OctaneRender™ Standalone v1.0 beta3.03 [obsolete]

[abstrax]

Im finding this release almost 2x slower than 2.58 on my setup

PT
3.03 = 0.29 Ms/sec
2.58e = 0.56 Ms/Sec

Which speed do you get with 3.01?

[indexofrefraction]

installed 3.03 + cuda driver 4.2.10 --> didnt work
cuda toolkit 4.2.9 --> didnt work

octane says :
cuda driver version 0.00, "cuda driver must be at least of version 4.2... cuda init failed. ... "

this is mac os x 10.7.5

[bepeg4d]

hi,
i have got the same problem
with os x 10.7.5 you must install cuda drivers version 5.0.24 or 5.0.36.
cuda 4.2.10 works on 10.7.4 only and you need nvidia drivers 270.0 also
ciao beppe

[indexofrefraction]

hey thanks that helped

still no reload mesh function in 3.03?
i thought i read smth like that in one of the beta release notes

[funk]

Im finding this release almost 2x slower than 2.58 on my setup

PT
3.03 = 0.29 Ms/sec
2.58e = 0.56 Ms/Sec

Which speed do you get with 3.01?

PT
2.57 = 0.71 Ms/Sec
2.58e = 0.56 Ms/Sec
2.59 = 0.43 Ms/Sec
3.01 = 0.31 Ms/sec
3.03 = 0.29 Ms/sec

It seems since the 3.x release it's about 2x slower on my gtx 285
UPDATE: I added more versions. It seems since 2.57, speeds have been dropping

[abstrax]

Im finding this release almost 2x slower than 2.58 on my setup

PT
3.03 = 0.29 Ms/sec
2.58e = 0.56 Ms/Sec

Which speed do you get with 3.01?

PT
2.57 = 0.71 Ms/Sec
2.58e = 0.56 Ms/Sec
2.59 = 0.43 Ms/Sec
3.01 = 0.31 Ms/sec
3.03 = 0.29 Ms/sec

It seems since the 3.x release it's about 2x slower on my gtx 285
UPDATE: I added more versions. It seems since 2.57, speeds have been dropping

Ok, thanks for the feedback. What's happening here is, that in 2.58 and 2.59 we did some optimizations for Fermi/Kepler, which unfortunately reduced speed on old Tesla GPUs (compute model 1.x). And with beta 3.x we introduced instancing which also doesn't run too well on old Tesla GPUs. Since most of our customers are using Fermi/Kepler (which are many times faster than Tesla GPUs), we are not considering Tesla GPUs anymore when we add new features or work on performance.

Cheers,
Marcus

[abstrax]

hey thanks that helped

still no reload mesh function in 3.03?
i thought i read smth like that in one of the beta release notes

There has been a reload button on the mesh node for ages? Or what exactly do you mean?

Cheers,
Marcus

[funk]

Ok, thanks for the feedback. What's happening here is, that in 2.58 and 2.59 we did some optimizations for Fermi/Kepler, which unfortunately reduced speed on old Tesla GPUs (compute model 1.x). And with beta 3.x we introduced instancing which also doesn't run too well on old Tesla GPUs. Since most of our customers are using Fermi/Kepler (which are many times faster than Tesla GPUs), we are not considering Tesla GPUs anymore when we add new features or work on performance.

Cheers,
Marcus

Yeah I understand Looks like it might be time for an upgrade

[ROUBAL]

Hi Marcus ! Have you seen my little bug report in a post above (page2) ? About node input not working with two links in output ?

[andw]

In fact, the problem occurs each time there is more than one link connected in output of the obj node.

I found similar bug with nodes connections.
How to reproduce it:
- Open sample scene from tutorial "23 - Instancing Placement" (http://www.youtube.com/watch?v=5PsGuKA3dZY). Direct link is http://uploading.com/files/get/2b98ma6b/Tutorial.rar.
- After load, disconnect link from topmost "Placement" node output to some of "Placement" nodes input ("Geometry" pin), placed below, in first row of five nodes.
- Then try to restore this link...it can not be done.