All the crazy crazy things this new GPU will do is truly a huge leap forward.
So the question is will Terragen look toward and type of GPU rendering/assistance to take advantage of this?
I'm sure the rewrite to make it happen is mind blowing, but goddamn, Terragen's glorious lighting model rendering at Warp 9 would be a huge win.
No specific plans for it as-yet, but the more hardware power and software resources are available to support this kind of thing, the more likely we'll be able to do it. So we're certainly getting closer with this release.
- Oshyan
If I'm not mistaken he NVIDA RTX runs exclusively on Vulta technology, and through the Unreal Engine. See my post here: https://planetside.co.uk/forums/index.php/topic,25345.0.html
Still very much in it's infancy and require specific executions functions. Volta with it's new microarchitecture has these capabilities, but I believe is a high-end GPU unit, though supposedly the "future" of all their GPUs
Quote from: WAS 8/21/2018, 2:35:46 PM
>> If I'm not mistaken he NVIDA RTX runs exclusively on Vulta technology, and through the Unreal Engine.
Apparently, NVidia's RTX is now supported by Unity:
https://wccftech.com/unity-collaboration-nvidia-rtx-engine/
I'm starting to get very interested in both.
Quote from: PabloMack on December 24, 2019, 12:54:44 PMQuote from: WAS 8/21/2018, 2:35:46 PM
>> If I'm not mistaken he NVIDA RTX runs exclusively on Vulta technology, and through the Unreal Engine.
Apparently, NVidia's RTX is now supported by Unity:
https://wccftech.com/unity-collaboration-nvidia-rtx-engine/
I'm starting to get very interested in both.
Yeah I saw that.
Having played with a 2080TI (16gb) I can tell you that RTX is a huuuuuuge gimmick. Their "specialized" cores are more conventional than specialized, as when not in RTX you can magically utilize all the card. I already had a hunch about this when AMD released models of their ray tracing without any specialized cores, and just a specialized API. And I can't wait for that API to be mainstream.
When you have RTX enabled, suddenly most the "important" (to visual quality) settings of your engine are grayed out, and defaulted to medium/low settings, making your game really look like crap compared to ultra settings without RTX.
It's good the technology was introduced, but like usual nvidia messed it up. Games that have been utilizing it besides Tomb Raider have been largely flops and revolving around the technology which was a terrible idea from the devs. CONTROL, one of the biggest advertisers of RTX is really a terrible game IMO, and a flash show with heavily exaggerated lighting and shafting to show off the tech.
WAS, with all that said, It seems to me that the need for real time is what forces compromises to be made just to meet target frame rates. The guys who want to make movies, though, don't strictly need quaranteed real-time frame rates. They just want very good render times. Even one frame per second would be really good as compared to packages like TG that often need many minutes or even hours per frame to render many detailed scenes. But movie-makers are a low priority for these tech providers so the controls we need will probably not even be available. When gamers are concerned about frames per second, I am just interested in an average frames per second. The controls we want are more related to guaranteed detail and image quality and we adjust those according to the render times we can live with in order to get it. I'm sure you know all this. I'm just thinking out loud.
Quote from: PabloMack on December 24, 2019, 04:36:29 PMWAS, with all that said, It seems to me that the need for real time is what forces compromises to be made just to meet target frame rates. The guys who want to make movies, though, don't strictly need quaranteed real-time frame rates. They just want very good render times. Even one frame per second would be really good as compared to packages like TG that often need many minutes or even hours per frame to render many detailed scenes. But movie-makers are a low priority for these tech providers so the controls we need will probably not even be available. When gamers are concerned about frames per second, I am just interested in an average frames per second. The controls we want are more related to guaranteed detail and image quality and we adjust those according to the render times we can live with in order to get it. I'm sure you know all this. I'm just thinking out loud.
Good thinking though....
Quote from: PabloMack on December 24, 2019, 04:36:29 PMWAS, with all that said, It seems to me that the need for real time is what forces compromises to be made just to meet target frame rates. The guys who want to make movies, though, don't strictly need quaranteed real-time frame rates. They just want very good render times. Even one frame per second would be really good as compared to packages like TG that often need many minutes or even hours per frame to render many detailed scenes. But movie-makers are a low priority for these tech providers so the controls we need will probably not even be available. When gamers are concerned about frames per second, I am just interested in an average frames per second. The controls we want are more related to guaranteed detail and image quality and we adjust those according to the render times we can live with in order to get it. I'm sure you know all this. I'm just thinking out loud.
RTX technology is exclusively real-time. You can bake on textures etc.and record that live RTX show. It's not part of any renderer, and would immediately compromise quality. They use much better systems like cycles, path tracing in TG, etc.
Quote from: WAS on December 24, 2019, 07:51:48 PMRTX technology is exclusively real-time. You can bake on textures etc.and record that live RTX show. It's not part of any renderer, and would immediately compromise quality. They use much better systems like cycles, path tracing in TG, etc.
WAS,
It's good to know that. I appreciate the information and insight. Looks like we are stuck with CPU multi-core for the foreseeable future.
Actually Vray and other renderers are taking advantage of RTX cores already: https://www.chaosgroup.com/blog/v-ray-gpu-adds-support-for-nvidia-rtx
Supposedly in Cycles too: https://code.blender.org/2019/07/accelerating-cycles-using-nvidia-rtx/
- Oshyan
Quote from: Oshyan on December 25, 2019, 01:52:06 PMActually Vray and other renderers are taking advantage of RTX cores already: https://www.chaosgroup.com/blog/v-ray-gpu-adds-support-for-nvidia-rtx
Supposedly in Cycles too: https://code.blender.org/2019/07/accelerating-cycles-using-nvidia-rtx/
- Oshyan
Granted both these use the API for a specific function of the API for calculation. Nothing is being rendered with RTX. Seems they are using the cores to calculate structure and intersection points. Not much different than GPU acceleration, just using the OptiX API and "RT" cores which as I mentioned seem more conventional than specialized, as they operate as generic GPU cores through any other API.
For NVIDIA workstation cards being premier it makes sense to take advantage of the API for specific things that would be implemented 3rd party in other APIs like Cycles in the past.
For actual RTX, you'd be recording live. It's Inherently Real-Time Raytracing. That's what the API is truly for, and why Cycles or Vray aren't actually using RTX, just the API for specialized math. Which is no different than using a GPU with CPU in end goal. Faster computation.
I'm not exactly clear what you mean when you say "RTX" then. You mean the specific RTX functionality as accessed through some particular API (e.g. DirectX)? In other words strictly how it is implemented in games?
- Oshyan
Quote from: Oshyan on December 25, 2019, 03:17:36 PMI'm not exactly clear what you mean when you say "RTX" then. You mean the specific RTX functionality as accessed through some particular API (e.g. DirectX)? In other words strictly how it is implemented in games?
- Oshyan
I edited my post to make that clear. RTX is Real-Time Raytracing. The API is using the GPU to render. What's being done here is using the API to calculate for specific end goal. Which itself isn't RTX.
This is just using OptiX API and doing some speedy calcs for their cycles/Vray to accelerate rendering.
It seems the benefits here is software based calcs done on CPU vs hardware based calcs on GPU, meant to do them on RT cores.
To put it simply Cycles is still the renderer, and rendering, Vray is still the renderer, and rendering. TG's Path tracer or w/e would still be the renderer and renderering. It wouldn't be RTX, which is hardware, real-time rendering, on the GPU. That would be a whole new renderer.
Arguably then you have a very specific/individual definition of "RTX":
https://developer.nvidia.com/rtx
- Oshyan
Quote from: Oshyan on December 25, 2019, 03:42:59 PMArguably then you have a very specific/individual definition of "RTX":
https://developer.nvidia.com/rtx
- Oshyan
You seem to be completely ignoring the whole "real-time" and rendering aspects where they are rendering on the Turing GPU, in real-time, or performing real-time AI. Nothing you posted previous is RTX... It's using the OptiX API to make some very specific calls, as needed. There is no actual RTX employed beyond a brand name GPU name. No real Real-Time Raytracing or AI computation, etc, outputted via Hardware.
Effectively Cycles and Vray are just using it as a calculator, none of it real-time RTX.
Did you look at the page I linked? Nvidia themselves define"RTX" as a whole platform which even includes AI. The specific component you're talking about is of course ray tracing, but even Nvidia don't differentiate between realtime and non-realtime uses:
QuoteRay Tracing (OptiX, Microsoft DXR, Vulkan)
From: https://developer.nvidia.com/rtx
- Oshyan
Quote from: Oshyan on December 25, 2019, 04:14:57 PMDid you look at the page I linked? Nvidia themselves define"RTX" as a whole platform which even includes AI. The specific component you're talking about is of course ray tracing, but even Nvidia don't differentiate between realtime and non-realtime uses:
Quote from: undefinedRay Tracing (OptiX, Microsoft DXR, Vulkan)
From: https://developer.nvidia.com/rtx
- Oshyan
I don't think you're understanding the pages technicalities. "The platform" specifies use of "real time" denotation to those features, on Turing GPUs (output). "real time ray tracing"
There is NO rendering happening with the RTX platform. The OptiX API is used for calculations. The whole inherent application of the technology is real time steaming on turing GPUs. AI for example, like deep neural networks, are suppose to be real time.
Again The RT in RTX stands for "Real Time". The X for Ray Tracing (it's main rendering goal). Actually I guess it's "Ray Tracing" (The RT), X isn't specified. I guess Technology similar to GTX.
Additionally, RTX Platform, refers to the GPU. RTX itself is OptiX, Microsoft DXR, Vulkan. NGX already existed, same for PhyX, etc. Those were already part of GPUs (like GTX cards [the GTX platform]) and not "RTX"
I'm not sure if I should chime into this discussion after being called "ignorant of TG" by you before little time ago in another topic, but I feel I need to back Oshyan in saying that VRay utilizes RTX now and what he means by saying that.
It simply means that VRay utilized GPU capabilities before, but was not utilizing the new tensor cores which made it into the mainstream market and is what actually defines the RTX product line.
No tensor cores is GTX and with tensor cores is RTX.
So VRay utilizes the RTX Tensor cores now and they actual have all sorts of benefits in relation to the type of calculations useful in raytracing scenarios, but also AI.
It's interesting cherrypicking and contradicting that you refer to RTX as also being an AI platform *because* of these tensor cores, but dismiss its usefulness for ray tracing and such and call it a "feature" when in ray tracing context.
From there the discussion became a language and blaming the ignorant one type of discussion I just referred to... AGAIN.
I don't know what happened in your offline life recently, but I think not many here enjoy the way you treat (other) people here when they oppose or even mildly question the tiniest bit of your assumptions/opinion/logic.
Quote from: Tangled-Universe on December 28, 2019, 04:17:01 PMI'm not sure if I should chime into this discussion after being called "ignorant of TG" by you before little time ago in another topic, but I feel I need to back Oshyan in saying that VRay utilizes RTX now and what he means by saying that.
It simply means that VRay utilized GPU capabilities before, but was not utilizing the new tensor cores which made it into the mainstream market and is what actually defines the RTX product line.
No tensor cores is GTX and with tensor cores is RTX.
So VRay utilizes the RTX Tensor cores now and they actual have all sorts of benefits in relation to the type of calculations useful in raytracing scenarios, but also AI.
It's interesting cherrypicking and contradicting that you refer to RTX as also being an AI platform *because* of these tensor cores, but dismiss its usefulness for ray tracing and such and call it a "feature" when in ray tracing context.
From there the discussion became a language and blaming the ignorant one type of discussion I just referred to... AGAIN.
I don't know what happened in your offline life recently, but I think not many here enjoy the way you treat (other) people here when they oppose or even mildly question the tiniest bit of your assumptions/opinion/logic.
Well, you're wrong, though. Because they are using the OpitX API to run calculations on RT Cores. They aren't using Tensor Cores. As they specifically state, this is for "RTX Class" GPUs. Not Volta GPUs... Tesla Volta GPUs have Tensor Cores... Not sure where you're getting your information from. Both Vray and Cycles are using the OptiX API for RT core calculation, two little functions to run structure and intersection calculations that RTX can use in real-time, just much higher fidelity settings in non-real time. However, again, as even NVIDIA specifies, RTX is real-time. The "RTX Platform" are "RTX GPUs" (which aren't Volta GPUs; AKA the Volta Platform...)...
You'll see in the future here, neither Vray, or Cycles being used in RTX demonstrations. Because it's not RTX. They're just mining a tiny bit of data that RT cores can run faster on the "RTX Platform" via the OptiX API. Nothing streaming on the GPU and outputting on the GPU.
In conclusion, cycles is transparent in its use, and Vray calls RTX "support". In general, just using any API for some features most certainly doesn't make your engine/system/website/whatever "It" lol. I'd say the API would have to be handling it all, and output.
Quote from: WAS on December 25, 2019, 03:20:44 PMQuote from: Oshyan on December 25, 2019, 03:17:36 PMI edited my post to make that clear. RTX is Real-Time Raytracing. The API is using the GPU to render. What's being done here is using the API to calculate for specific end goal. Which itself isn't RTX.
This is just using OptiX API and doing some speedy calcs for their cycles/Vray to accelerate rendering.
RTX = Ray Tracing eXtentions. It's not limited to realtime, or to gaming. Optix is using the hardware to enhance ray tracing performance.
There's a reason that even though AMD implemented ray tracing using its general purpose compute, it's developing hardware ray tracing for itself as well as for the next generation PlayStation and XBox. And the speedup from using Optix in Cycles has been more than double.
nVidia isn't exactly new to 3D rendering; it's been using general purpose compute to accelerate ray tracing for years with IRay, and the RTX cores grew out of that.
Quote from: WAS on December 28, 2019, 06:25:35 PMWell, you're wrong, though. Because they are using the OpitX API to run calculations on RT Cores. They aren't using Tensor Cores. As they specifically state, this is for "RTX Class" GPUs. Not Volta GPUs...
Tesla Volta GPUs have Tensor Cores... Not sure where you're getting your information from. Both Vray and Cycles are using the OptiX API for RT core calculation, two little functions to run structure and intersection calculations that RTX can use in real-time, just much higher fidelity settings in non-real time. However, again, as even NVIDIA specifies, RTX is real-time. The "RTX Platform" are "RTX GPUs" (which aren't Volta GPUs; AKA the Volta Platform...)...
You'll see in the future here, neither Vray, or Cycles being used in RTX demonstrations. Because it's not RTX. They're just mining a tiny bit of data that RT cores can run faster on the "RTX Platform" via the OptiX API. Nothing streaming on the GPU and outputting on the GPU.
In conclusion, cycles is transparent in its use, and Vray calls RTX "support". In general, just using any API for some features most certainly doesn't make your engine/system/website/whatever "It" lol. I'd say the API would have to be handling it all, and output.
Not only are you wrong, but you're also contradicting yourself. The RTX cores are what Optix uses to speed up its ray tracing and BVH traversal operations, because the hardware is customized for them, rather than being repurposed general purpose computing hardware like AMD is using to tide itself over until it has ray tracing hardware.
Blender makes the tensor cores available via an optional denoise shader node, because it's not strictly necessary; instead it's there so that the artist using the renderer can use the denoiser to reduce the number of rays required to achieve a quality render. In other words, it's there as a time saving option.
Your definition of what qualifies as taking advantage of RTX cores is nonsensical, and obviously exists just to justify your claim that hardware ray tracing extensions can only be used in real time applications, no matter how false.
The way Optix is designed is basically to enable renderer developers to offload the ray tracing calculations to dedicated hardware without having to specifically code to that hardware. The point of that is so that nVidia's developers take care of the under the hood work of identifying how many GPUs with RTX hardware are available and how many RTX cores each has, so that a renderer like Cycles, Vray (or KeyShot, etc) can scale up with more RTX GPUs without the renderer's developers having to handle that part on their own. Hence when nVidia launches Ampere, the same renderers will be able to transparently get a speed boost without having to change their own code.
Quote from: Tamerlin on December 30, 2019, 04:11:24 PMQuote from: WAS on December 25, 2019, 03:20:44 PMQuote from: Oshyan on December 25, 2019, 03:17:36 PMI edited my post to make that clear. RTX is Real-Time Raytracing. The API is using the GPU to render. What's being done here is using the API to calculate for specific end goal. Which itself isn't RTX.
This is just using OptiX API and doing some speedy calcs for their cycles/Vray to accelerate rendering.
RTX = Ray Tracing eXtentions. It's not limited to realtime, or to gaming. Optix is using the hardware to enhance ray tracing performance.
There's a reason that even though AMD implemented ray tracing using its general purpose compute, it's developing hardware ray tracing for itself as well as for the next generation PlayStation and XBox. And the speedup from using Optix in Cycles has been more than double.
nVidia isn't exactly new to 3D rendering; it's been using general purpose compute to accelerate ray tracing for years with IRay, and the RTX cores grew out of that.
OpiX is the API to use the platform (the GPU, like accessing RT Cores; otherwise you're using just going to be using it like any other legaxy GPU with another API). It's not accelerating anything... and the X most certainly doesn't mean "extensions" as there are no "extensions". It's a new platform and not an extension of one.
You're shooting yourself in the foot out the gate bud, and does nothing for your credence.
You should at least try to research what you talk about. Your shenanigans about OptiX are hilarious.
Quote from: WAS on December 30, 2019, 06:00:20 PMOpiX is the API to use the platform (the GPU, like accessing RT Cores; otherwise you're using just going to be using it like any other legaxy GPU with another API). It's not accelerating anything... and the X most certainly doesn't mean "extensions" as there are no "extensions". It's a new platform and not an extension of one.
You're shooting yourself in the foot out the gate bud, and does nothing for your credence.
You should at least try to research what you talk about. Your shenanigans about OptiX are hilarious.
I did the research, and you're wrong. Anyone with a few minutes can see that just by looking at the overview in the nVidia developer documentation; it doesn't even require any programming knowledge to recognize this.
Your flat earther style debating tactics won't get you anywhere except with people just as ignorant as you.
Quote from: Tamerlin on January 02, 2020, 02:44:31 AMQuote from: WAS on December 30, 2019, 06:00:20 PMOpiX is the API to use the platform (the GPU, like accessing RT Cores; otherwise you're using just going to be using it like any other legaxy GPU with another API). It's not accelerating anything... and the X most certainly doesn't mean "extensions" as there are no "extensions". It's a new platform and not an extension of one.
You're shooting yourself in the foot out the gate bud, and does nothing for your credence.
You should at least try to research what you talk about. Your shenanigans about OptiX are hilarious.
I did the research, and you're wrong. Anyone with a few minutes can see that just by looking at the overview in the nVidia developer documentation; it doesn't even require any programming knowledge to recognize this.
Your flat earther style debating tactics won't get you anywhere except with people just as ignorant as you.
Their own documentation is so clear they are careful with denotations about real time. Sorry you can't see this. Any modern GPU can do ray tracing, RTX has dedicated hardware for streaming this in real time. That's why Cycles is very clearly transparent on its use of RTX as a means to accelerate intersections and structure, because these things on CPU or GPU through 3rd party software would be slower. This is where RTX is only accelerating a simple function of Cycles, their own renderer. It's not actual RTX. The fact you can't comprehend this, and how APIs being used third party act within your software. So maybe you should be a developer before making a comment. 😂🤣
Also weird you mention new Xbox and PlayStation when Xbox is going to be using AMD, and PlayStation may likely do the same. Price vs performance and all.
If they were creating whole new renderers based on OptiX and RTX platform, it would be RTX. But no, it's simply aided, in a veeeerry small aspect towards the goal of the render.
Again, by inherent definition of it's use with Cycles, etc, it's being used as a calculator. Raw numbers. Takes awhile to process intersection points and structure. So speeding that up is fantastic... But no, the GPU isn't rendering anything with RT cores, or even on main GPU cores where it's using legacy methods for GPU render acceleration.
A great example of this is Terragen using RTX cards and their RT cores for intersection calculation, etc, it will still be Matt's renderer doing the work, thus it's not RTX. The OptiX API is being used to calculate, not render anything. Those raw calculations would than be used by Terragen's renderer, which is not RTX... In the immediate past, GPUs were already doing this. Now they're just making use of hardware specific tasks on the GPU where before it was just software-based, calculated on the GPU.
So Terragen's Renderer, it's own thing, is aided by RTX platforms, but is not RTX, and even would obviosuly have fall-backs for no RTX cards (like Cycles).