Page 1 of 1
Start time run 1 million polygons in OCtane render
Posted: Wed Aug 10, 2016 9:15 am
by Goldisart
I think that this is definitely a bug in the compatibility of hardware and drivers are faced with the problem of launch images - very slow 12 seconds of video attached
Octane render - 12 second
https://www.youtube.com/watch?v=DZ8eF-x ... e=youtu.be
Vray 3.4 - 4 second
https://www.youtube.com/watch?v=vXCe521 ... e=youtu.be
it is possible that this problem on the delay time, only on my computer .... please note thank you
770- 4 780- 6 780- 6
18 GB - RAM
nvidia 368.81
octane 3.15
p.s. - asks regular users Octane render to try this test yourself and to pinpoint the start time
Re: Start time run 1 million polygons in OCtane render
Posted: Wed Aug 10, 2016 10:12 am
by Lewis
Can you test same in octane 2.25 just for comparison. To me it also seems that loading time into GPUs in OR 3.x is long(er) than before.
Re: Start time run 1 million polygons in OCtane render
Posted: Wed Aug 10, 2016 6:20 pm
by HHbomb
3.x is slower to load in gpu and is not fully multithreaded, but in your case it seems to be very slow.. even on my laptop just 4 second to load 1M+ polys. (i7 3630qm gtx 670mx / octane 3.03.1 4.8)
Re: Start time run 1 million polygons in OCtane render
Posted: Wed Aug 10, 2016 9:15 pm
by ThaRaven
10 sec for 2,6Mil Polys here...
Re: Start time run 1 million polygons in OCtane render
Posted: Thu Aug 11, 2016 7:39 am
by Goldisart
a bit of analysis... 1 video card to run 780 - 6 sec 1 780 graphics card - 8 seconds . 1 video card 770 - 12 seconds - it is obvious that the octane render starts only after start ( response) all three cards....
////////////////
interesting dear programmers - you will respond to this question that would be a little bit to understand how it works ... -- and why it doesn't happen in VRAY
I think this is due to the bandwidth of PCI EXPRESS
Re: Start time run 1 million polygons in OCtane render
Posted: Fri Aug 12, 2016 4:23 am
by azen
Hi,
Quite a lot of stuff that was being handled in GPU memory, is now handled in CPU memory (e.g. samples integration, frame buffers). Are you using PCIe 1x buses? If you are using slow PCI buses. Performance hits on slow PCI buses (e.g. PCIe 1x ) have been as high as 30%.
These are drawbacks to an overhaul to Octane Kernels. These overhauls were needed in order to accommodate the new features of V3. Work to offset these performance drawbacks is ongoing. In saying that, these are indeed drawbacks rather than bugs - necessary sacrifices in some areas to make V3 possible. A lot of much needed features simply could not be implemented without this refactoring at the core level. Apologies for the inconvenience.
Re: Start time run 1 million polygons in OCtane render
Posted: Fri Aug 12, 2016 7:09 am
by Goldisart
Thanks for the detailed response... We are patiently waiting for reconstruction ....
I need to change the motherboard .... according to GPU Z - GPU 770 (1 - win) works on the PCI-E bus 2.0x16@x2. ----- GPU (2) 780 @x8 ---- GPU (3) 780@x4 - the real weak link is 770
Re: Start time run 1 million polygons in OCtane render
Posted: Mon Aug 15, 2016 1:08 am
by azen
Goldisart wrote:Thanks for the detailed response... We are patiently waiting for reconstruction ....
I need to change the motherboard .... according to GPU Z - GPU 770 (1 - win) works on the PCI-E bus 2.0x16@x2. ----- GPU (2) 780 @x8 ---- GPU (3) 780@x4 - the real weak link is 770
That sounds about right. It may become necessary pretty soon to move away from the Kelper GPUs as soon as reasonably possible.