With the recent posting about Octane going to be using OpenCl in version 3 I have some queries about the difference between the OpenCl and the CUDA version, especially when it comes to gfx cards. Some have argued that AMD cards are better performance wise for OpenCl than Nvidia cards. Are there any performance hits for NVidia gfx card users when Octane switches to OpenCl? For example if you have an Nvidia 980 if you render a scene in OpenCl on Octane V3 will there be a performance loss vs rendering the same scene using CUDA?
I am thinking of getting a new gfx card in the next few months and am thinking of switching from Nvidia to AMD. The only reason I went with Nvidia is because Octane at the time was strictly CUDA based.
Octane V3 - OpenCl vs CUDA performance difference.
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
+1
GTX 1080Ti 11GB (3x), Water-cooled
Intel i7-5820K 6-core @ 3.3GHz
Windows 10 Pro 64-bit, 32GB RAM
Intel i7-5820K 6-core @ 3.3GHz
Windows 10 Pro 64-bit, 32GB RAM
- mib2berlin
- Posts: 1194
- Joined: Wed Jan 27, 2010 7:18 pm
- Location: Germany
Hi, it would be nice to get exact numbers but it makes no sense to use OpenCL on a Cuda card if you can use native Cuda.
Octane 3 is Cuda and OpenCL.
Btw. GTX 900 is also very fast on OpenCL.
http://render.otoy.com/forum/posting.ph ... =9&t=45604
Cheers, mib
Octane 3 is Cuda and OpenCL.
Btw. GTX 900 is also very fast on OpenCL.
http://render.otoy.com/forum/posting.ph ... =9&t=45604
Cheers, mib
Opensuse Leap 42.3/64 i5-3570K 16 GB
GTX 760 4 GB Driver: 430.31
Octane 3.08 Blender Octane
GTX 760 4 GB Driver: 430.31
Octane 3.08 Blender Octane
- gabrielefx
- Posts: 1701
- Joined: Wed Sep 28, 2011 2:00 pm
I'd like to know how Otoy ported all the Cuda kernels to work with ATI gpus...
Also I'd like to know if it's possible to work with ATI and Nvidia gpus mixed render farm.
regards
Also I'd like to know if it's possible to work with ATI and Nvidia gpus mixed render farm.
regards
quad Titan Kepler 6GB + quad Titan X Pascal 12GB + quad GTX1080 8GB + dual GTX1080Ti 11GB
I don't think OpenCL and CUDA will work together, you have to choose how it will calculate since they're totally different things
Nvidia cards can run CUDA and OpenCl, AMD cards can just run OpenCl. I'm not 100% sure that there are going to be two different versions of Octane from version 3 onwards, like one for CUDA cards and one for OpenCl and it might just be one version being an OpenCl one. If there is only going to be 1 version of Octane V3 and it is going to be OpenCl then Nvidia cards might perform very differently compared to when the card is using CUDA.
Looking at Luxmark which uses OpenCl the difference in performance between the AMD R290X and the Titan X is quite large with the R290X coming ahead.
Benchmark results for Luxmark using OpenCl.
Score (thousands of samples per second) higher is better:
AMD R290X : 2950
Titan X : 2150
Bear in mind the R290X is priced in the UK at £279 and the Titan X £868 and in this benchmark the R290X wins against the Titan X using OpenCl.
Link to the benchmark : http://www.anandtech.com/bench/product/1447?vs=1439
I am glad that Octane is going to be using OpenCl. As for gfx cards I am going to wait until I see the benchmarks for Octane 3 using OpenCl before I make the jump on buying either an AMD or Nvidia card. I AM NOT going to spend over £800 on an Nvidia card IF a £300 AMD card can not only match it, but beat it in performance. I really am going to wait and see what the benchmarks are like.
Looking at Luxmark which uses OpenCl the difference in performance between the AMD R290X and the Titan X is quite large with the R290X coming ahead.
Benchmark results for Luxmark using OpenCl.
Score (thousands of samples per second) higher is better:
AMD R290X : 2950
Titan X : 2150
Bear in mind the R290X is priced in the UK at £279 and the Titan X £868 and in this benchmark the R290X wins against the Titan X using OpenCl.
Link to the benchmark : http://www.anandtech.com/bench/product/1447?vs=1439
I am glad that Octane is going to be using OpenCl. As for gfx cards I am going to wait until I see the benchmarks for Octane 3 using OpenCl before I make the jump on buying either an AMD or Nvidia card. I AM NOT going to spend over £800 on an Nvidia card IF a £300 AMD card can not only match it, but beat it in performance. I really am going to wait and see what the benchmarks are like.
- taylorcoxmotion
- Posts: 4
- Joined: Fri Jan 17, 2014 7:10 pm
Add in another "+1" for me.
We've been led to believe up to this point that OpenCL just isn't nearly mature enough a platform to do what Octane has been doing with CUDA. So my question now is: Has Otoy captured some sort of magic? Or is OpenCL support just a noble attempt to make sure that everyone can at least run Octane, with the unfortunate caveat of decreased performance?
If OR 3 on an AMD card is at least able get close to similar performance from an NVIDIA card (like 70% or better), then that'll be good enough for me. Especially if the next Mac Pro incorporates their version of Dual FirePro 9100s, rocking 16 GB of VRAM apiece. The current Mac Pro D700 cards are based on FirePro 9000s, so this hopefully is a safe assumption.
We've been led to believe up to this point that OpenCL just isn't nearly mature enough a platform to do what Octane has been doing with CUDA. So my question now is: Has Otoy captured some sort of magic? Or is OpenCL support just a noble attempt to make sure that everyone can at least run Octane, with the unfortunate caveat of decreased performance?
If OR 3 on an AMD card is at least able get close to similar performance from an NVIDIA card (like 70% or better), then that'll be good enough for me. Especially if the next Mac Pro incorporates their version of Dual FirePro 9100s, rocking 16 GB of VRAM apiece. The current Mac Pro D700 cards are based on FirePro 9000s, so this hopefully is a safe assumption.
Well AMD has been fixing their issues recently and keep doing it. Thea and Blender Cycles are also getting OpenCL ports. Oh wait AMD has assigned a dedicated team to get Cycles to work on their card(huge patch submitted a few days ago). 

This is very interesting. I was also on the verge of upgrading my GPUs. I'll wait now and see some benchmarks. Nvidias monopoly in all these CUDA applications has lead to some ridicules pricing for their top CUDA performers simply because people have nowhere else to go. This will shake things up a bit and hopefully teach Nvidia a lesson *fingers crossed*
Yeah i'd like to know which will be better choice. I'm going to go NY in October for 10 days and decided to buy 2x Titans for $999 for each where $1800 in my country
But
would be great to know what AMD will give us before we buy new gpus.


http://www.sadece.com.tr
--------------------------------------------------------------------------
i7 990x // 12GB Ram // 2x Titan X + Titan + 780 // Octane for C4D // Win7 x64
--------------------------------------------------------------------------
i7 990x // 12GB Ram // 2x Titan X + Titan + 780 // Octane for C4D // Win7 x64