Best version for RTX support Tweet

[samsue]

I tested it on my workstation at home and the same problems happen, this update for 3ds max, turn my week into a nightmare ... In fact it is always the same, as with Chaos Group, super news you are all amazed, so you decide to install ... big mistake. If always had problem, but in the same scene in version 2018 works fine

CUDA error 2 on device 0: out of memory
-> failed to allocate device memory
device 0: failed to allocate tone-map buffers
device 0: invalid memory pointers pass

Failed to free page-locked memory
CUDA error 700 on device 0: an illegal memory access was encountered
-> failed to destroy texture objects


CUDA error 700 on device 0: an illegal memory access was encountered
-> failed to bind device to current thread
device 0: failed to initialize context
CUDA error 201 on device 0: invalid device context
-> failed to create link
device 0: failed to create OSL compile context:

device 0: failed to compile module 0

UPDATE:

The problem is GPU Room, although in previous versions I have not changed this value and even reading the documentation, I can not understand the logic.

Well ... problem solved ... 5 days later ... with a solution that does not seem to me to have the slightest logic.

https://docs.otoy.com/StandaloneH_DEV/S ... ead%20room

GPU Head Room [MB]

Users can specify the amount of space reserved for other GPU users. Octane maximizes the amount of memory up to the headroom limit specified here. Note that GPU memory is faster than the system memory, therefore GPU Head Room tends to be set to a minimal level since it is practical to have the maximum amount of texture and geometry data fitted into the GPU VRAM.

[nildoe]

samsue Hi... so do you increase or decrease the default 300 value of the GPU Head room? to how much?

[samsue]

Sorry for the late answer.

It is a bit confusing, I increased from 300 to 600, corrected the problem, then without changing the scene, had to increase to 1000 and sometimes to the maximum. This has solved my problem with gpu fail.

But if you read the Octane documentation (which is quite outdated for 3d max) you don't feel like changing this parameter as I understand it.

[paride4331]

Hi samsue,
GPU Head Room specifies the amount of space reserved for other GPU users. OctaneRender maximizes the amount of memory up to the headroom limit specified here.
VRAM is faster than RAM, therefore GPU Head Room tends to be set to a minimal level since it is practical to have the maximum amount of texture and geometry data fitted into VRAM if you can dedicate one GPU to the OS display, and then dedicate one or more GPUs for rendering. The smoothest user experience with OctaneRender is dedicating one GPU for the display and OS to avoid slow and jerky interaction and navigation. The dedicated video card could be a cheap, low-power card since it will not be used for rendering, and it should be disabled under CUDA devices in Device Manager > Preferences.
In this situation, It is recommended matching the rendering GPUs in model and VRAM size. You can do multi-GPU rendering, but the OS interface may still be slow as all the GPU processing power is dedicated to the rendering process. In multi-GPU setups, the amount of RAM available to OctaneRender is not equal to the sum of the VRAM on the GPUs, but it is restricted to the GPU with the smallest amount of VRAM.
regards
Paride

[Rik]

Hi Paride

I think I understand how you could prevent the low power card from rendering (by disabling cuda or just unticking the 'active' box in the devices tab?)

But how do you tell the high power card to only do rendering and not waste resource running the OS display?

Also, with the 2 card limit, could you use 1 low power OS display card (deactivated from rendering) and 2 high power cards for the rendering? Or would that count as 3 cards?

[paride4331]

Hi Paride

I think I understand how you could prevent the low power card from rendering (by disabling cuda or just unticking the 'active' box in the devices tab?)

But how do you tell the high power card to only do rendering and not waste resource running the OS display?

Also, with the 2 card limit, could you use 1 low power OS display card (deactivated from rendering) and 2 high power cards for the rendering? Or would that count as 3 cards?

Yes, you can. Keep in mind connecting your monitor/s to non active GPU.
For a proper functioning, usually it should be in PCIe 16_1 slot.
Regards
Paride

[Rik]

Thanks Paride.

Would it not be better to have the rendering cards in the best PCIE slots to speed up the 'scene loading' and 'building render target' waiting times? Or does the number of PCIE lanes not matter for this part of the process? (Which takes ages)

[paride4331]

Hi rik,
PCIe lanes are crucial during the scene loading from PC to GPU.
At least 8x would be recommended.
Regards
Paride