Page 1 of 1
Render speed comparison V2 vs V3
Posted: Sun Dec 08, 2013 11:41 pm
by umdrehung
Hello,
I was wondering, if there is any comparison between render speed of V2 and V3 on the same machine with the same settings?
Kind regards
Umdrehung
Re: Render speed comparison V2 vs V3
Posted: Mon Dec 09, 2013 12:39 am
by Bubbaloo
What scene? Some scenes will render faster, depending on the content.
Re: Render speed comparison V2 vs V3
Posted: Thu Dec 12, 2013 12:09 pm
by umdrehung
Ok, so if some scenes will render faster, there is an average speed gain.
That's exactly what I am interested in. How much ist the 'average' speed gain?
Umdrehung
Re: Render speed comparison V2 vs V3
Posted: Thu Dec 12, 2013 3:41 pm
by Polyxo
umdrehung wrote:Ok, so if some scenes will render faster, there is an average speed gain.
That's exactly what I am interested in. How much ist the 'average' speed gain?
You're asking for completely meaningless numbers here. I'm glad that NL doesn't spam us with statements like
speed increase up to 800%! *
They could have done this, they could even have made up a number which represents the average frequency of Motion Blur usage in Rendering and
have dialed in a number. But was this any meaningful? No.
*only valid on certain files which use Motion Blur
Re: Render speed comparison V2 vs V3
Posted: Fri Dec 13, 2013 10:41 am
by umdrehung
Polyxo,
if there is only a speed gain only for renders with motion blur, then the numbers are indeed meaningless for me too (maybe not for others).
If there is an average speed gain for any renders this may be meaningless for you but that doesn't mean, that it's meaningless for everyone ( I know - it's hard to believe not to be the center of the universe).
Re: Render speed comparison V2 vs V3
Posted: Sat Dec 14, 2013 1:28 pm
by Polyxo
umdrehung wrote: ( I know - it's hard to believe not to be the center of the universe).
It might be more helpful to read release notes or to watch the Novedge Seminar instead of resorting to
undue statements like this one.
Re: Render speed comparison V2 vs V3
Posted: Sat Dec 14, 2013 4:59 pm
by kami
I just rerendered all of our projects of the last year with maxwell v2 and now with v3 on the same machine. The speed average speed gain was 7,51% but if you'd know how my grades in mathematics on my school report were you'd better not trust this number too much!

Re: Render speed comparison V2 vs V3
Posted: Sun Dec 15, 2013 12:36 am
by umdrehung
@Polyxo
Sorry for that.
@kami
Thank you.
Re: Render speed comparison V2 vs V3
Posted: Sun Dec 15, 2013 1:30 am
by vizport
in my tests its around 10% so far.
Re: Render speed comparison V2 vs V3
Posted: Sun Dec 15, 2013 6:33 am
by Mihai
It won't make much sense to look at SL alone. Especially if you had a scene with many additive materials - check the noise levels of the V2 version. Then compare the noise levels with the V3 version.
Re: Render speed comparison V2 vs V3
Posted: Sun Dec 15, 2013 4:36 pm
by kami
Sorry, I forgot the [ironic] bracket ... to be honest, I didn't rerender a few hundred images. But I wanted to try out a few selected projects and compare the speed/noise, which I haven't done yet.
I hope scenes with a lot of hidden objects (
http://maxwellrender.com/forum/viewtopi ... 97&t=41018) will also perform faster now, especially when using instances.
Re: Render speed comparison V2 vs V3
Posted: Sun Dec 15, 2013 6:38 pm
by pat@patrickfaith.com
vizport wrote:in my tests its around 10% so far.
I'm finding about the same number if I keep everything the same.
But when I start changing things, I am finding i can drop my sample level by two with V3 (which is a exponential decrease in time). With the ability to alpha masks, deep exr, and the other features I am doing a lot more in compositing in nuke and not rerunning things in maxwell. So I think overall, I have about 3 times speed increase in my entire workflow which I think that's pretty amazing. (really means maxwell guys were listening to production workflow problems/comments when they did all these changes)
This obviously is dependent on peoples cases: I do a lot of caustics, and is for film (2k dcp), with things moving (i.e. motion blur applied in nuke) with respect to film footage that has grain … so I don't have to be as fanatical on renders as people doing things like fully graphically generated ads.