OctaneRender® 2026.4 for 3ds max® v26.11

[paride4331]

Hi mbetke,
I don’t think this is actually an issue. When you create a new material, the default setting is applied correctly.
The issue occurs when you open older materials. In that case, the material keeps its original color space settings, which can conflict with the new color space set for the scene. That actually makes sense, so the material’s color space needs to be changed manually to match the scene’s color space.
Regards
Paride

[coilbook]

Why does octane render empty screen so slowly? 5090 and 2 x 3080ti at only 350m? Old octane used to render with one 780ti at 1000 plus speed?

[paride4331]

Hi coilbook,
Could you please run a few tests so we can narrow down where the slowdown is coming from?
enable Load and Render Performance Report under Octane Preferences > Troubleshooting and run the same empty scene
First, disable both RTX 3080 Ti cards in Octane's Devices settings and leave only the RTX 5090 enabled. Start the Octane Live Viewer and let it run until the initial startup/initialization period is finished. Please report how long it takes from starting the Live Viewer until rendering actually begins, and then the stable Msamples/s after the initial startup period.
Then enable the RTX 5090 and only one RTX 3080 Ti, leaving the second 3080 Ti disabled. Run the same test and report the startup time and stable Msamples/s.
Finally, enable all three GPUs and repeat the same test, again reporting the startup time and stable Msamples/s.
Please post the complete report log.
Regards
Paride

[coilbook]

Hi Paride,

I am following up on slaves not using out of core ram. Even 32 gb did nothing. So the scene is borderline what gpu can hold in vram. I attached the image. master has 5090 and 2 x 3080 ti with 12 gb of vram. But the slaves have some 2080ti with 11 gb of vram. It seems that octane tries to use vram only therefore slaves fail. How can I tell octane after 10Gb of vram usage start using out of core?
Does all mesh must fit vram or textures?

UPDATE:
After hiding some hi poly meshes, slaves render fine as long as vram is bellow 11 gb due to 2080ti. So out of core is not working.

Thank you





"16 GB o-o-c limit could be the bottleneck. If the scene exceeds the available VRAM by more than those 16 GB, Octane may still run out of memory and crash instead of using more system RAM. Also, some data may still need to fit in VRAM, so large or complex meshes, displacement, subdivision, volumes, hair, etc. could be the actual cause. Since most textures are procedural, I’d look at the geometry and check the VRAM/o-o-c usage in the Octane log right before the crash.
Regards
Paride"

[coilbook]

Hi,
After adding cloth bind tyflow event, octane doesn't render. The scene is attached, Thank you

[paride4331]

Hi Paride,

I am following up on slaves not using out of core ram. Even 32 gb did nothing. So the scene is borderline what gpu can hold in vram. I attached the image. master has 5090 and 2 x 3080 ti with 12 gb of vram. But the slaves have some 2080ti with 11 gb of vram. It seems that octane tries to use vram only therefore slaves fail. How can I tell octane after 10Gb of vram usage start using out of core?
Does all mesh must fit vram or textures?

UPDATE:
After hiding some hi poly meshes, slaves render fine as long as vram is bellow 11 gb due to 2080ti. So out of core is not working.

Thank you
"16 GB o-o-c limit could be the bottleneck. If the scene exceeds the available VRAM by more than those 16 GB, Octane may still run out of memory and crash instead of using more system RAM. Also, some data may still need to fit in VRAM, so large or complex meshes, displacement, subdivision, volumes, hair, etc. could be the actual cause. Since most textures are procedural, I’d look at the geometry and check the VRAM/o-o-c usage in the Octane log right before the crash.
Regards

Paride"

Hi coilbook,
Since I don't have enough context about the scene itself, I'm approaching this more from a hardware architecture perspective and thinking about what could potentially be causing the issue.
Personally, I'd move the 3080s to a separate PC, or, if possible, pair the 3080 with the 2080s instead. That way, you could make better use of the 5090's full VRAM capacity, which is currently being heavily limited by the lower VRAM capacity of the 3080s.
The 3080s and 2080s have roughly the same amount of VRAM, so they should behave more consistently as a group, while using out-of-core memory to effectively make up for the 5090's additional VRAM.
Also, keep in mind that if hiding the geometry in out-of-core allows space for textures and additional geometry as well, the issue might simply be a matter of increasing the amount of out-of-core RAM. However, there's a practical limit to this. If you allocate more out-of-core memory than the system's total RAM can comfortably handle, you'll eventually start hitting system paging. At that point, performance will drop dramatically and the whole setup could become extremely slow, potentially to the point of being unusable.
Regards
Paride

[mbetke]

Is there an error in Out of Core display? VRam is full, System Ram filled with 100GB. But display says all fits into VRay. I set head room to 512MB. Also the rendering speed "feels" like there is OOC used. It's slower than usually.

[paride4331]

Is there an error in Out of Core display? VRam is full, System Ram filled with 100GB. But display says all fits into VRay. I set head room to 512MB. Also the rendering speed "feels" like there is OOC used. It's slower than usually.

Hi mbetke,
Out-of-core is working, but it’s not visible in the displayer. I believe this may be related to an SDK issue, as I’m seeing the same behavior in another plugin as well.
Regards
Paride

[coilbook]

Hi Paride,

I am following up on slaves not using out of core ram. Even 32 gb did nothing. So the scene is borderline what gpu can hold in vram. I attached the image. master has 5090 and 2 x 3080 ti with 12 gb of vram. But the slaves have some 2080ti with 11 gb of vram. It seems that octane tries to use vram only therefore slaves fail. How can I tell octane after 10Gb of vram usage start using out of core?
Does all mesh must fit vram or textures?

UPDATE:
After hiding some hi poly meshes, slaves render fine as long as vram is bellow 11 gb due to 2080ti. So out of core is not working.

Thank you
"16 GB o-o-c limit could be the bottleneck. If the scene exceeds the available VRAM by more than those 16 GB, Octane may still run out of memory and crash instead of using more system RAM. Also, some data may still need to fit in VRAM, so large or complex meshes, displacement, subdivision, volumes, hair, etc. could be the actual cause. Since most textures are procedural, I’d look at the geometry and check the VRAM/o-o-c usage in the Octane log right before the crash.
Regards

Paride"

Hi coilbook,
Since I don't have enough context about the scene itself, I'm approaching this more from a hardware architecture perspective and thinking about what could potentially be causing the issue.
Personally, I'd move the 3080s to a separate PC, or, if possible, pair the 3080 with the 2080s instead. That way, you could make better use of the 5090's full VRAM capacity, which is currently being heavily limited by the lower VRAM capacity of the 3080s.
The 3080s and 2080s have roughly the same amount of VRAM, so they should behave more consistently as a group, while using out-of-core memory to effectively make up for the 5090's additional VRAM.
Also, keep in mind that if hiding the geometry in out-of-core allows space for textures and additional geometry as well, the issue might simply be a matter of increasing the amount of out-of-core RAM. However, there's a practical limit to this. If you allocate more out-of-core memory than the system's total RAM can comfortably handle, you'll eventually start hitting system paging. At that point, performance will drop dramatically and the whole setup could become extremely slow, potentially to the point of being unusable.
Regards
Paride

Hi Paride,

I think there may be a bug with slaves and out-of-core rendering, particularly with detecting the card with the lowest VRAM and setting that as the VRAM limit.

For example, the master has a 3080 Ti with 12 GB of VRAM, while the slaves have 2080 Ti cards with only 11 GB. For some reason, Octane seems to think that 12 GB should be the limit because it only checks the master and ignores the VRAM of the cards on the slaves. This creates a problem showing in octane viewport 12 gb being used even though it should show 11 gb of vram used and 1 gv of ram used there fore slaves crash with out of memory errors.

Could you please check if Octane can detect the card with the lowest amount of VRAM across both the master and all slaves and use that as the VRAM limit?

Thank you.