PDA

View Full Version : Speeding up renders



twister47
11-07-2004, 11:13 AM
Hey guys, I just discovered this myself, and I'm sure MOST of you know this already, so I guess this is more directed towards the newbies, like myself.

I totally never thought about the effects of breaking up images into segments, based on the segment memory option in the camera settings.

And this is probably more effective when using the spinning lights trick, as I was doing for this render, since it breaks the frame into so many different segments.

So I noticed the segment memory limit was set at 8 Megs and I realized, okay, I have a gig of RAM, why on earth am I rendering it that low. At the 8 meg limit, a frame at HD was broken into 11 segments and rendered in 5 minutes. I cranked the limit up to 800 megs and it rendered in 1 segment in 70 seconds.

So for all of you who knew about this before, please laugh all you want at me :rolleyes:

And for those who didnt, I hope I can save you a LOT of rendering time.


Nate

toby
11-08-2004, 09:04 PM
You can actually waste memory if you set it higher than it needs to be. Generally, set it just high enough to render in one segment - I find that 48mb is usually enough, unless you're doing large images.

cgbloke2004
11-08-2004, 10:31 PM
This is what I thought for years - i always set it at the point at which my image would render out in 1 segment - PAL D1 was always - 24-32mb (i cant remember at the moment - i'm a bit sleepy)..
And so my memory segments have always been set at that, raising or lowering as required, but always to match a little over the requirement for a one segment render.

But i did keep reading in the manual [ottomh]:
'you can assign as much memory as you want, as Lightwave will only ever use what it needs to'
- and so i guessed this meant that you could set it to 512mb, and even though it took LW, say, 64mb to render a frame out in 1 segment - the rest of the memory space thats been assigned is still not lost to Lightwave (and therefore wasted) and is still accessible as normal by the system..?
Or is that other 448mb lost to LW..?
Or did i miss something..?

twister47
11-09-2004, 09:36 AM
I think you might be right.

Because the first time I did it, I set it to 100 MB, and it rendered in 91 seconds

the second time, the 800 MBs, and it did it in 71 seconds

I mean, I assume it the frame needed more then the 24-48 you guys were talking about becuase I had 3 full sized Snowboard graphics (which are 24 megs themselves)

But still it does say 'limit' So I assume it'll use as much as it can up to that?

toby
11-09-2004, 01:32 PM
Actually what I said may only apply to the old OS9 version -

Marcia
11-14-2004, 05:15 PM
I didn't realize, from what I'd read in the manual, that you could actually crank the segment memory up to 800 mb. I have 2 gig of RAM and am trying to render an image at 3636 x 2727 for print; the file size is approximately 2.5 mb with about 325 mb of texture maps. Bumped the RAM up to 30 mb, which brought the number of segments down from 30-something to 14, but I discovered after about 24 hours of rendering, that the render was going slower and slower (by my calculations, it would have taken about another 2 weeks to finish!).

My question is, what's the rule of thumb for determining how much RAM you can allocate to a rendering before it starts paging the hard drive and really slowing things down?

I'm running OS X 10.3.5, which uses dynamic memory management, so there's no way to allocate a specific amount of RAM to the program. System Profiler no longer tells you how much memory is devoted to the OS (fairly stripped), so I'm not sure how much to subtract from working memory.

model is 2.5 mb; maps are 325 mb... I assume all of this is being held in RAM during rendering (???)

Does anyone know how much of the 2 gig I can reasonably allocate to the memory segments?

(Sorry if this is repetitious, but this rendering is giving me a migraine . Any advice appreciated.