OTOY Forums

renderer hangs up at the end of the frame (after 50 frames or so) but no error message. 2.04 did not do that
And 3ds max would just crash sometimes no reason
Thanks

nevermind maybe it is my GPU.
Error 700 on device 1 was on my slave render

i noticed i get most of my errors when i animate camera movement like a driver in the car POV therefore motion blur is everywhere when camera moves. maybe it is so demanding it crashes system

I'm still also getting this same problem, renders hang at the end of the frame. It can be after 2 frames, or 200 frames, but it will eventually freeze. Even on scenes that aren't too heavy / demanding, and with and without network rendering. I've sent some logs to Abstrax, hopefully he can work out what the problem is. All GPU temps are staying low, and I've set TDRDelay to 20 as recommended, also tried adding more fans to keep the temperatures really low, vram is more than enough, PSU is 1500w so more than enough, everything is stable except for Octane, it will hang at the end of a frame so I have to check my renders every 30 mins overnight to restart the render. Hopefully this can be sorted asap, or if someone can point us in the right direction if it is a hardware issue as it is a major problem.

mine did because my TEMP TARGET on EVGA precision was set to 80. I had to lower to 69 degrees and power target no more than 75%. On slave power target is set to 62% or it becomes too unstable but rendering speed stay about the same

Thanks for the suggestion, I don't really want to use anything other than stock settings Wight the cards. I intentionally bought the 780ti cards (and the more expensive optimised ones with the two fans) so that I didn't have to void the warranty by over or under clocking them. I'm sure most Octane users would be the same, just want to use a nice GPU out of the box to render in Octane. The max temp on my cards is suggested to be 95 degrees so I think if I tweak that then I won't be getting the same performance. This does seem like an Ocrabe issue as my cards work for anything else. Surely Octane can manage when a frame freezes to carry on, especially as if I restart the render immediately it carries on without a problem, I'd just rather not have to do this manually.

I see. It would be nice if Octane can autorecover and continue render

Hi,

I have also seen this problem. This can happen on a single frame or any frame in an animation sequence (which freezes the render). Can be complex or simpler scene. Once this happens, render won't restart without a full restart of the computer, all that gets displayed is the coloured square, no polygons, textures etc are registered all numbers are zero at the top of the render window. GPU devices are still registered with octane. I recall seeing this problem pop up in version 2.04.

Ta,
N.

I have also seen this problem. This can happen on a single frame or any frame in an animation sequence (which freezes the render). Can be complex or simpler scene. Once this happens, render won't restart without a full restart of the computer, all that gets displayed is the coloured square, no polygons, textures etc are registered all numbers are zero at the top of the render window. GPU devices are still registered with octane. I recall seeing this problem pop up in version 2.04.

For me, (and I think it is the same for Coilbook and others) this problem is slightly different. In my case the frames render and the progress bar reaches the end. The Log from Octane will say things such as:

An exception occurred on the device while executing a kernel
ERROR : CUDA error 719
failed to deallocate device memory

CUDA error 702 on device 1: The device kernel took too long to execute. This can only occur if timeouts are enabled. The context cannot be used anymore and must be destroyed. All existing device memory allocations from this context are invalid and must be reconstructed.
-> kernel execution failed (dl)
device 1: direct lighting failed
CUDA error 702 on device 1: The device kernel took too long to execute. This can only occur if timeouts are enabled. The context cannot be used anymore and must be destroyed. All existing device memory allocations from this context are invalid and must be reconstructed.
-> failed to deallocate device memory
CUDA error 702 on device 1: The device kernel took too long to execute. This can only occur if timeouts are enabled. The context cannot be used anymore and must be destroyed. All existing device memory allocations from this context are invalid and must be reconstructed.
-> could not get memory info
device 1: failed to upload data texture 4 of context 0