https://en.wikipedia.org/wiki/Vulkan_(API)pixym wrote:Is Vulkan kernel a new true kernel you will have in Octane or just an integration of an Nvidia technology ???
New RTX will be ~30% faster than Pascal running CUDA API but running Vulkan API, RTX will be 6x~8x faster than Pascal.dtxtreme wrote:NVIDIA’s RTX speed claims fall short without game API support..
Expect 30-40% more speed over 1080TI for pure CUDA computing according to this article.
https://www.engadget.com/2018/08/25/nvi ... ccounter=1
I hope the CUDA benchmarks will turn out better than this.
RTX feels a bit risky to invest in right now.
Sure but will Octane still be the unbiased Octane we love using the new APIs??nuno1980 wrote:New RTX will be ~30% faster than Pascal running CUDA API but running Vulkan API, RTX will be 6x~8x faster than Pascal.
Several things :dtxtreme wrote:Sure but will Octane still be the unbiased Octane we love using the new APIs??nuno1980 wrote:New RTX will be ~30% faster than Pascal running CUDA API but running Vulkan API, RTX will be 6x~8x faster than Pascal.
viewtopic.php?f=9&t=68584
I'm a bit worried!
I realized i am totally clueless now, when it comes to these different APIs and whatnot. I mean, there was never any doubt i am total layman and have no clue about any details regarding this stuff, but for years it was said that Octane is CUDA-based - whatever that meant, LOL. Now i read Octane is "built" on Optix, DirectX, will be on Vulkan, Metal, etc... what the hell? Does it mean that all these stuff, all these APIs - are equal - to what CUDA is? Or what is the difference?Grumly wrote:Several things :dtxtreme wrote:Sure but will Octane still be the unbiased Octane we love using the new APIs??nuno1980 wrote:New RTX will be ~30% faster than Pascal running CUDA API but running Vulkan API, RTX will be 6x~8x faster than Pascal.
viewtopic.php?f=9&t=68584
I'm a bit worried!
- a GAME need real time and must work on other hardware... do RTX on Game is biased.
- Octane does not want to be real time, just quicker but withiout beeing biased.
Several Technology in the RTX chips can help to stay unbiased but be quicker : the AI tensors can reduced the number of itteration needed with AI Denoiser and reduced the number of ray needed with selecting only ray which ended in the camera.
Vulkan gives a API to use these hardware, but does not decide if it's unbiased or not, it's the software, which use the API, which decides. For exemple a game can decide to use raytracing only for the sadows but a biased result for other part. Octane does not do that.
Why would you buy the cards before they get official support? It makes no sense.Timmaigh! wrote:
I realized i am totally clueless now, when it comes to these different APIs and whatnot. I mean, there was never any doubt i am total layman and have no clue about any details regarding this stuff, but for years it was said that Octane is CUDA-based - whatever that meant, LOL. Now i read Octane is "built" on Optix, DirectX, will be on Vulkan, Metal, etc... what the hell? Does it mean that all these stuff, all these APIs - are equal - to what CUDA is? Or what is the difference?
And in addition to this question, one practical thing of primary interest to me, which i alluded to in one of my previous posts, but nobody cared to answer: say i preorder or simply buy on release one or 2 RTX cards, while simultaneously selling my current Pascal cards.... i read that Turing needs CUDA 10, which has not been released yet... will Octane work for me anyway, cause its OptiX, rather than CUDA based now? Or will i need to wait for the release of some CUDA 10 based Octane version for Turing cards to work at all? Like it used to be in case of Pascal and CUDA 8?
I am not asking now about all those hypothetical speed-ups, i get that is 2019 thing - i just care about basic functionality. I would hate to sell my Pascals, replace them with RTX card(s), only to find out current Octane release does not recognize them and i have to wait for a release of next version god knows when...
Well, i dont really want to, hence my question. If the response would be support would come in December, then obviously no buy now and keeping my Pascals until then. But per the other topic, the support should come in September, so i can go on with it, if i decide.fedegarciaf wrote:Why would you buy the cards before they get official support? It makes no sense.Timmaigh! wrote:
I realized i am totally clueless now, when it comes to these different APIs and whatnot. I mean, there was never any doubt i am total layman and have no clue about any details regarding this stuff, but for years it was said that Octane is CUDA-based - whatever that meant, LOL. Now i read Octane is "built" on Optix, DirectX, will be on Vulkan, Metal, etc... what the hell? Does it mean that all these stuff, all these APIs - are equal - to what CUDA is? Or what is the difference?
And in addition to this question, one practical thing of primary interest to me, which i alluded to in one of my previous posts, but nobody cared to answer: say i preorder or simply buy on release one or 2 RTX cards, while simultaneously selling my current Pascal cards.... i read that Turing needs CUDA 10, which has not been released yet... will Octane work for me anyway, cause its OptiX, rather than CUDA based now? Or will i need to wait for the release of some CUDA 10 based Octane version for Turing cards to work at all? Like it used to be in case of Pascal and CUDA 8?
I am not asking now about all those hypothetical speed-ups, i get that is 2019 thing - i just care about basic functionality. I would hate to sell my Pascals, replace them with RTX card(s), only to find out current Octane release does not recognize them and i have to wait for a release of next version god knows when...
