OctaneRender® 1.024 beta 2.48b TEST (lin/win) [OBSOLETE]

[kivig]

Regarding the black screen when loading a scene with linear tonemapping: Thanks for the report and it will be fixed with the next release.

Cheers,
Marcus

I understand it's not a bright idea to base any work on non-stable version, but I suddenly have a need for dispersion, so I'm interested to know if there may be any prediction on next version release timing. Thank you.

[graffer]

Hi!

There was a problem with displaying the device in the new release 2.48b 64bit win. Downloaded this one file - Windows 64bit: http://www.refractivesoftware.com/rcdow ... in_x64.zip. I can not test the new release.
Drivers 275.33-desktop-win7-winvista-64bit-international-whql. What to do?
Thanks in advance!

Previous release 2.47 PMC stable.

Shortly before I post screenshots, but no one responded to my question.
I repeat: in the new release 2.48b no active device in my case the GTX480. In the bookmarks I can not remove it from inactive to active move. In release 2.47 PMC everything is working, what could be wrong?

[pixelrush]

I think there is a bug already reported concerning the 470 not being selectable. a silly coders error apparently. probably this is related..
Wait and see how it is for test 'C' due soon

[mirobimbo]

I noticed, Octane never crash to me, but 2.48 crash when I drag and drop material from live library to my scene.

Very thanks for alphashadows for directlight.

[kiwimage]

Hi Abstrax and the developpers team,

I am working on a big scene (professionaly) on Octane 2.46 and i'm reaching the maximum number of textures allowed. I wasn't expecting this, especially so close of my deadline...

Anyway, the error message is : "Maximum number of 8bpp Y textures reached, please reduce scene complexity"

I read somewhere in this forum that it is due to the limitations of Cuda 3.

Is there any chance that Cuda 4 (and by extension Octane 2.48) can manage more textures than Cuda 3 did ?
In this case I would switch to 2.48.

Second question :
What are these limitations exactly ? because if I can add quite a few more 16bpp and 32 bpc textures, maybe I could reach the end of my project on 2.46 wich I guess is more stable.

thanks for your time

[kubo]

kiwi, I'm not sure about cuda 4, in anycase, you should try to pack them, bake several of your textures in a single one to fit more, say you have trees with 4 different leaves, pack those leaves in a single texture and adjust your uvws

[Timmaigh]

I tried yesterday 2.48b finally and it seems PMC is slower than on 2.47. I tried to render the very same interior picture, i rendered on 2.47 before, i just changed the material for carpet from Diffuse to Glossy (so its possible this was the reason for the slowdown). On 2.47 it took 30 min to get cca 3200 samples (and the sample rate was 4,55 Ms per second). On 2.48 it managed only 2100 samples in the same timeframe with a sample rate of 2,88 Msps.
On the positive side, the mesh relink button works fine as already reported by ROUBAL.

Can you let it render for 30min and then compare the noise? Although the samples/second is lower in beta 2.48b than in beta 2.47 it should have less noise. Could you please verify this?

Thanks,
Marcus

I am going to rerender the same pic one more time and will return the carpet material back to diffuse, i will pay closer attention to noise regardless of sample amount this time. Are you suggesting that with 2.48b the picture could be less noisy, even if its rendered with less samples? That is possible, i let the pic render up to 5000 Msps though, so i cant compare the final resulting pics, as its 5000 to 3200 samples.
Anyway, it is still brilliant work on your part, it took 1 hour and 5 minutes to get to those 5000 samples and although it is still slightly noisy (at 1920*1200), it can be considered as final image, good for facebook gallery presentation That was nowhere near possible with the PT, so keep up the good work!

[kiwimage]

@Kubo, yeah, I though about this too, it could be a solution, but it 'll be my last resort.
It can be very time consuming to do that, and not optimised in term of re-usability of the models.

I've gathered my textures to count them and I'm a bit surprised : I have only 41 JPGs and 15 PNGs in my scene + a few more in material macros from the live DB.

[Jaberwocky]

@Kubo, yeah, I though about this too, it could be a solution, but it 'll be my last resort.
It can be very time consuming to do that, and not optimised in term of re-usability of the models.

I've gathered my textures to count them and I'm a bit surprised : I have only 41 JPGs and 15 PNGs in my scene + a few more in material macros from the live DB.

I recollect somewhere else in the forum that Radiance said that the maximum texture count in Cuda is 64 different Textures? So if this is still the case then Yes , it looks like your near or over the limit.

[kiwimage]

kiwimage wrote:
@Kubo, yeah, I though about this too, it could be a solution, but it 'll be my last resort.
It can be very time consuming to do that, and not optimised in term of re-usability of the models.

I've gathered my textures to count them and I'm a bit surprised : I have only 41 JPGs and 15 PNGs in my scene + a few more in material macros from the live DB.

Jaberwocky wrote :
I recollect somewhere else in the forum that Radiance said that the maximum texture count in Cuda is 64 different Textures? So if this is still the case then Yes , it looks like your near or over the limit.

Thanks, you're right Jaberwocky, it was there : http://www.refractivesoftware.com/forum ... r+textures :
But now this limitation compromises a major project for me, on wich I've been working for months and wich is 80% finished. I may have to restart over...or not. So I need facts, I need more than "if it's still the case...". I need to know for sure if whether or not I have to completely change my workflow, if i have to switch to another rendering solution OR simply just change my drivers and Octane version, since they have upgraded to CUDA.4 (wich may imply improvements in terms of performance). That's why I ask Octane developpers, they must know for sure the capabilities and limitations of their software and of the CUDA technology.