OctaneRender® for Maya® 2.26.1 - 7.20.1 Win [OBSOLETE] Tweet

[JimStar]

Otoy® is proud to announce the availability of a new version of OctaneRender™ for Maya®
The world's fastest and most feature-rich unbiased render engine that integrates completely into Autodesk® Maya®.

Maya® Version Requirements

This release will work with Maya® Versions 2015, 2016, 2016.5 and 2017 64-bit on MS Windows operating systems.

COMPATIBILITY AND OCTANERENDER STANDALONE REQUIREMENT

To run OctaneRender™ for Maya®, you need to also have an activated OctaneRender™ Standalone copy activated on the machine you wish to install the plugin onto.
You cannot purchase and use only the Maya plugin and use it without also owning an activated copy of OctaneRender™ Standalone on your machine.

CHANGES SINCE THE PREVIOUS VERSION

• New 2.26.1 rendering core.
• Implemented the asynchronous mode for LiveDB/LocalDB dialog. Now you can acquire multiple materials from DB not closing the dialog.
• Implemented support of different directories for LocalDB data, see the new setting in Octane globals. Current Octane core requires you restart a plugin to see the effect of this setting change.
• Improved displaying of Octane textures in Maya's viewport.
• Fixed default passes channels naming. Common layer IDs for multilayer EXRs are now used (same IDs as produced by standalone for multilayer EXRs) if no special names are set by a user.
• Fixed camera mapping functionality bug (wrong mapping on some geometry types).
• Minor fixes and improvements...

DOWNLOAD

OctaneRender for Maya 2.26.1 - 7.20.1

For network slave and daemon please download standalone installer here:
viewtopic.php?f=24&t=57368


Yours,
The OctaneRender™ Team.

[itsallgoode9]

finally, thank you! will test it out this weekend.

[renmaxhb]

demo version?

I need it.

[itsallgoode9]

so, I just installed it and it does in fact render on my 1080's

BUT.....

1080 FTW card renders @3.66Ms/sec
780 card renders at @ 4.59 Ms/sec

any clue what is up with that? The 1080 FTW should be nearly twice as fast as my 780's but it's actually slower it seems?

[itsallgoode9]

double post. is there a delete button anywhere?

[SimonPercy]

So it seems to give me the same beauty output as 7.14 so thats good.
Only since the AOV system has changed my old renders will not output the old AOV setups any more... (which is fine... just not going to change all my scenes so close before my deadline...)

Anyway, seems stable, not fast indeed!


@ItsAllGoode9
Maybe this?

Since 2.26 reduced the PMC performance on pre-Pascal GPUs and PMC didn't run on Pascal GPUs at all, we had to switch back to CUDA 6.5 for pre-Pascal GPUs and enable PMC for Pascal GPUs.

Unfortunately using CUDA 6.5 + 8.0 for Pascal GPUs, crashes the PMC kernel, forcing us to use CUDA 7.5 + 8.0 for Pascal GPUs, which means that PMC runs a little bit slower on Pascal GPUs than on the corresponding Maxwell GPU. Since we don't have the resources to investigate this problem further, there isn't much we can do here about it.

[itsallgoode9]

@ItsAllGoode9
Maybe this?

Since 2.26 reduced the PMC performance on pre-Pascal GPUs and PMC didn't run on Pascal GPUs at all, we had to switch back to CUDA 6.5 for pre-Pascal GPUs and enable PMC for Pascal GPUs.

Unfortunately using CUDA 6.5 + 8.0 for Pascal GPUs, crashes the PMC kernel, forcing us to use CUDA 7.5 + 8.0 for Pascal GPUs, which means that PMC runs a little bit slower on Pascal GPUs than on the corresponding Maxwell GPU. Since we don't have the resources to investigate this problem further, there isn't much we can do here about it.

so technically they processed are request by getting pascal to work with v2 as we requested but....

it's and expected outcome that our 10xx series cards and they have not plan to fix that?


am I understanding that blurb right?
I am rendering pathtrace, not PMC and I am still getting that drop in speed. according to this, I shouldn't be affected.

[calus]

1080 FTW card renders @3.66Ms/sec
780 card renders at @ 4.59 Ms/sec

any clue what is up with that? The 1080 FTW should be nearly twice as fast as my 780's but it's actually slower it seems?

Right, according to the tests in standalone V3, the gtx1080 is 1.75x as fast as the gtx 780.

You should check if you have the problem in standalone 2.26.1:
viewtopic.php?f=24&t=57368#p294643

If it is the same issue, the problem is in Octane 2.26.1 core,
path tracing loosing half of pascal GPU speed was not expected at all,
Abstrax only talks about PMC runing a little bit slower on Pascal.
So should be reported in the 2.26.1 standalone topic.

Then check also previous standalone version 2.26.0 with pascal support: viewtopic.php?f=24&t=56982#p292546
(maybe this huge slow down comes from the fix in 2.26.1 to avoid "reduced PMC performance on pre-Pascal GPUs and PMC not working on Pascal GPUs".)

[itsallgoode9]

Thanks! I will do that today

[itsallgoode9]

so...

standalone, 2.26.1:
1080 gpu: pathtrace @ 4.71-ish Ms/sec
780 gpu: pathtrace @ 5.56-ish Ms/sec


standalone, 2.26.0:
1080 gpu: pathtrace @ 4.71-ish Ms/sec
780 gpu: pathtrace @ 5.56-ish Ms/sec

so looks like it's core performance issue? I'll post there and see what's up. Maybe that's expected in V2 with Pascal and I wrongly assumed that you'd get the speed benefits of the new GPU and not just basic functionality so it can run along side the Maxwell GPUS