OctaneRender® beta 2.43 (lin/mac/win) [OBSOLETE]

[matej]

When Octane can't find an image (it was deleted / renamed or else..) and you click cancel, the whole loading process stops. Wouldn't be better if a "skip" method would be added, and the missing texture data would be replaced by some default RGB node.

Also, during loading you are asked multiple times for the same missing image file - for every time that this image file is referenced in nodes. There should be distinction between image datablock and texture datablock. The latter can be derived multiple times from the former (as is done in every other graphic program I know). I've written my opinion of this non-distinction in Octane as a feature request.

[Jaberwocky]

This is more usability problem than anything else.

Whilst doing my Competition Entry i found that everytime i altered a setting in the node inspector , the Tree in the outliner collapses.I have to go and expand the outliner if i need to adjust another item.Can't this be made sticky...IE it remains open rather than keep collapsing?

If it was mentioned elseware before then apologies.

Not a problem with the Software as such but i noticed the competition has ended early.I though it was open till 8.0PM GMT Monday 7th ? That's what it said on the competition front page.?

[Synthercat]

Speed issues...

1.40 MS/Sec Ubuntu 10.10 64-bit
1.47 MS/Sec Win XP 32-bit !!!
(That's Megasamples and not Microsofts per second lol)

Isnt'a weird?
I did this with The famous trench benchmark setting the Kernel to path
I was about to move to 64bits and now I dunno what to do!

I did use CUDA 3.2 and 260.19.06 nVidia
I have an AMD Phenom 9950 working on 196% (why?) and a GTX-470 768MB

Anyone did a WinXP 32 vs Win7 64 speed test?
Anyone tried this baby with different Linux distros? I bet Sabayon & Gentoo will be faster

[GeoPappas]

Anyone did a WinXP 32 vs Win7 64 speed test?

I upgraded from WinXP (32 bit) to Win 7 (64 bit) last summer and ran a few speed tests to compare Octane performance between the two O/Ses. I found that Win 7 (64 bit) was slightly slower than WinXP (32 bit), but nothing to be alarmed about. This makes sense, since WinXP is generally faster than Win Vista/7 for most things.

For testing purposes, I used two of the demo scenes (Chess Set (Image Maps) and Spaceships) and ran them for 500 samples. In both cases, the time to complete the samples was 0.8% to 0.9% slower on Win 7 (64 bit) than on Win XP (32 bit).

Please note that these tests were under 1.022 Beta 2.2 Final, but I am pretty sure that results would be very similar for the most recent version.

[dave62]

Speed issues...

1.40 MS/Sec Ubuntu 10.10 64-bit
1.47 MS/Sec Win XP 32-bit !!!
(That's Megasamples and not Microsofts per second lol)

Isnt'a weird?
I did this with The famous trench benchmark setting the Kernel to path
I was about to move to 64bits and now I dunno what to do!

I did use CUDA 3.2 and 260.19.06 nVidia
I have an AMD Phenom 9950 working on 196% (why?) and a GTX-470 768MB

Anyone did a WinXP 32 vs Win7 64 speed test?
Anyone tried this baby with different Linux distros? I bet Sabayon & Gentoo will be faster

hey,
maybe try the cuda3.0 version for ubuntu.
cant say they if its really faster, but at least much more stable on my linux sys.
i dont think that it will differ much under your mentioned linux distros, but if you made any experience with it let me know (i think about changing to mint or centos from ubuntu)
if there is any testscene available, i will do some benchmarks.
which settings/scene do you used?
cpu usage is the same on my machine no matter the cuda version.. maybe some compiling stuff needed for cuda.

cheers dave

[dave62]

hey marcus,

not shure if there is any bug/todo list (would prevent or at least decrease double posts)?
one thing for linux version: please make the enter button on the num pad work.
it is still not working in 2.43.

btw previous crashes on my machine were cuda3.2 related! since i changed to cuda3.0 it works really rock solid! 36hrs up time and no crash

great work guys, keep on!

dave


(ubuntu10.10 64 bit)

[matej]

Gah!, this is still present..

err.jpg

When I load a saved material macro which has an embedded image, use it in the scene, save the scene, then reload that scene - sometimes it works, sometimes it complains that it didn't succeeded to load the embedded image. Problematic file attached.

Please: make it so that if it cant load or find images, the texture is replaced by some default and the scene continues to load - currently this bug effectively corrupts your whole scene file. Until this is 100% fixed, we cant use our .ocm material database, without the fear that it will corrupt our work.

Or make and option that for .ocm intended for local use, the images are just linked from outside, not embedded in the .ocm (it wastes space, it's not flexible and has problems reloading).

[YoonKyung]

I did run multiple GPU mode, with Octane beta 2.43 | 4 x GTX580 | CUDA 3.2 driver | Windows XP 64 bit,
and got about 60Msps performance. What's up? So I tried with one GTX580 and got about 15Msps.

[abstrax]

I did run multiple GPU mode, with Octane beta 2.43 | 4 x GTX580 | CUDA 3.2 driver | Windows XP 64 bit,
and got about 60Msps performance. What's up? So I tried with one GTX580 and got about 15Msps.

Did you use the CUDA 3.0 build? If yes, then that's expected. Actually it's even better than I would expect. If you used the CUDA 3.2 build then that would be surprising.

Cheers,
Marcus

[pixelrush]

multi GPU is broken with Cuda 3.2
use Cuda 3.0 and the Octane build for that
if all else fails read the instructions...

bah beaten my seconds - I'll post it anyway