Page 1 of 1

Cuda error (2)

Posted: Thu Jun 29, 2023 5:15 pm
by Rik
Hi
Any idea what this means please...

Started logging on 29.06.23 17:55:39

OctaneRender Studio+ 2022.1.1 (12000102)

Can't connect node 'tx-Map #773141-1076' (NT_TRANSFORM_SCALE, 00000000FEB63750) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 00000000FEB63290)
Can't connect node 'tx-Map #2122715324' (NT_TRANSFORM_SCALE, 000000013E82C2C0) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000013E82C780)
Can't connect node 'tx-Map #2122715253' (NT_TRANSFORM_SCALE, 000000014B1A5E60) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000014B1A5150)
Can't connect node 'tx-Map #773141-1076' (NT_TRANSFORM_SCALE, 000000014B1AB510) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000014B1AD3F0)
Can't connect node 'tx-Map #773169-400' (NT_TRANSFORM_SCALE, 000000013E489AA0) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000013E48A8E0)
Can't connect node 'tx-Map #773141-1076' (NT_TRANSFORM_SCALE, 000000014ABA6100) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000014ABA3FC0)
Can't connect node 'tx-Map #2122715365' (NT_TRANSFORM_SCALE, 000000014ABA8E20) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000014ABA66F0)
Can't connect node 'tx-Map #2122715373' (NT_TRANSFORM_SCALE, 000000014ABABB40) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000014ABACBE0)
Can't connect node 'tx-Map #2122715253' (NT_TRANSFORM_SCALE, 000000014ABAB1C0) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000014ABAF7D0)
Can't connect node 'tx-Map #773141-1076' (NT_TRANSFORM_SCALE, 000000014ABB11F0) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 000000014ABB4890)
Can't connect node 'tx-Map #772863' (NT_TEX_FLOATIMAGE, 0000000168634900) to static pin 'displacement' (P_DISPLACEMENT, type PT_DISPLACEMENT) of node 'Glossy material' (NT_MAT_GLOSSY, 00000001686347D0)
Can't connect node 'tx-Map #773141-1076' (NT_TRANSFORM_SCALE, 00000001686437B0) to static pin 'projection' (P_PROJECTION, type PT_PROJECTION) of node 'Noise texture' (NT_TEX_NOISE, 0000000168643680)
CUDA error (2) on device 1: out of memory
-> failed to allocate device array


The scene I'm rendering easily fits in the 2800ti
I have rtx and denoising off

I always have parallel samples at 32 without problems.
I've just reduced it down to 12 and it renders OK. The vram usage is about 2.5gb and the 2080ti has 11gb

Any idea what could be going on please?

Also, snip n sketch doesn't work when the octane viewport is open, even when it's finished rendering. weird flashing screens happen as though 2 processes are trying to use the gpus at the same time. As soon as I shit the octane viewport it works fine.

Re: Cuda error (2)

Posted: Fri Jun 30, 2023 10:29 am
by paride4331
Hi Rik,
Are you using any displacement textures map?
Would you share your scene, if you want in dm.
Regards
Paride

Re: Cuda error (2)

Posted: Fri Jun 30, 2023 11:19 am
by Rik
Oops, just read my post. Think I meant 'shut' rather than 'shit' :lol: :lol:
Probably a good description of the day I was having though. ;)

I am using displacement maps and also large textures, 34mb jpegs 9k resolution for the brickwork. I appreciate these might be large but I've been using the scenes, textures and materials for a few years now and never had this issue.

I updated to the latest release recently. Sorry, can't remember which version I was using previously.

I'm now having to replace the larger textures with much smaller versions and hoping it won't be noticeable.

Seems something has changed with the parallel samples. Previously, I've always had it at max, 32. Now I have it at the minimum, 1.
Don't think it made much of a difference to Vram usage in the past but now it seems to make a huge difference.

I have an octane proxy of a tree I've used for years with parallel samples at 32.
Here's how the parallel samples affect the Vram now.

PS 1 = 0.19GB
PS 32 = 2.38GB

Another example is a few simple planes with tree images on them...

PS 1 = 0.088GB
PS 32 = 2.28GB

So now I'll fill my 11GB Vram 2080Tis with just a few different trees unless I minimise the parallel samples :(

What's the best way to send you a scene? (using max r14 :oops: )

Just to emphasise the point - previously when trying to squeeze out a bit more speed I've been frustrated by the fact that that max parallel samples were limited to 32. If I could have gone higher I would.
Now I'm having to use the minimum.

Re: Cuda error (2)

Posted: Mon Jul 03, 2023 12:54 pm
by paride4331
Hi Rik,
You can use wetransfer then share wetransfer download link in private message.
in any case parallel Samples Controls how many samples OctaneRender calculates in parallel.
Smaller values require less memory to store the sample's state, but increase render time. Larger values require more memory, but reduce render time.
So by reducing it, you increase the available space in VRAM.
Also you can increase the "the out of memory" space.
Regards
Paride