bad allocation - crashing during renders - windows vr.

[michael_brazelton]

hey guys - so i've been getting this bad allocation error every since I started using the windows vr. of silhouette - when I used the linux vr. at weta I never got it - talked to jbills about it and he thought it was the ram cache - so Ive been playing with the values there to see if it'll help - and it has'nt really - when I have a long roto with lots of shapes and layers it crashes out on me after about 100 frames -

I have upgraded to 2.2 - and did'nt seem to help - anyone know how I can fix this?

thanks!@!!!:eek:

[michael_brazelton]

ok - just through messing around found that turning off cache output in the renders preference allows me to render through the whole frame range...

[marcop]

The fact that you need to restart Silhouette before the new cache settings take effect probably through you for a loop.

[jbills]

many programs, when sensing these types of changes, pop up a dialog - "you must restart for your settings to take effect."

[orphanage]

I have de-selected cacheOutput, restarted the application as well as the computer, but still get the bad allocation error after about 100 frames or less. I see the cache-o-meter filling up as it goes through each frame. Should this still be happening if I de-select the cacheOutput during a render?

One work around I've found, and this is what I would assume de-selecting cacheOutput does in principal, is to void the link between the footage and just let it export the alpha without pulling in the plate. Of course this will only work if you are exporting luma mattes, not RGBA renders of your source.

Any suggestions on how to avoid the Bad Allocation errors?

Thanks.

[paulm]

How much physical RAM do you have, and what is the maxCache preference set to?

Try setting maxCache to half your physical RAM.

[jbills]

ya, maxRAM is the pref you want - just drop that down. it varies from OS to OS. But you should be able to experiment around and find a magic number that is stable.

I assume you're on windows - you may also look at adding the 3GB switch and see if that improves anything. I believe there were some instructions in the latest release notes (I could be wrong).

the renderer loads the plate, so although it's not caching output it's still bringing the FG into ram. Paul's aware this could be optimized so that it doesn't even pull in the FG when rendering a mask only project to alpha.

[orphanage]

Is there anything in the future to eliminate this problem? I feel a little like I'm driving a race car but am only able to fill the gas tank up 1/2 way. Any way this RAM issue can be addressed? How does it feel to be driving a 4GB car and only able to fill it up to 1.5GB?

Thanks.

[jbills]

are you on windows? if so, did you enable the 3gb switch? it's easy. give it a google.