Here’s the link to the thread:
http://www.fstormrender.com/forum/forum/off-topic/1703-andrey-kozlov-vs-otoy
Funny to see that Otoy is claiming that the 3ds max Environment Tab belongs to their product - the really have no idea what they’re doing.
I don’t think it’s a good idea to publish everything in a public forum though. Hope this turns out well for Andrey (Karba), he’s a good guy.
Yeah, what the f*ck is this ?!
Btw, what happened to the company founder (Refractive, not Otoy) ?
Regarding F-storm.
What I like:
Karba is fast. You ask him, you get the feature next-day.
They have good material, after it was requested by users. Now they have correct fresnel and even glossiness attentuation towards edges. Very cool.
Cool glare and tonemapping. Good to see this being priority :- )
What is OK.
It’s not fast, c’mon, but not slow either. Pretty much like Octane, the interior GI is still slow. Sometimes a user will post “2 minute !!!” render only to be followed that actual, production quality 3k+ interiors are still 1 hour + even with 4x Titans. But Titans are still cheaper and more stackable than CPUs :- )
What I dislike:
Every properietary crap. No support for 3dsMax instances ? SERIOUSLY ?
No normal map support, or any plan to. SERIOUSLY ?!?! Big NO.
No support for 3dsMax physical camera but new F-Storm camera ? Nobody wants no f-storm camera.
So it’s pretty damn good in some areas, and pretty bad for any actual stable production on others. Will watch cautiously.
Have to agree with almost everything, Juraj - GPU rendering seemed so promising at the beginning. But you’re very quick back to the ground of reality with more complex interiors taking quite some time even with a serious hardware setup. It’s not the holy grail everyone was wating for. Also Corona is getting better each day 
Otoy did a great job in screwing Octane completely up; features nobody ever requested, confusing marketing (thinking about the V3 presentation video…omg), wired decisions.
Releasing new versions while breaking stuff that has worked before; the 3ds max plugin is filled with test releases and unstable versions.
They have hired some good guys at the forums though, professionally playing along customers, delaying promised updates/features - they’re doing a remarkable job with this.
I think GPU rendering could have been much further if developed the right way…
You are also right about the cons for FStorm but please consider that it’s a very young engine; first release was only six months ago and it’s still in alpha stage. So hopefully all this will be added at some point.
Adaptive sampling is great feature though, cutting down render times quite a bit. I hope Andrey will add an integrated noise reduction solution one day - assuming everything works out good for him with the lawsuit.
When I realize there is no max maps support (procedurals, bitmaptexture etc) it’s usually the last time I try the renderer. It’s a no-go for production for me and at this point I can only regard it as a (o)toy.
No normal map support is ridiculous though, can’t imagine he’s not planning to implement it.
I must admit that I never really understood these religious CPU vs GPU debates. As a hardware designer, I am looking forward to the upcoming Xeons combined with FPGAs for the ability to design custom hardware 
But to be serious, I think it would be better for everyone to combine the best of both worlds. Use CPU for more complex tasks and then use GPU for stuff like postprocessing, where you apply a relatively simple operation to lots of data.
People who do those cpu vs gpu wars often don’t understand quite well what they are actually debating, it’s just another variation of being part of cool camp. But ‘religious’ is overstatement, have a look at PC vs Console, nVidia vs AMD type of debates :- ) Those are beyond any kind of salvation.
c’mon guys, Fstorm render is now still a cute baby, but I heard it developed so fast then I’m sure it will become a beast in the future not just a toy.
hi Ryuu, Can you make hardware acceleration for corona denoising just like the “other” do? 4k render will be happy for this acceleration (and especially me who don’t have hi-end cpu). or I didn’t realize it is already use GPU acceleration…?
cheers
Instancing works, you have to select it in the fstorm properties of the instance to work.
So things like Forrest pro work.
Things I agree with
No Normal map support, that sucks
Fstorm Camera, yeah he should support Physical camera
With enough convincing I’m sure he’ll put them in
I’ll do some more testing but I’m pretty sure I’ve had decent results within 2to3hrs on one GTX 970 at 3k.
Does Corona use gpu’s at all ?
I think it’s such a waste to have powerful gpu’s in computer and to not use them at all.
But I understand the argument that you lose the flexibility with GPU and CPU is best used for calculating light transport and complex shaders and all.
But GPU not even used for post processing and denoising ? Sure a gpu can be used efficiently to calculate glare, blurs and post process ? And maybe denoising. Either alone or in combination with the CPU for an even faster speed. Or maybe calculating post processing in parallel with the CPU ? so you don’t have to wait a few more seconds for post processing after the render is finished.
Altus Denoiser makes use of GPU’s.
https://www.innobright.com/2016/02/29/innobright-altus-1-3-now-available/
It is also waste not to use better sampling techniques, not having optimized SSS/skin/hair shader, not having multilight, etc… it is all about priorities
I definitely can, but the key questions are:
Besides those questions, there is another issue with the fact that bloom&glare + denoising now take only a small fraction of the whole rendering time. Imagine you have a 30 minute render, where the denoising takes only the last 2 minutes. Even if we optimize the denoising infinitely (so that it is calculated in zero time), it will actually result in a 28 minute total render time. Amdahl’s law is a bitch
Not at this moment. The system definitely uses GPU for displaying the VFB, but that is outside of Corona.
I definitely feel your pain. I always buy the high end GPUs + CPUs in the hopes that I will have time to play with some interesting optimizations, only to watch them gather dust for a few years before buying new ones
I don’t think this is doable. To be able to calculate the post processing effects, you already need to have the whole rendered image. So even when using GPU, you can start the post processing only after the CPU has finished rendering (+ some additional time to transfer data from CPU to GPU). It is certainly possible to have GPU compute post processing on one frame and have the CPU start rendering the next frame in parallel, but that would be a nightmare to implement & debug + it would probably consume significantly more memory.
So exactly what I said, it doesn’t support the native instancing directly ;- ). In all the years of Vray usage, I almost never clicked on the VrayProperties of object. I can’t imagine having the need to do so on top.
that questions kicks my smaller-brain enough, I have no any idea about that people here are always dreaming until zero rendertime comes true
btw, the corona trello roadmap which contains kickass features I loved is still in hi priority.. keep rocking ryuu!
I meant in parallel like octane calculates glare and other things during rendering and not at the end.
I just tested FStorm to compare glare and to see if it has the lens distortion capability of Octane and it looks like Octane is much better in this regard. It seems that at least these parts were not programmed by the FStorm developer in Octane.
Compared to I think every other renderer, glare speed and looks in octane is amazing. I don’t know how they did it. While changing the glare settings during rendering it’s so fast it seems like adjusting a layer in photoshop. An already rendered layer. While in FStorm is much slower.
I remember talks in the past on Adobe forum and how some would say GPU’s are for gaming and there’s no mixing that with a 2d editor, and then Adobe added GPU support to Premiere, and it made such a huge difference. Not only many effects like color correction and blurs become real-time, but you were able to playback in real-time (24 fps) not just one layer of full hd movie, but several. And with added gaussian blur and color correction on top, without needing to render first.
I used to think that the reason a program like Premiere can’t playback high res movies in realtime because a spinning hdd was not fast enough for that, to move all that data. Seems not.
It began to be interesting.I couldn’t believe when I saw this game video I thought it was real…
SW:B maps make me cry evertim.
Such quality… I love where are in real-time. Exciting times.
In all honesty, and as stupid as this may sound, I am holding off investing into anything OTOY related because of crap like this. Voting with my wallet because my livelihood doesn’t depend on it.
To be perfectly frank, the C4D Octane integration seems exemplary, really really good but since I can afford to stay on the CPU I will wait for the dust to settle and see how things progress. I do need to reiterate here, Octane C4D seems superbly well done!
Other than that, I think GPUs have a good future ahead of them, the generational leaps we are making so far are, as always, impressive!
juraj you have experience with realtime.Can you do such thing with bake or using unreal engine would it take a lot of time ?
Yeah, that can definitely be done in Unreal 4. Frostbite engine isn’t probably too different, although they have bigger graphic research team, similar to Crytek’s one. Check out their Siggraph and GDC papers on Frostbite 3 PBR, it’s incredible where their engine is graphically and the amount of work they put into it.
I am pretty sure they use Enlighten GI on all their games ( SW:B, Mirror’s Edge2, DG:I,etc..), which while branded as real-time, is still a precomputed bake.
This kind of quality takes incredible amount of time regardless of platform or engine.