render nightmare help badly needed.

[ThE_JacO]

it would be helpful to know WHAT you're trying to render.

does it crash even if you're rendering a sphere?

[Sil3]

Try disabling Norton

[desiredfx]

No it lets me render spheres and anything like that

It is basically as soon as it gets a big file or a large amount of stuff to calculate. I just tried a volumic only pass on my scene and that made it crash out...there are only two volumic lights.... didn't even manage frame1!
but yet sphere with volumic light renders fine. so i'm guessing it is a file size issue..... but i thought it would just take longer or at least i could break it down into passes.
thanks for ongoing help

[KurtLawson]

try turning the progress reports on in your logged messages in render settings. then open your script editor so you can see it. try to render and see if it shows any errors before it crashes out. that would be a good start.

also trying hiding everything in your scene and pulling up one object and with one light at a time and see what object it crashes on. if this works you can replace the object and make a new scene.

godd luck

also check your services to see if mental ray is running. i but that wouldnt make since if you can render sometimes. its voodoo you appear to have a curse of some kind.

[Atyss]

Another thing you may want to try out is to render with XSIBATCH instead of XSI. It seems you are having some memory issues, xsibatch handles more memory than xsi. Btw, would you mind posting the settings you are using in the Optimization tab of the render options?


Cheers
Bernard

[runejw]

Just for fun try disabling hyperthreading in the PC's BIOS - if Windows (XP/2K/2K3) then usually the OS likes fewer (logical) CPU's or it could get stressed too much.

Another thought - size of swap file(s) (pagefile.sys) should be bigger than the RAM installed and preferrably not on the same drive as the OS to get max performance.

Otherwise: need more details. Error messages - especially error numbers are important. Event logs? etc


Cheers,
Rune

[jcorkery]

Just for fun try disabling hyperthreading in the PC's BIOS - if Windows (XP/2K/2K3) then usually the OS likes fewer (logical) CPU's or it could get stressed too much.

Interesting. Yesterday, I was trying to render a couple of HDTV-resolution images from some fairly (geometrically) simple test scenes of metallic and glass surfaces just to see what I could expect when I get around to much more complex scenes and print-resolution renders, but XSI crashed on me repeatedly. After I disabled hyperthreading, the scenes rendered fine.

I can't say I'm thrilled about turning off hyperthreading, since I paid extra for it. Does anyone have any suggestions on how to get XSI and hyperthreading to coexist peacefully?

[runejw]

Well, hyperthreading has some pros and some cons.

Basically hyperthreading is not magic, so you will not get double the amount of transistors - it's just a clever way of dividing the CPU logically so that it looks to the operating system as 2 x CPU's.

The pro about that is that one logical CPU can be dedicated to rendering at 100% while the other devotes more to other tasks. So basically it lets the CPU not completely "lock up" doing just one thing. Great, so now you can write emails or similar and do something useful while at the same time rendering a lengthy animation.

The bad about multiple logical cpu's is that
a) programmers tend to think linearly as opposed to parallell tasking, meaning the programmers are not that great at making programs which keeps both logical CPU's optimally busy. That is why most programs scale OK up to around 4 cpu's but adding additional cpu's give only marginally better performance for a given application. This will improve as compilers and programmers get better at writing paralell-processing programs...
b) if situations arise where both logical cpu's are very busy, then a disproportionate amount of time is used to control the switching between logical cpu's - so it may actually be counterproductive...

Both a) and b) may appear as problems in the OS or application itself, which is probably what you are seeing. Those things are often cured by "Service Packs" or "Hotfixes" or sofware updates (apps).

Invariably there may also be timing issues (hardware/software interaction) and hidden problems in drivers which are exposed by the combination of high cpu-load and cpu-switching. These problems are hard to catch during formal testing but may arise in high-load situations (games, rendering, Terminal Services with high user load etc.) Patches/hotfixes are often made - but only after the bug has been identified and been around for a while...

Regarding operating system support for hyperthreading - both Windows 2003 and XP Pro has full support for it, while Windows 2000 does not fully utilize it (= disabling it is recommended).

Bottom line though is that it could be the application, the operating system, the BIOS itself, some driver (disk, network, anti-virus, video, or other kernel driver) or a combination that caused the problem.


Cheers,
Rune

[jcorkery]

Thanks for that explanation.

I knew that hyperthreading wasn't going to magically multiply my processing power; I basically understood it to be a means of optimizing the processing power of my single CPU. What I hadn't considered before was that any benefit to be gained is dependent on a combination of so many factors. Seems like I should've just opted for a non-HT processor...

[runejw]

Naw, I think Intel did allright with HT. It comes for free in all P4's and it's a cost effective way to introduce paralell-processing.

Also they did good - and something that most programmers forget - to still give the option of disabling a new feature.

Next generation willl be the "dual-head" cpu's where they have actually fit two physical cpu's into one packaging - also with the minimum cost overhead.

Cheers,
Rune

[desiredfx]

still no joy.
Tried everything suggested but still no joy. I think it is a memory issue as it only happen on more complicated scenes. As soon as the size of the scene drops everything is fine. A friend of mine has just started rendering his final project and is having the same nightmare.
IS there a memory cut off point or something. I would have thought that a complex scene would just take longer to render not just bottle out.
I have 2 gig of ocz 400 so that should be plenty. Are there any settings in Xsi or windows that should be changed.
Could really do with help on this one crunch time is approaching!
cheers

[withanar]

Quick question: How many polygons are in your scene?

Are we talking 500K or 5 million?

The default BSP tree settings and memory settings for Mental Ray are fine for small to medium sized scenes. In order to render bigger scenes you will need to do some tuning.



still no joy.
Tried everything suggested but still no joy. I think it is a memory issue as it only happen on more complicated scenes. As soon as the size of the scene drops everything is fine. A friend of mine has just started rendering his final project and is having the same nightmare.
IS there a memory cut off point or something. I would have thought that a complex scene would just take longer to render not just bottle out.
I have 2 gig of ocz 400 so that should be plenty. Are there any settings in Xsi or windows that should be changed.
Could really do with help on this one crunch time is approaching!
cheers

[desiredfx]

there are just over 4 million polys in the scene. In this scene there are 8 fairly hight poly characters sitting in a church. If they are hidden the church renders fine and relatively fast even with fg on but as soon as they are unhidden. It just thinks for a second or two and then crashes. I have watched my system resources and nothing peeks out of control. So I thought it must be something to do with settings either in Xsi or windows. This problem is really limiting what i will be able to hand in for my project...gggrrrrrrr.
Could you explain what was said about adjusting mental ray settings as I have o idea how to go about it.
Thank you

[phonicsmonkey]

How about working around the problem? maybe render the background first and on a seperate pass render the characters...then comp the 2 together. I dunno if this will work for you, but it was just a thought :shrug:

[desiredfx]

yeah multi pass is my only option but it think the scene is about 15 passes now but hey at least its rendering...I know 4 million is a hell of a lot but there must be a way around this problem. Would like to know for future. Any ideas? I know a few guys who are having the same problem so we can't be the only ones
cheers for all help given

[ThE_JacO]

-warning- slight semplification ahead:

what you are hitting here is most likely a problem with discretizing the scene.
BSP is the system MRay uses to separate and order the scene elements and then sort them for rendering.

while BSP is incredibly efficient in most situation, there's a particular case where it will fall down in a blaze, which is the ball and stadium situation (an extremely detailed ball in the middle of a lowpoly stadium).

in situation such as this the process of ordering the triangles goes tits up very easily, because for every branch MRay will try to split and allocate in leaves massive amounts of data, all this because it's been thrown off balance from the weirdly distributed density of your scene.
this is also the reason why you don't see a CPU burnout neither a memory hog, it doesn't get to that point.

optimizing the BSP tree settings is most likely your only bet to get out of this without splitting the characters and the BG in a different pass.

I have to say though that I don't get why you would need such massively detailed characters in the middle of the scene.

[JDex]

On this high poly scene... is there a lot of displacement maps? It sounds like you may be chewing up all your memory w/ displacement data. If you are using displacement try using the 4x4 ender tile size in the render options, optimization tab... it may render slower, but will likely render...

[Atyss]

Keep in mind that BSP is relevant only as long as you have raytracing, that is, raytraced reflection, refracion and shadows. If you don't have any of these, you'll have to experiment with alternatives, like:

- enable Automatic Map Conversion or convert manually your texture to pyramidal map files
- using rapid motion scanline
- user smaller task size
- split high polygon count objects into smaller pieces and render them separately
- render with mr standalone if you can
- render in XSI linux if you can



Cheers
Bernard

[zukezuko]

if you render something with autosave enabled xsi will behave weird
and maybe crash:twisted: during rendering.

solution : turn off autosave before rendering

propably not your case
but does this happen to anyone else?

[CGTalk Moderation]

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.