OctaneRender™ Standalone 2.26

[abstrax]

Hi all,

This is probably the last maintenance update of v2 (unless we broke something important with this release). It adds support for Pascal GPUs and provides various bug fixes that have accumulated over the last months.


Changes since version 2.25:

• Added support for Pascal and switched from CUDA 6.5 to CUDA 7.5/8.0.
• Added confirmation to menu item File -> Save as default....
• Extended the value range of the max. samples pins of the various kernel nodes to 1 .. 1,000,000 and limited the slider range to 1..100,000.
• Fixed rendering of the object layer ID.
• Fixed orientation of the tangent vector when the UV coordinate system is left-handed (turning from U to V is clockwise).
• Fixed incorrect initialization of the shutter time in the render engine after loading a project.
• Fixed missing geometry update if the shutter time changed and the geometry has vertex speeds, but not animated vertices positions.
• Made importance sampling of environments numerically more stable and work correctly if it's completely black.
• Disabled sampling of disabled portals.
• The render engine will be updated after finishing a Lua script that was rendering a temporary render target.
• Improved geometry compilation time when a scene loaded from Alembic is re-compiled, due to a time change, but not all meshes are animated.
• Fixed a bug where partial geometry updates on the slaves were missing to update the extra primitive data, causing the slaves to crash during rendering.
• Fixed random crashes on Linux, caused by OpenSSL when the DNS lookup timed out.
• Fixed empty drop down when you pick a material from a mesh with no material names specified.
• Made sure that the rendered image is displayed sharp at 100% zoom level.
• Fixed incorrect handling of file names that start with "..".
• Sped up deletion of multiple items from the node graph editor.
• The text display of the scatter node will be hidden if the scatter node stores a large number of transformations.
• Eliminated pop-up dialog when an empty script graph is being copied.
• Lua: Added octane.gui.createSettingsComponent() to create simple settings components.
• Lua: Script graphs can now override the default node name via the entry _name.
• Lua: Added staticPinCount to octane.NODE_ITEM_PROP.
• Lua: If setting a property for a node item or a GUI component fails, an error is raised.

Downloads:

Windows
- 64-bit commercial version (installer)
- 64-bit commercial version (ZIP archive)

Mac OS X
- 64-bit commercial version (DMG image)

Linux
- 64-bit commercial version (ZIP archive)

Happy rendering,
Your OTOY NZ Team

[nuno1980]

Good news for 1MS/px. But it would be excellent news for continuous (0 as infinity S/px).

[Notiusweb]

Wow, great support! Thank you for this release

[nuno1980]

This new version is ~1.24x slower than v2.25 at PMC kernel with parallelism 1. Why??????????????

[abstrax]

This new version is ~1.24x slower than v2.25 at PMC kernel with parallelism 1. Why??????????????

Don't know. Will have a look at it next week.

[coon_ra]

Are there any specific settings, or version of drivers, or else to make 1080 work in pmc of Octane 2.26? EVGA 1080 FTW failed to render with pmc kernel but works with PT and DL. Same time 580s work fine, just each is about half of 1080 performance.
PS: I got Octane 3.04 Demo, in PMC mode it renders on both current GPUs unlike 2.26 that uses only Geforce GTX 580 in PMC. Any idea how to make 2.26 use 1080 in PMC? Or it is impossible at all?

[nuno1980]

Will have a look at it next week.

Do you forget this?

[abstrax]

Are there any specific settings, or version of drivers, or else to make 1080 work in pmc of Octane 2.26? EVGA 1080 FTW failed to render with pmc kernel but works with PT and DL. Same time 580s work fine, just each is about half of 1080 performance.
PS: I got Octane 3.04 Demo, in PMC mode it renders on both current GPUs unlike 2.26 that uses only Geforce GTX 580 in PMC. Any idea how to make 2.26 use 1080 in PMC? Or it is impossible at all?

I had a look and found that it was an oversight during the merge into the v2 branch. I will release an update in the next couple days which will fix this problem.

[abstrax]

This new version is ~1.24x slower than v2.25 at PMC kernel with parallelism 1. Why??????????????

I had a look and the problem was introduced by the switch to the CUDA 7.5+8.0 toolkit combo. I did this because kernels would fail on Pascal with CUDA 6.5+8.0. Unfortunately this seems to break PMC for at least Maxwell. It not only renders slower, but also incorrect. I'm trying to find a solution for this issue.

[karanis]

Another VERY important fix for me in Linux is :

Standalone in command line does not need Xserver running anymore.

That was an important fix for me, automating renders with lua etc.

Thanks