render however my "challenge" is to try to see how far i can get with AR3, at the moment i am not that impressed. There sems to be a definite issue with TP and shadows, i will try to figure oiut how to use mograph for a fix but what if you don't have mograph
The other issue is the flickering of the ground which ( apart from the gi flicker - which according to claims should not exsist) i think is down to AA- the cure would be very high AA and or blurring of the ground. I am always unhappy with having to blur textures to cure texture crawl- what is the point of having a great texture if we have to blur the life of it ?
paulselhi
09-08-2008, 11:34 PM
sadly mograph clones of the TP do not solve the shadow issue
Per-Anders
09-08-2008, 11:37 PM
Cache the mograph using the mograph cacheing tag, that should solve it. The issue is that TP itself doe not have deterministic randomization, so each time it plays it plays slightly differently, as the caches are built on the scene if that changes between passes and the ifnal render then the cache itself of course is completely out (i.e. it's a TP issue, which hopefully will be resolved in future versions).
paulselhi
09-09-2008, 12:32 AM
ok i will try that but i don't think caching is the problem as the issue still occurs with brute force renders. it may be a bug with r11 as i am testing on 10.5 and it seems fine
and if there is a tp issue to be fixed then i wonder if there would be any chance to get a full upgrade to TP..oh maybe asking to omuch..TP is only..how many years old ?
Per-Anders
09-09-2008, 01:03 AM
You're getting flickering a missing particles when rendering using QMC only (not IR + QMC mind)?
paulselhi
09-09-2008, 12:58 PM
It is not so much missing particles as missing shadows. Go to frame 193 render with gi and note there is no shadow, switch gi off..still no shadow, switch the TP emitter off..still no shadow but then trurn off the particle geometry and the shadow appears
So this is not a gi issue but a bug with, at least this scene. shadows and particle geometry in r11
even if i switch off sky and add an infinite light it is the same- particle geometry is interfering with shadows
even using a mograph cloner object for the particles interfers with the shadows
paulselhi
09-10-2008, 11:19 PM
Here is a render using R10 showing that shadows work fine in R10. The render took 20 hours using stochiastic mode aka brute force ( 50 hours faster then R11 and it's object animation mode and much better)
I think the GI is quite good, still got some texture crawl on the floor and it seems that this is more obvious when you are looking across the floor at eye level, from above the crawl is not so obvious. I will try with higher AA
http://www.black-and-white-to-color.com/stuff/r10sto.mov
CGTalk Moderation
09-10-2008, 11:19 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
vBulletin v3.0.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.