Tweet
so
in this thread:
http://www.chaosgroup.com/forums/vbu...ad.php?t=42881
I reported an error in a way that reflections/shadoing was being calculated.
after some trial/error it was apparent that it was not an issue with vray (thanks Vlado!).
i felt hopeless, as i could't nail the bastard and the animation had to be finished with those artifacts....... i don't have to tell you how frustrating that can be.
finally (but after finishing the job) i found that it happened only when the object was smooth-shaded. flat-shading produced desired results, but at the cost of faceted surfaces and the like.
THEN i stumbled upon ANOTHER VERY ANNOYING problem.
...
nv4_disp.dll BSOD in max/d3d.
it happened before, every time i tried to open heavy scenes with d3d and cached3xmeshes enabled. without it viewports were as sluggish as in opengl - not what i would expect from 8800gtx. for some time i've been working using opengl - if the scene had less than 1M polys it was ok. but recently i was trying to work with a scene that was 13M polys/33M verticles and had about 0.3 fps, and redrawing of the viewports took ages....
reinstalling drivers - nothing
updating dx - nothing
reinstalling max - nothing
i even considered soft-modding my gfx into quadro, but decided to do it later (for fun / experience)
...and then I noticed that my max was not updated with sp2.
installing the patch didn't help with the BSOD thing, but it SOLVED the previous problem - no more different buckets in DR!!
(i didn't think that sp2 had such changes that it affected the way that objects are being shaded in rendering - my workstation produced darker buckets with visily different shading than patched to sp2 render nodes.)
so it was 1:0 for me, and only one to go...
after 3 days of web browsing and trying to come up with something on my own i found info about /3GB switch causing some problems with drivers.
of course i was already aware of that, but as always couldn't believe that it would have any negative effect on MY hardware
what i found was this:
http://thatworks.blogspot.com/2007/07/3gb-switch.html
in short - using only /3GB is not recommended, BUT adding /userva='something below 3000' (like /userva=2900) increases stability, at the cost of a little over than 100 megs of memory.
i tried this
and it WORKS!
no more nv4_disp.dll BSOD in max, 13Mpoly scene now at >20fps, viewports redrawing almost instantly, work is now much more comfortable and stress-free.
this tweak was posted earlier, but it was linked to video card disappearing in the system, not BSODing all the time...
conclusion:
shading issues as observed in my previous thread - conflict between max versions on workstation and render slaves. my mistake, as i should have checked this before bothering You with the problem. but maybe it will help someone with similar issue, who - as I - forgot to update max on all machines and has version conflicts
nv4_disp.dll BSOD in d3d w/cached3dxeshes on - /3GB issue, solved with adding /userva='between 2900 - 3030'. i hope that this solution is solid, and the driver won't go haywire again....
cheers and thanks!
Maciej
in this thread:
http://www.chaosgroup.com/forums/vbu...ad.php?t=42881
I reported an error in a way that reflections/shadoing was being calculated.
after some trial/error it was apparent that it was not an issue with vray (thanks Vlado!).
i felt hopeless, as i could't nail the bastard and the animation had to be finished with those artifacts....... i don't have to tell you how frustrating that can be.
finally (but after finishing the job) i found that it happened only when the object was smooth-shaded. flat-shading produced desired results, but at the cost of faceted surfaces and the like.
THEN i stumbled upon ANOTHER VERY ANNOYING problem.
...
nv4_disp.dll BSOD in max/d3d.
it happened before, every time i tried to open heavy scenes with d3d and cached3xmeshes enabled. without it viewports were as sluggish as in opengl - not what i would expect from 8800gtx. for some time i've been working using opengl - if the scene had less than 1M polys it was ok. but recently i was trying to work with a scene that was 13M polys/33M verticles and had about 0.3 fps, and redrawing of the viewports took ages....
reinstalling drivers - nothing
updating dx - nothing
reinstalling max - nothing
i even considered soft-modding my gfx into quadro, but decided to do it later (for fun / experience)
...and then I noticed that my max was not updated with sp2.
installing the patch didn't help with the BSOD thing, but it SOLVED the previous problem - no more different buckets in DR!!
(i didn't think that sp2 had such changes that it affected the way that objects are being shaded in rendering - my workstation produced darker buckets with visily different shading than patched to sp2 render nodes.)
so it was 1:0 for me, and only one to go...
after 3 days of web browsing and trying to come up with something on my own i found info about /3GB switch causing some problems with drivers.
of course i was already aware of that, but as always couldn't believe that it would have any negative effect on MY hardware
what i found was this:
http://thatworks.blogspot.com/2007/07/3gb-switch.html
in short - using only /3GB is not recommended, BUT adding /userva='something below 3000' (like /userva=2900) increases stability, at the cost of a little over than 100 megs of memory.
i tried this
and it WORKS!
no more nv4_disp.dll BSOD in max, 13Mpoly scene now at >20fps, viewports redrawing almost instantly, work is now much more comfortable and stress-free.
this tweak was posted earlier, but it was linked to video card disappearing in the system, not BSODing all the time...
conclusion:
shading issues as observed in my previous thread - conflict between max versions on workstation and render slaves. my mistake, as i should have checked this before bothering You with the problem. but maybe it will help someone with similar issue, who - as I - forgot to update max on all machines and has version conflicts
nv4_disp.dll BSOD in d3d w/cached3dxeshes on - /3GB issue, solved with adding /userva='between 2900 - 3030'. i hope that this solution is solid, and the driver won't go haywire again....
cheers and thanks!
Maciej
Comment