View Full Version : BSP diagnosis
Hi there;
Iīm trying to understand how the MR diagnosis mode works...
I am diagnosing the BSP size for example..., I know I have to look at the colors of the resulting image but I donīt know if I have to look to its rendertime also..., I mean, what if I have an image with lots of red (wich itīs supossed to be good) and the rendertime is good, and another with less red and a rendertime that is twice the last rendertime?.
Thank you for your guidelines guys.
-joie
|
|
razorback
07-10-2007, 06:45 AM
Diagnostic speed doesnt equate to rendering speed :) It just checks to see if you've blown the limit on something, or how often you're doing it.
Red means its recursed through your max raytrace / FG / BSP levels or whatever it is that you were diagnosing without reaching the end. The cooler the color the less it has recursed to reach the desired end point.
EG : 3 mirrors facing each other. raytrace set to 2 will return red after 2 bounces. raytrace set to 4 would return a yellowish color after 3 bounces, but would have taken longer.
JacobCarmona
07-10-2007, 08:42 AM
Hi Joei.
If you want to understand BSP diagnosis perfectly, I recommend to you the Gnomon mentalray dvd's. It has some videos dedicated to that subject.
Good luck.
OK.
So, to simplify things, the more red I have on my render, the better, do you mean that?.
It doesnīt matter if my render takes longer to render, if it is reddish, itīs OK.
I think Iīm going to check those Gnomon DVDs..., they are expensive, but they may be suitable for my situation...
Thankīs both of you.
razorback
07-10-2007, 02:23 PM
No... the LESS red the better. Seeing red means you aren't getting an accurate value returned.
sixbysixx
07-10-2007, 04:34 PM
Are you sure about this?
I thought the BSP tree is just about how MR tesselates for rendering and it's just a matter of finding a good compromise of leave size vs. tree depth. It will always render correctly, but maybe slower if you set it up the wrong way...
At least that's how I understand it - please somebody correct me if I'm wrong!
razorback
07-10-2007, 04:54 PM
Ok I think I got ppl (and myself included) confused with the reflection example.
The docs say that diagnostics of the photon map can tell you if you're using too little or too may photons hence the inaccuracy of the output. But when it comes to BSP it doesnt really say, but im assuming that if your getting alot of red, alot of info is being written to disc, hence the slow down. I think quality is not effected.
tfritzsche
07-10-2007, 05:41 PM
another good source for deciphering BSP settings is lamrug.org, and it's free
good luck
thomas
According to Digital tutors CD about mental ray tunning, they say the more red in your BSP diagnosis, the smaller pieces of information are processed, so better for CPU computation..., but they donīt put much interest on those matters, so Iīm a bit confused right now...
I may check lamrug for inspiration! :D
sixbysixx
07-10-2007, 06:32 PM
http://lamrug.org/resources/charoptions.html#bsp
I found this a lot more useful and effective than using the diagnose samples!
SanjayChand
07-10-2007, 09:55 PM
you want a good balance of red and blue. Alot of red is bad.
thematt
07-11-2007, 09:15 AM
yes lots of red is really bad..you want the minimum red, only maybe on some part that need a lot of tesselation to render, small wrinkle ect..
cheers
Lots of red is bad?, so is the Digital Tutors DVD wrong?, I canīt believe it...
Iīll perform more test so I can clear things up.
thankīs
http://lamrug.org/resources/charoptions.html#bsp
I found this a lot more useful and effective than using the diagnose samples!
I think it's most effective if you just copy & paste the average size and depth from the output window from your last test render.
sixbysixx
07-11-2007, 06:13 PM
I think it's most effective if you just copy & paste the average size and depth from the output window from your last test render.
No, not really.
Your average depth will always be lower than the max, so if your max is set too low it won't make any sense whatsoever to copy and paste your average settings, right?
OK, my point was that pasting the settings from the output settings takes just a few seconds and gives you a pretty good approach to the "best" setting. Diagnosing BSP more precisley can give you better results by taking time you could already be rendering your final image.
coccosoids
07-11-2007, 08:30 PM
One question: would it be wise that for determining the best bsp settings to render the whole frame - or could you get away by rendering just a "representative" render region?
Thanks.
:)
thematt
07-11-2007, 09:14 PM
Lots of red is bad?, so is the Digital Tutors DVD wrong?, I canīt believe it...
lol do they really say that in the DVD hard to beleive.
But yes if your all image is red it can multiply your rendering time by many, I've experience it before so I'm pretty sure about it, although i'm by no mean a mental ray guru.
cheers
Ironhalo
07-11-2007, 09:30 PM
we're lucky enough to have a platnum license here at work, the white pages entry on diagnostics basically says red = bad.
from the white pages:
mental ray for Maya - BSP Depth Diagnostic.
BSP Depth set to Default of 40.
The color (Blue > Green > Yellow > Red) range indicates areas of geometry that have been broken down into voxels based on geometry detail. Blue areas indicate low depth values in the BSP. Colors change as Depth values increase. Red areas indicate the Depth default of 40.
Render Time (for relative comparison only) = 0:05:10.02
------
BSP Depth set to 50 Render Time (for relative comparison only) = 0:04:15.39
Notice the absence of red areas as compared to the first image. In this particular example, tuning the BSP Depth attribute reduced render time considerably.
Emil3d
07-12-2007, 07:22 AM
... red = bad...Hmm I thought so too but I made a test that doesnt agree with this.
I made a simple scene with several spheres on a plane, with one of them smoothed with a lot of subdivision levels. I made 9 diagnostic rendering tests starting with the default BSP values and after each rendering test, the average depth and leaf size from the Output window were used for the next test. Rendering test 8 was the fastest with a depth size twice as small as the first test with the default values. Its size diagnoses rendering produced evenly distributed blue over the entire scene. Its depth diagnoses rendering showed much redder on the highly subdivided sphere than the first rendering test, yet it was faster.
RENDER TEST 1_________________________________________
RCI 0.3 info : average depth : 34 .................. was default 40
RCI 0.3 info : average leaf size : 7 ................. was default 10
RC 0.3 info : wallclock 0:00:04.88 for rendering
RENDER TEST 2_________________________________________
RCI 0.3 info : average depth : 32 .................. was entered 34
RCI 0.3 info : average leaf size : 8 ..................was entered 7
RC 0.3 info : wallclock 0:00:04.48 for rendering
RENDER TEST 3_________________________________________
RCI 0.3 info : average depth : 30 .................. was entered 32
RCI 0.3 info : average leaf size : 13 ................was entered 8
RC 0.3 info : wallclock 0:00:04.26 for rendering
RENDER TEST 4_________________________________________
RCI 0.3 info : average depth : 30 .................. was entered 30
RCI 0.3 info : average leaf size : 14 ................was entered 13
RC 0.3 info : wallclock 0:00:04.26 for rendering
RENDER TEST 5_________________________________________
RCI 0.3 info : average depth : 28 .................. was entered 30
RCI 0.3 info : average leaf size : 26 ................was entered 14
RC 0.3 info : wallclock 0:00:03.95 for rendering
RENDER TEST 6_________________________________________
RCI 0.3 info : average depth : 25 .................. was entered 28
RCI 0.3 info : average leaf size : 47 ................was entered 26
RC 0.3 info : wallclock 0:00:03.74 for rendering
RENDER TEST 7_________________________________________
RCI 0.3 info : average depth : 21 .................. was entered 25
RCI 0.3 info : average leaf size : 132 .................. was entered 47
RC 0.3 info : wallclock 0:00:03.52 for rendering
RENDER TEST 8_________________________________________
RCI 0.3 info : average depth : 16 .................. was entered 21
RCI 0.3 info : average leaf size : 652 .............. was entered 132
RC 0.3 info : wallclock 0:00:03.49 for rendering
RENDER TEST 9_________________________________________
RCI 0.3 info : average depth : 9 .................. was entered 16
RCI 0.3 info : average leaf size : 3984 ...........was entered 652
RC 0.3 info : wallclock 0:00:05.56 for rendering
JacobCarmona
07-12-2007, 11:01 AM
In the gnomon dvd's also they comment that the red points are bad. Possibly in dvd of digitalis tutors committed an error.
CGTalk Moderation
07-12-2007, 11:01 AM
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.