Poll

3 features you want the most:

PBR Style material (Disney, Unreal Engine, etc..)
78 (9.8%)
Refraction/Reflection working with masking render elements (CMasking_Mask,CTexmap, etc...)
49 (6.2%)
Render-time booleans (cut/slice objects)
42 (5.3%)
Dedicated CarPaint Shader
13 (1.6%)
Memory optimization (lower memory usage in 3ds Max; not out-of-core)
59 (7.4%)
GPU/Hybrid rendering
108 (13.6%)
Displacement improvements (2D displacement, autobump, memory efficiency, ...)
53 (6.7%)
Speed improvements
83 (10.4%)
Cryptomatte
24 (3%)
Geopattern
92 (11.6%)
Toon shader
31 (3.9%)
Reworking tone mapping (DSLR-style tonemapping)
96 (12.1%)
Interactive rendering in 3ds max viewport (with gizmos, object selection, manipulation, ...)
29 (3.6%)
Adding own materials to Corona Material Library
39 (4.9%)

Total Members Voted: 331

Author Topic: The most wanted feature?  (Read 198175 times)

2018-09-14, 12:37:51
Reply #510

jpjapers

  • Active Users
  • **
  • Posts: 1099
    • View Profile
I'm so happy with Corona, but the following three features would make it unbeatable:

1) GPU/Hybrid - I'm now starting to feel that if Corona isn't announcing plans for the future here I'm getting worried. I'm due to building a new machine next year or so and spending my hard earned money on GPUs seems to be a better bet than multi core CPUs

2) Completely automatic import of Vray models and scenes. Still ways to go.

3) Fstorm LUTs - not sure what they do with their color mapping, but I want it!  ;)

Completely agree with the GPU comment

2018-09-16, 18:03:05
Reply #511

Juraj Talcik

  • Active Users
  • **
  • Posts: 3435
  • Tinkering away
    • View Profile
    • studio website
I would not expect the devs to jump on the GPU vagon now, esp. since they never wanted it, and now Vray is focusing more and more on it.

They did promise at least off-loading the framebuffer&post onto it in some way, didn't materialize yet in any way either. But Optix is kick-ass ;- ).

Would be nice though, let's see how fast Vray is going to be once RTX embargo is out and Vlado can disclose some benchmarks.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik    My Behance portfolio
lysfaere.com   Something new

2018-09-17, 11:24:50
Reply #512

jpjapers

  • Active Users
  • **
  • Posts: 1099
    • View Profile
I would not expect the devs to jump on the GPU vagon now, esp. since they never wanted it, and now Vray is focusing more and more on it.

They did promise at least off-loading the framebuffer&post onto it in some way, didn't materialize yet in any way either. But Optix is kick-ass ;- ).

Would be nice though, let's see how fast Vray is going to be once RTX embargo is out and Vlado can disclose some benchmarks.

Its true that they never wanted it. But lets be honest, Its becoming more and more apparent that its the cheaper way of operating and scaling your operations.
If i wanted a new render node id have to shell out a few thousand for a dual xeon node. Or i can pay around a thousand for a new gpu and pay less in terms of power draw too.
I just think with fstorm becoming so popular since they won the lawsuit and large chunks of free models being released exclusively for fstorm and with vray showing more gpu rendering is it not a worry that eventually the cpu renderers would get left behind just because the cost of setting up a gpu farm can be alot less?

Nvidia are really keeping their cards tight to their chest with the embargo on RTX but i kind of agree with what some people are saying that this could be like PhysX all over again where it never really became an integral part of the gaming experience. However it would be interesting to know what impact tensor cores could have on rendering applications if any.

2018-09-17, 12:29:19
Reply #513

Juraj Talcik

  • Active Users
  • **
  • Posts: 3435
  • Tinkering away
    • View Profile
    • studio website
Yup, GPUs definitely have better future in rendering and it's much easier to scale it.

But then again...look at what nVidia introduced... massive overpricing. To get the most out of GPU rendering, you would need Quad (two pairs of NV-Link to get memory pooling) 2080ti= 5200 Euros + price of workstation.
So majority of people will not be able to afford it either it's just that at certain high-price point, GPU overcomes CPU in price/effectivity a lot.

Truth be told I thought cpu rendering could be dead by now...but imho 32-core 2990WX is much better progress than 2080ti itself. In next two years, I fully expect 48-64core 3rd/4th gen 3990/4990WX getting another massive 100perc. speed boost while I only expect 7nm nVidias to cost even more for even less speed-up. So the competition is strangely equal right now and people should stick with what they have right now and see how it unfolds.

But since most of us have at least one GPU in each PC, it's shame we can't use it to anything.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik    My Behance portfolio
lysfaere.com   Something new

2018-09-17, 13:02:01
Reply #514

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 8860
  • Turning coffee to features since 2009
    • View Profile
If you read our announcements carefully, you can see we are getting more and more open about the possibility. If you look retrospectively at GPUs and technology available 8 years ago, you will surely agree that doing CPU rendering then would be unfeasible. Yet even then people claimed CPU rendering is dead and GPU future is right behind corner.

Since then GPU rendering got better and better, but with smaller, incremental steps, there is no clear "switch now" moment. I am lot less nervous about this now since we have large GPU team inside Chaosgroup that I can consult on the feasibility and how-to of GPU rendering. This might be the biggest benefit of RL-CG merger for the future.
Rendering is magic.
Private scene uploader | How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2018-09-17, 14:19:17
Reply #515

jpjapers

  • Active Users
  • **
  • Posts: 1099
    • View Profile
If you read our announcements carefully, you can see we are getting more and more open about the possibility. If you look retrospectively at GPUs and technology available 8 years ago, you will surely agree that doing CPU rendering then would be unfeasible. Yet even then people claimed CPU rendering is dead and GPU future is right behind corner.

Since then GPU rendering got better and better, but with smaller, incremental steps, there is no clear "switch now" moment. I am lot less nervous about this now since we have large GPU team inside Chaosgroup that I can consult on the feasibility and how-to of GPU rendering. This might be the biggest benefit of RL-CG merger for the future.

Interesting. Do you guys feel there would ever be that switch moment?
Perhaps when GPU prices settle back down to where they actually should be rather than a good 30% above what the RRP should be?

2018-09-17, 16:19:53
Reply #516

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 8667
  • Marcin
    • View Profile
Perhaps when GPU prices settle back down to where they actually should be rather than a good 30% above what the RRP should be?
I don't think this is going to happen in this millennium, considering how GPU rendering/ai/cryptocurrencies are getting more and more popular.
Same with RAM.

2018-09-17, 18:02:08
Reply #517

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 8860
  • Turning coffee to features since 2009
    • View Profile
If you read our announcements carefully, you can see we are getting more and more open about the possibility. If you look retrospectively at GPUs and technology available 8 years ago, you will surely agree that doing CPU rendering then would be unfeasible. Yet even then people claimed CPU rendering is dead and GPU future is right behind corner.

Since then GPU rendering got better and better, but with smaller, incremental steps, there is no clear "switch now" moment. I am lot less nervous about this now since we have large GPU team inside Chaosgroup that I can consult on the feasibility and how-to of GPU rendering. This might be the biggest benefit of RL-CG merger for the future.

Interesting. Do you guys feel there would ever be that switch moment?
Perhaps when GPU prices settle back down to where they actually should be rather than a good 30% above what the RRP should be?

30% is not enough to switch, it would have to be lets say 10x. Or solving some outstanding issues of GPU programming, such as lack of function pointers, bad scaling with complexity/divergent code, etc. Or just going step by step. We are in contact with VRay GPU team and we are discussing this from time to time.
Rendering is magic.
Private scene uploader | How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2018-09-17, 18:05:01
Reply #518

Juraj Talcik

  • Active Users
  • **
  • Posts: 3435
  • Tinkering away
    • View Profile
    • studio website
10 times what :- ) ?
talcikdemovicova.com  Website and blog
be.net/jurajtalcik    My Behance portfolio
lysfaere.com   Something new

2018-09-17, 18:06:11
Reply #519

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 8860
  • Turning coffee to features since 2009
    • View Profile
10x better performance per $ or watt (so you can pick between GPUs getting 10x faster or 10x cheaper ;))
Rendering is magic.
Private scene uploader | How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2018-09-17, 20:46:26
Reply #520

Benny

  • Active Users
  • **
  • Posts: 176
    • View Profile
What is the status of Optix support in Corona?

2018-09-17, 20:48:26
Reply #521

sebastian___

  • Active Users
  • **
  • Posts: 129
    • View Profile
When/if Corona would add GPU rendering - it would be ideal if it would be a seamless experience. I mean just add another CPU and another GPU and watch your rendering gets faster and faster. No Corona split in two like Corona CPU and Corona GPU.

 And if the GPU would be hard or unfeasible to program in a way CPU is, then the GPU should be used at least for the operations in which GPUs excel. So some things would be done by the CPU and simultaneously the GPU would start doing those other things which comes easy for a GPU. And I don't mean just the post effects, but maybe some parts of the shading or lighting or I don;t know, cause I have no idea :)

2018-09-17, 22:10:14
Reply #522

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 8860
  • Turning coffee to features since 2009
    • View Profile
That is another topic. There is not a single renderer on the market that sucessfully switched from CPU to GPU rendering. There are probably some reasons behind this. One might be that you not be able to get full feature set of a CPU renderer on GPU. At this point I think it would then make more sense to remove the same features from CPU version than keep 2 renderers with 2 separate feature lists. And I think most people would still not like it...
Rendering is magic.
Private scene uploader | How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2018-09-18, 02:57:53
Reply #523

Benny

  • Active Users
  • **
  • Posts: 176
    • View Profile
But isn't it possible to have a supplementary strategy, like Optix for noise or just having the GPU doing some other taxing task?

2018-09-18, 09:57:44
Reply #524

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 8860
  • Turning coffee to features since 2009
    • View Profile
that was also never successfully done. The overhead and synchronization just makes it not worth it. You would have to transfer tens of gigabytes of data per second between CPU and GPU to make this work, and you would be limited by the slower component of the two
Rendering is magic.
Private scene uploader | How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)