Octane Render 1.022 beta2.2 RC3 available [obsolete]

[Proupin]

dof is circular, it's a circular aperture, there's nothing wrong with that...
Radiance

yes there is. attached the same render with previous RC2 comparision, note how the higlight is circular, but now in RC3 it's a RING. In the highlights you can see best what's the shape of the Depth of Field.

about the devices, if you have 2 identical devices, we take the 2nd device.
the 2nd device is what the cuda driver gives us as device number 2, and as the cuda docs state, the order of devices returned is not always the same.
i suggest you attach your monitor to the other card, so octane uses the other one by default.
in 2.3 we'll have an app config file so you'll be able to store your setting(s).
Radiance

I don't have identical devices. I use the 220 for display and to prevent slowdowns when rendering with the gtx470... you are suggesting to plug the 470 to the display instead?

[thwak]

Well... with 3 SLI GTX 480 it´s getting the same 1.49 Megasamples with pathtracing...

I didn't think Octane worked with SLI'd cards?

[YoonKyung]

A multi-gpu scalability report

RC3 ///////////////////////////////////////////////

RES : 1024x512
1x GTX480 : 4.25 Msamples/sec
8x GTX480 : 20.97 Msamples/sec (61.67%)

RES : 1920x1080
1x GTX480 : 4.35 Msamples/sec
8x GTX480 : 27.65 Msamples/sec (79.45%)

RES : 3840x1260
1x GTX480 : 4.72 Msamples/sec
8x GTX480 : 32.17Msamples/sec (85.19%)

RES : 4096x2048
1x GTX480 : 4.53 Msamples/sec
8x GTX480 : 31.22 Msamples/sec (86.14%)

RC2 ///////////////////////////////////////////////

RES : 1025x512
1x GTX480 : 4.19 Msamples/sec
8x GTX480 : 20.97 Msamples/sec (62.55%)

RES : 1920x1080
1x GTX480 : 4.32 Msamples/sec
8x GTX480 : 28.25 Msamples/sec (81.74%)

RES : 3840x1260
1x GTX480 : 4.65 Msamples/sec
8x GTX480 : 32.57 Msamples/sec (87.55%)

RES : 4096x2048
1x GTX480 : 4.47 Msamples/sec
8x GTX480 : 31.22 Msamples/sec (87.30%)

////////////////////////////////////////

There seems to be dependency between resolution and scalability.
And, RC2 seems to be faster than RC3, but the differences are within error range.

[r10k]

A repeatable crash...

- Create a material (new scene, no objects)
- Click on it to show the viewport preview
- Delete the material node
- Try and select the material in the viewport using the picker

[YoonKyung]

A multi-gpu scalability regarding scene complexity

RES : 1280x720
SCENE : 1.468M Triangles

1xGTX480 : 1.75 Msamples /sec
8x GTX480 : 8.39 Msamples /sec (59.92%)

//////////////////////////////////////////////////////

The scalability does not depend on scene's complexity.

[scandog72]

Render windows gets stuck

I have been playing with RC3 and have noticed that if I render a picture to the max samples that once finished I can't restart the render or do anything to affect the render windows. I can't select materials or anything.

Is anyone else finding this?

[InfoCentral]

Do I need to delete previous version of Octane before installing this release?

[YoonKyung]

For more accurate performance check, I like to suggest to display average Msamples/sec.
If the value of Msamples/sec would be the average of some period, such as last one second,
we could check it more accurately, I think.

[Elvio]

My scene has a resolution 1920X1080 and it´s getting really slow with pathtracing, the scene has a high complexity, however the spped difference between the direct lighting kernel and pathtracing is drastic!

And another big factor is the little difference in speed that many video cards interconnected are doing when the pathtracing is on, in my scene hardly made any difference to have three video cards in SLI, the speed with pathtracing is almost the same with one or three cards (1.49 / MegaSamples 1.85). I will use only one card, this way im saving energy.

I will wait a scene from radiance to do more tests.

Regards.

Elvio Cavalcante

[YoonKyung]

My scene has a resolution 1920X1080 and it´s getting really slow with pathtracing, the scene has a high complexity, however the spped difference between the direct lighting kernel and pathtracing is drastic!

And another big factor is the little difference in speed that many video cards interconnected are doing when the pathtracing is on, in my scene hardly made any difference to have three video cards in SLI, the speed with pathtracing is almost the same with one or three cards (1.49 / MegaSamples 1.85). I will use only one card, this way im saving energy.

I will wait a scene from radiance to do more tests.

Regards.

Elvio Cavalcante

Disable SLI. The SLI integrates your three GPUs into one GPU.
So there were not the performance increase.