RE: Mental Ray Standalone distributed rendering slaves render really slow!
| Date : Wed, 30 Apr 2008 16:42:03 -0400 |
| To : <XSI(at)Softimage.COM> |
| From : "Halfdan Ingvarsson" <hingvars(at)Softimage.COM> |
| Subject : RE: Mental Ray Standalone distributed rendering slaves render really slow! |
|
No. It only has to be transmitted once. But the data is
requested on demand. Ie. an object's geometry is only requested at the
first tile it shows up in but it stays on the slave until no longer needed (or
if it runs out of memory).
- ½ From: owner-xsi(at)Softimage.COM [mailto:owner-xsi(at)Softimage.COM] On Behalf Of Mathieu Leclaire Sent: 30-Apr-2008 16:28 To: XSI(at)Softimage.COM Subject: RE: Mental Ray Standalone distributed rendering slaves render really slow! Oh but
I?m still curious to know if the data has to be repassed for each tile or if it
stays in memory for the next tiles once it gets it the first
time? Mathieu
Leclaire R&D
Programmer Hybride
Technologies "Yesterday is history,
tomorrow is a mystery, today is a gift and that is why it's called the
present" -----Original
Message----- The
slaves have to get their data from somewhere. If the time of sending the data,
over the network to the slave, exceeds that of the time it takes to render the
whole frame on the master, then it's not worth pursuing. Increasing the tile
size isn't going to do much for you in this case. Distributed
rendering is good for some things (large image size, or small scene + lots of
FG/GI), bad for others (large scene data). -
½ From:
owner-xsi(at)Softimage.COM [mailto:owner-xsi(at)Softimage.COM] On Behalf Of Mathieu Leclaire So what
does that mean? It?s normal? I should use larger tiles to maximize the slaves
resources or once the tesselated scene data is passed, the slave keeps it in
memory for future tiles and it shouldn?t affect the speed
anymore? Mathieu
Leclaire R&D
Programmer Hybride
Technologies "Yesterday is history, tomorrow is a
mystery, today is a gift and that is why it's called the
present" -----Original
Message----- The
cost is mostly in the master sending the tesselated scene data
over. -
½ From:
owner-xsi(at)Softimage.COM [mailto:owner-xsi(at)Softimage.COM] On Behalf Of Mathieu Leclaire Hi
guys, I?ve
configured my Mental Ray Standalone for distributed rendering. When ray3 is
launched, the master can easily render 50 times faster then the slaves can. Is
this normal? Did I miss-configure something in my XSI installation or is there a
problem with my network? It?s faster to render all the tiles directly on my
master even though it is a pretty heavy scene. I don?t get it. Is anyone able to
enlighten me? Thanks. Mathieu
Leclaire R&D Programmer Hybride Technologies "Yesterday is history, tomorrow is a
mystery, today is a gift and that is why it's called the
present" |
- References:
- RE: Mental Ray Standalone distributed rendering slaves render really slow!
- From: "Halfdan Ingvarsson" <hingvars(at)Softimage.COM>
- RE: Mental Ray Standalone distributed rendering slaves render really slow!
- From: "Mathieu Leclaire" <mleclair(at)hybride.com>
- RE: Mental Ray Standalone distributed rendering slaves render really slow!
| DATE: | << | >> | THREAD: | << | >> | INDEX: | Main | Thread |
|---|
- Previous by Date: A reminder to list members under NDA
- Next by Date: A reminder to list members under NDA
- Previous by Thread: A reminder to list members under NDA
- Next by Thread: A reminder to list members under NDA
- Index(es):
| Search the XSI List archives here or use the advanced search form to search across mailing lists. Searching help is available. |