Constant crashes in LV and PV.

[FAZ]

Hello,
I am getting constant crashes in either the LV or PV. If I am in the LV I get a red error message. If rendering to PV, the screen freezes and computer auto-restarts. (log attached). The larger the size and samples the renders are, the sooner it will crash after a few frames. Might be memory related? This project is setup to be 4k pre-rendered 360 VR content. So my camera is using the octane panoramic lens. Using path-tracing with conservative settings, nothing fancy at this stage of development. Mainly rendering at 2000x1000 for quicker playtesting. Smaller the render, the more frames it outputs... but always crashes after about 200-700 frames in.

My current hardware setup...

Windows 7 pro
Cards: Three gtx 1080s in my tower; and two 980ti on an amfeltec open air frame connected via host card.
Memory: 8gb vram on 1080s but i think it is only using 6gb from the 980s.
Driver: 372.90
Octane c4d plugin: 3.03.3-R1
Host App: Cinema 4D R17.053

Notes/Questions:
We are using active booleans in c4d. I was thinking that maybe Octane doesn't like booles?

This is pretty urgent since we have not had any problems until just recently (1 year Octane user) and are due to present a pre-viz to client very soon. Any advice on what to try out on our end in the meantime would be much appreciated.

Thanks!
Alex

[bepeg4d]

Hi FAZ,
in general, if the machine auto-restarts, there could be a power issue.
Could you try by removing just one card from your setup?
Do you have the same issue if you disconnect the amfeltec?
ciao beppe

[FAZ]

Hi FAZ,
in general, if the machine auto-restarts, there could be a power issue.
Could you try by removing just one card from your setup?
Do you have the same issue if you disconnect the amfeltec?
ciao beppe

Hi Beppe,

I don't think it is a power problem because we are using less power than we used to, and the scene is not as complex. We used to have x4 980tis in this tower and never had a freeze up due to power. Right now we are only use x3 1080s which are supposed to be more power efficient. The amfeltc is on its own power supply.

I am trying your suggestion now with just the tower running, w/o the amfeltec connected. I'll let you know how it goes. Thanks!
Alex

[FAZ]

@beppe,

So I tried taking out the amfeltec rig like you mentioned. It seems to work well without it, so far. Then, I hooked it up to a co-worker's workstation and it worked fine. Only difference between his and my workstations were the card models... Mine we're mixed- 1080s+980s. His were all 980s (6 total).

Have you heard anything about mixing card models on the same system with a gpu expander? Seems like a weird hardware conflict since they are all nvidia brands.

So now it seems to work for both of us since we had an extra amfeltec rig.

My setup...
x3 1080s in tower
x2 1080s in the amfeltec rig

His setup...
x2 980ti in tower
x4 980ti in amfeltec rig

Both working as long as all the gpu models are the same on each setup.

Can anyone else confirm this matching card models per rig issue? Maybe there are other model combos that work too? and maybe the 1080s are the conflict with other cards?

I'll will be rendering these overnight @ 2,700 frames. I'll let you know if there were any other problems.
Thanks!
Alex

[bepeg4d]

Hi Alex,
interesting info and thanks for sharing your tests.
There was an user in the general section with a problem like yours when a 1080 was mixed with Maxwell GPU with a riser
It seems an issue introduced with Pascal GPU, I guess
ciao beppe

[FAZ]

All frames rendered flawlessly overnight! Yes, that makes sense because I remember people mixing maxwell cards w/o a problem.

Question:
What have you heard to be more efficient for large and numerous frame animations through Octane?... Network Rendering or the Team Render? Our final output will be 4k panoramics at about 1500 PT samples.

Thanks,
Alex