Re animation the single biggest issue is updating the geometry, say you're doing a test, the actual render takes only a few secs to get a simple preview, but the geom update takes 30 secs that means what should only take half an hour, now takes 5-6 hours, for a simple test animation render!!!... and there's little you can do with a decent amount of geometry that changes (so instancing doesn't help)!
Is there anything in the works to alliterative this major problem, what about the fact that only one core is used during voxelization?
animation voxelization way tooo slow!!
Are you sure it's the voxelization process that's too slow?
The larger part for me is copying the actual geometry to the VRAM, which as far as I know is as fast as standard hard drives can handle anyway...
I do see your point, though.
The larger part for me is copying the actual geometry to the VRAM, which as far as I know is as fast as standard hard drives can handle anyway...
I do see your point, though.
Octane for Rhino | Windows 8.1 x64 | i7-3820 OC | GTX970 4GB OC & GTX 560 Ti 1GB OC | 32GB DDR3
I don't think the loading is the problem, at least not for the max plugin. All the materials are kept in ram afaik and I don't think the geometry takes that long to load, so it must be the voxelization. When the voxelization occurs I see one of my 8 cores being used - that's where the problem is, I assume!
Can't get an answer from Otoy as well, sent them an email, posted in the forum before.... no response. Simple question: is there anything that can be done to speed up voxelization or how do I improve geometry updates for test renders?
Can't get an answer from Otoy as well, sent them an email, posted in the forum before.... no response. Simple question: is there anything that can be done to speed up voxelization or how do I improve geometry updates for test renders?
Win 7 64 | 3 x GTX 580 | i7 950 | 24GB
- mauricesteenbergen
- Posts: 31
- Joined: Tue Dec 04, 2012 9:37 am
My workflow is C4D/Octane Standalone, and I do my testrenders from C4D using the Hardware (OpenGL) renderer, it's from RAM, has textures, and it renders at 1 second per frame. I can also use the external compositing tags with this, so i can already start to work on compositing footage into screens etc, in After Effects. While I'm working on all that, Octane Standalone renders away on a dedicated GPU.
Ofcourse it would be more optimal if that entire export process would be faster, but at the moment export and voxelisation of larger scenes takes about 60 seconds, and my renders are about 2-3 minutes each. In C4D, using the physical render engine, and GI, AO etc enabled it would easily take 45 minutes per frame. Plus with Octane my renders don't look a like a cartoon anymore, but beautiful like they should. I can still use this workstation while rendering too (AE, Premiere etc).
Just my 2 cents.
Ofcourse it would be more optimal if that entire export process would be faster, but at the moment export and voxelisation of larger scenes takes about 60 seconds, and my renders are about 2-3 minutes each. In C4D, using the physical render engine, and GI, AO etc enabled it would easily take 45 minutes per frame. Plus with Octane my renders don't look a like a cartoon anymore, but beautiful like they should. I can still use this workstation while rendering too (AE, Premiere etc).
Just my 2 cents.
Render: Geforce GTX480 (EVGA) (1.5GB) | OC @ 1.0V / 750Mhz
Display: Geforce 8800GT (NVIDIA) (512MB)
OS: Win 7 x64 | CPU: Quad Intel Xeon 2.66GHz (12GB)
Display: Geforce 8800GT (NVIDIA) (512MB)
OS: Win 7 x64 | CPU: Quad Intel Xeon 2.66GHz (12GB)
I'm curious too if this process can be optimised?
With a decent scene it is sooo slow.
Only thing that helps is to set everything to proxy. Then geometry transfer is faster after 1st frame, but rendering is slower.
I don't know what happens internally when using instances (proxies), but i feel there is a room for optimisation.
Suv
With a decent scene it is sooo slow.
Only thing that helps is to set everything to proxy. Then geometry transfer is faster after 1st frame, but rendering is slower.
I don't know what happens internally when using instances (proxies), but i feel there is a room for optimisation.
Suv
I have no idea if this would be possible but could you divide your scene into static and dynamic pieces? The static pieces would be loaded once and only the dynamic pieces would need to be voxalized for each frame. otoy would need to provide a mechanism to do this in the standalone but plugins might be able to handle this. Just a thought...
Win8/64, I7-4770K (3.5Ghz) 24GB, GTX-980ti (6GB) / GTX-TITAN (6GB)
- BorisGoreta
- Posts: 1413
- Joined: Fri Dec 07, 2012 6:45 pm
- Contact:
It is just a matter of making voxelization process multithreaded, at the moment it's single threaded.
19 x NVIDIA GTX http://www.borisgoreta.com
well yes that's exactly what I'm saying (or assuming), but apparently it does show up on any of Otoy's todo lists, which is frankly very strange, because animation has to be top priority, or at least it ought to beBorisGoreta wrote:It is just a matter of making voxelization process multithreaded, at the moment it's single threaded.
Win 7 64 | 3 x GTX 580 | i7 950 | 24GB
Just wanted to say I noticed a 30 % bump in voxelization speed when I switched to a 680 pci-e gen 3 card from a 580. Depends on the scene though.
SuperMicro X9DAI
Dual Xeon E5-2670
64 GB DDR3-1600
Dual EVGA 680 4 GB
Win 7
Nvidia Driver 310.71
ASROCK Extreme 11
Intel 3970X OC'd 5.2 GHZ
64 GB DDR3-2400
Tri EVGA 680 4 GB
Win 7
Nvidia Driver 310.71
Dual Xeon E5-2670
64 GB DDR3-1600
Dual EVGA 680 4 GB
Win 7
Nvidia Driver 310.71
ASROCK Extreme 11
Intel 3970X OC'd 5.2 GHZ
64 GB DDR3-2400
Tri EVGA 680 4 GB
Win 7
Nvidia Driver 310.71
brue wrote:Just wanted to say I noticed a 30 % bump in voxelization speed when I switched to a 680 pci-e gen 3 card from a 580. Depends on the scene though.
This might be because the 680 is PCI 3.0 where the 580 is 2.0.
is the slot you have it plugged into a PCIe 3.0 slot?
I never knew any actual data on speed increase in voxelization, but this makes me happy because I am looking at getting a new motherboard with 3-4 3.0 slots and place 2 690s in there and maybe 1-2 more if everything works out. We are doing animations now and every second you can speed a frame up really counts.
Also, I am getting a SSD drive for my main drive instead of an HDD, this should also speed it up during load. Anyone try both and see a speed increase?
Intel quad core i5 @ 4.0 ghz | 8 gigs of Ram | Geforce GTX 470 - 1.25 gigs of Ram