Community
VRED Forum
Welcome to Autodesk’s VRED Forums. Share your knowledge, ask questions, and explore popular VRED topics.
cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

RTX Support

33 REPLIES 33
SOLVED
Reply
Message 1 of 34
Neil_Cross
9702 Views, 33 Replies

RTX Support

Nvidia recently announced their new Quadro and GeForce RTX cards which ray tracing and tensor cores, the major thing being real time ray tracing support on the GPU. 

 

If you're able to comment, do you guys have any plans to implement this into Vred, it could potentially be a game changer for visualisation software? If so, any idea on a timeline for support?

33 REPLIES 33
Message 21 of 34
seiferp
in reply to: Neil_Cross

That guy made a video with his mobile from the RTX GPU demo at the Autodesk University Nvidia booth. Quality isn’t brilliant but shows where it’s going.

 

https://youtu.be/5hbOeRRKuVE

Message 22 of 34
eid-o
in reply to: seiferp

I am not quite sure if this is the best thread to ask this question, but since you're discussing the raytracing mode running on two GPUs I thought it might be a good idea to ask...

Any idea if the future VRED releases will be able to render using the GPU/GPUs? It's been years since all other major renderers are utilising the power of multiple GPUs with CUDA cores and yet VRED is stuck to CPU only? The only way to accellerate that is by adding render nodes, which is very expensive and just requres more and more machines while the only thing you need from the extra nodes is the CPU processing power really. So it's quite an inefficient way to increase productivity. Instead it would be much better just to stick a few more GPUs in your current machine.....

So any idea on that?

Message 23 of 34
michael_nikelsky
in reply to: eid-o

This whole thread is about GPU rendering, RTX is the name of the latest NVIDIA GPUs and we have shown a prototype running on it. I think that should answer you question already.

The reason why we did not do this earlier is simply because the GPUs did not have the memory we need and the performance benefit you got from a GPU did not really exist compared to a dual CPU workstation. 



Michael Nikelsky
Sr. Principal Engineer
Message 24 of 34
eid-o
in reply to: michael_nikelsky

Thanks for the reply Michael, much appreciated!

In this case shall I assume that it's better to wait and invest in adding more GPUs (currently running on Quadro) as opposed to buying more computer nodes? I understand you can't reveal the full details on the current developments, but it seems this way..

Message 25 of 34
michael_nikelsky
in reply to: eid-o

Hi,

 

there is no time frame to when we will release the GPU based raytracer yet. I would not assume it to be on feature parity with the CPU tracer as well since some things just don´t work efficiently on a GPU. 

In any case we will continue to develop and improve the CPU raytracer as well since there are just use cases the GPU can not handle and the number of cores in a CPU is (finally!) rising again with 64 core / 128 thread CPUs on the horizon.

 

Kind regards

Michael

 

 

 

 



Michael Nikelsky
Sr. Principal Engineer
Message 26 of 34

Hi guys,

does the 2020 support RTX cards?

 

I have the opportunity to test 2 quadro RTX 5000 with my HP partner.

 

Thanks.

Chris

Christian Garimberti
Technical Manager and Visualization Enthusiast
Qs Informatica S.r.l. | Qs Infor S.r.l. | My Website
Facebook | Instagram | Youtube | LinkedIn

EESignature

Message 27 of 34

@Christian_Garimberti in OGL realtime you mean? Yes RTX cards do work. I have 2x RTX 6000 in SLI Mode and did not run into any problems yet.

The RTX GPU Raytrace demos you might have seen is still just a Tech. Preview and not released in VRED 2020.

Message 28 of 34

Thank You @seiferp .

In OGL i know it is ok. I tested some of my samples and i noticed a 10 to 15% performance more compared to an "old" Quadro P5000 on the same workstation.

For the RTX GPU Raytrace i will wait...

 

Chris

Christian Garimberti
Technical Manager and Visualization Enthusiast
Qs Informatica S.r.l. | Qs Infor S.r.l. | My Website
Facebook | Instagram | Youtube | LinkedIn

EESignature

Message 29 of 34

will it work as a common OGL visualization or will it integrate, for example, raytraced ambient occlusion in lieu of screen space AO?

_________________________________________________________

almost a showcase PRO user 😛

if you find my post useful, please give kudos or mark as solution 😄
Message 30 of 34
michael_nikelsky
in reply to: cikho

It will, at least in the first iteration, not be hybrid rendering. It will be an option for the raytracer to use the GPU instead of the CPU. We might add hybrid features later but it is currently not in development.



Michael Nikelsky
Sr. Principal Engineer
Message 31 of 34
itsupport
in reply to: Neil_Cross

Anyone have experience (success or otherwise)  using an Nvidia RTX 5000 card in a laptop (in my case, it will be a Dell Precision 7740) ?

At this stage, the mobile version of this card is not on the Inventor Certified web site. Just wondering if there is a problem or just time before it is added to the list (the card was released in May)

 

Thanks

Message 32 of 34
seiferp
in reply to: itsupport

This is not the Inventor Forum just in case! If you are searching for current supported cards for VRED try here:  https://knowledge.autodesk.com/certified-graphics-hardware

 

 

Message 33 of 34
andreasK3K4G
in reply to: seiferp

Hi,

for me the big advantage using the RTX cards, it´s not to do gpu rendering, cause they are not any better then the non rtx cards. (just slighty faster cause of the new generation)

The use of the Tensor Cores for AI acceleration for denoising is the point.

Using one or two RTX cards in addition to a small cpu cluster will do renderings much faster than a big cpu cluster.

Message 34 of 34

The RTX cards should be certified for VRED already including the RTX5000, which was missing initially for some unkown reason (my guess is we simple didn´t get a card in time for the certification).

 

Speaking the the AI denoising: This only works well with one GPU at the moment, the necessary data transfers between the GPUs is often more expensive than just denoising on one GPU. 



Michael Nikelsky
Sr. Principal Engineer

Can't find what you're looking for? Ask the community or share your knowledge.

Post to forums  

Autodesk Design & Make Report