Need help with large tour

[Dman3d]

Hello all,

I need help with a current problem.
We are working on a rather large tour but not much bigger than anything
else we have ever done.

At this point I can only describe it to you
Single condo/structure with water, boats, land, roads, trees and lowpoly
supporting objects such as stand in buildings and the like.

All of the trees are proxy and even some of the floors, the ones that are
identical are proxy and instance.

At one point I was able to render this with somewhat low settings but
something has happened that I cannot get it to render on all machines
consistently. Let me explain.

We have a mixed bag farm of around 60 machines total, I am working
with around a dozen of them. Some are workstations that are slaving at
night.

I am getting unexpected exceptions and on occasion other memory related
errors. I have cut down just about everything in this file. No textures, with
textures, with the main building, without and I keep getting mixed results.

For more wierdness. There are 4 machines in particular that are rendering
but slowing as they are single processor machines with 1GB of RAM, the
problem is the big dual machines with 2GB do not, they all throw
the unknown error.

Now here is more. The proper "slaves" are all new clean installs of Max8
with sp3 on XP. The older user machines that are slaved at night and rendering
are mixed bag, Max8sp1, sp2 and sp3 they all have 1.47.037 installed but
I cant imagine why this would make a difference but they have Max5,
Max7 and 7.5 update on them as they are old and have not been
burned down. They also are not using bitmap pager, neither are the others
it still throws errors whether bp is on or off at any value (it appears)
They are also mixed OS Win2K and WinXPSp2. I cant seem to isolate
what is or is not the same easily.

Anyone have any ideas on a good way to start trouble shooting again, I
am at a loss at this point and this project is due in 4 days with 2500 to be
rendered.

More to this, each of our departments are gettings some unexpected errors
working with files we have had no problems with in the past. Some images
are causing "cannot load filter tables" alot of these have been in use for ever
and the files themselves used for some time.

There are approx 502000 poly count, with proxies I have no idea a zillion
I suppose with all the trees and bushes. File is around 40MB.
Not sure what the total load is for textures, anyone know a quick-ish way to figure that out?

While writing this post I was trying different things, I just tried rendering
without the VFB, either one and it rendered on more machines but two
the "big" machines threw unexpected exceptions.

I also tried recalc on the LC using world and it crashed about 75% in.

------
Ok found some animated water, animated noise maps tossed that and things sped up

-----
Now find one of the "big" machines running four copies of raysat_3dsmax8.exe isnt that for mental ray? Two loaded but when I slave it 4 show up
-----
Well still no good on the "big" machines. I can only assume something
is screwy with their RAM or network rights or something else I know
nothing about at this point. Will check with IT monday see if we can
check the config on these.

Other problem is some of the machines that are rendering are having
the unexpected exceptions and cant load filter tables errors but not
consistently. The camera move in this is at a medium to slow pace so
there are no dramatic changes.

I am going back to the manual and going through the trouble shooting again.

------

[Dman3d]

Well another update

It is rendering quick
IRMAP -4-2
45/30
.3 5/.2

LC
screen (crashes in world mode)
3000 samples

Adaptive sub
0/4

but!
I found one of our team was, lol, doing mental ray distributed tests on the machine while they were rendering. But once this was sorted they still
have typical mem errors while the smaller boxes crank away.
It is narrowed down the our Dell power edge 1850's now
with occasional failures on the other boxes.

So still going

But I suppose the question now is less about obviously being out of memory but why the smaller machines with less resources are rendering and the bigger ones are not.

[jacksc02]

dman,
try to keep all of your render slaves on the same version of max and vray

[gilicom]

i had all sorts of problems that soud just like yours , I've made the 3gb switch on all of the xp machines and I must say that things got way better, much less crashes .
do you have the 3gb switch on those slaves?
Gili

[Dman3d]

All of the slaves, proper slaves are same OS and same MAX/patches
its just the floor/user boxes that are mixed. I assume that these are
units IT has not gotten to. We are currently in a phase of standarization.
But outside that, the mystery of why some work and some dont. What
is worrisome about this is the new "big" boxes are the standard but
the older or smaller sloppy ones are working fine.

I think we tried it 3GB, and forgive me for being uncertain, in the mosh of
the past week we have had a couple people probing around in the slaves
I think we did one and it worked but need to verify.

It is so odd to me about the smaller machines.
I am going on the notion at the moment that having less physical memory
causes windows to handle it all differently, that is the only reason
I can guess that the smaller ones render and the big ones dont.

I am going to attempt a more thorough grid of data on all the machines
today if possible to get a better picture of what is and is not working.

I am trying to find a way to light the scene differently too. Right now
just have a direct with Vray shadow on the scene, had to shrink it
then the hotspot to focus on the main area only toreduce render time.
Anyone got a tip?

On a better note the job rendered over the past two days without the help
of the "big" machines but I still need to figure this out. I have another
job due to go in the queue right after.

[gilicom]

try this line for 3gb switch on the slaves , worked on all of my servers and they are very mixed.

"multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Micr osoft Windows XP Professionaluserva"/3gb /USERVA=2800 /fastdetect /NoExecute=OptIn "

[Dman3d]

Will do tomorrow. Have the /3GB on one of the suspect machines now waiting for result

[Dman3d]

Nope just hangs forever, no errors though I guess thats something.
(I keep coming back to the office in the middle of the night, sick eh?)

Im going to put Task Info on there tomorrow and closely watch everything
see if there is a critical mass, leak or what on those particular boxes.

[Dman3d]

well now, here is the latest, as we know memory for sure
I have culled down a couple rogue materials that I had assumed were
ok, that taught me a lesson.

But over that, almost all of the machines are rendering now except for,
you guessed it the "big" dell power edge 1850's so tomorrow I will
have their guts all over the floor.

Thanks everyone for the pointers, sorry for the obvious panic but
I suppose we've all been there.

If anything worth reporting pops up i will update this post.

[gilicom]

" have culled down a couple rogue materials that I had assumed were
ok, that taught me a lesson. "
what exectly did you do?

[Dman3d]

Found 2 materials that were over 4k pixels each

[Dman3d]

Ha! got it. There were bad 3dmax.ini files. Previous IT had put
them on all the machines by copying one he set up, by removing
them the "big" dells run as expected.

One of the guys mentioned it in regards to a nother workstation issue
tried it and it worked.

------
Whoop spoke too soon only one of them is working, gotta do the comparison tomorrow.

[Dman3d]

Have one more update. They failed the last time and I found the Mental
ray raystat running a render again, killed it and its fine. Added 3GB
switch and not a problem now, along with no mental ray renders at the
same time.

Anyone know if mental ray uses ram only when doing DR?
We have a disagreement in office right now as to wether this is the
cause of the problem or not. One side yes due to resources being taken
and no saying the resources are so low its negligible. Would that not
be scene dependent?

[jacksc02]

good to hear you have it sorted