I recently purchased the Asus Strix GTX 1080 - viewport is great and the RT hasnt WOW’d me on heavier scenes. Sure when i look at a shaderball it works great but so did my old GTX 780.
Of course, the viewport becomes sluggish when RT is activated, do i get around this by adding another GTX? Its far too sluggish to be practical for larger scenes. I was thinking getting a 1070 or even a 1060 - how does that work in terms of 3DS Max/Vray - Can i allocate which card does the RT and which does the viewport? and when im not doing RT, can I still make the change as to which card does the viewport? Hoping that makes sense.
Also, is there a test scene available as a benchmark?
I have a current interior scene at 1.7 million polys. At a resolution of 640x480, ive set the rendering time to 1 minute (CUDA) and i get 159 passes. Not sure how that compares to anything as I dont know what to compare it to - so like i said if there was some form of benchmark that would give me some peace of mind.
Thanks guys and looking forward to some responses.
Of course, the viewport becomes sluggish when RT is activated, do i get around this by adding another GTX?
You can try lowering rays per pixel to 1 in the RT Settings to see if this helps.
I was thinking getting a 1070 or even a 1060 - how does that work in terms of 3DS Max/Vray - Can i allocate which card does the RT and which does the viewport? and when im not doing RT, can I still make the change as to which card does the viewport?
Yes, absolutely. If you have two GPUs you can plug the monitor into one of them and turn it off from the “V-Ray RT Render Devices Select” list in the RT GPU settings.
Also, is there a test scene available as a benchmark?
We have a very old one here. 1080 is roughly as fast as a Titan X (Maxwell).
Make sure to check the (really) short FAQ here, it might save you some troubles :).
What super gnu said. I am even using AMD RX 480 as a GPU for the monitor and computations and I have another GTX 1080 for computations. In the case with AMD for monitor and GTX for computations, I have tested only with NVIDIA GPUs that have TCC mode (980ti, Titan X, 1080, Quadro, Tesla, etc) and I don’t know if it works with different ones.
If you have all NVIDIA or all AMD setup you should not have any troubles.
The annoying part though is that if you have and old graphics card like for example a GTX 780 or something, and then buys a 1080 or the new Titan X you must dedicate that new card for Vray RT but are stuck with the old 780 for everything else, like the Max viewport and other applications. Isn’t that true Blago or is there some way around this without having to move the video cable and reboot all the time? Feels like such a shame to dedicate the latest and greatest to pure RT use.
Another option - use the new GPU for viewport, old for RT preview. When you render for final frame, use both of them for the rendering. You can choose which GPUs to be used for RT GPU from the V-Ray RT GPU settings tab.
Yes, but that means all my previews and active shade in RT will be on the old card, which defeats the purpose of all the CUDAs on the new card. Besides, I always do my final render on Vray Adv.
What about automatically using only 85% or so for everything but production render mode? Isn’t that the way Redshift and Fstorm seems to handle it?
You either don’t utilize fully the GPU or you get sluggish UI. If you don’t utilize the GPU well it can get much slower … You can go for lower utilization with RT GPU with rays per pixel 1, but for many scenes it is not low enough. I am thinking of adding option for that on-purpose underutilization. I will count your post as a request for that :).
So ive got the gtx 1070 now - and boy does it make a difference having 2
Anyway, i have another slight issue. Currently I have the GTX 1080 which is plugged directly into my monitor (I only have 1x34") and now the new GTX 1070. See image attached. 1080 is index 0 and the 1070 is index 1. When i uncheck the GTX 1080 under the RT devices select, it uses the 1070 for the RT rendering and the 1080 for the viewport (correct me if i am wrong). However when i uncheck the 1070 and leave the 1080 checked, it uses the 1080 as the RT device BUT the viewport is super-slow. It doesnt seem to be using the 1070 for the viewport - it seems to use the 1080 for both RT and viewport - Is this becuase the 1080 is plugged into my monitor directly?
The viewport uses the GPU with the monitor attached and we have no control over this.
V-Ray will use whatever you select and if nothing is selected - will use all the available cards.
So in your case - the viewport will use the 1080 no matter what you select in the V-Ray RT Render devices selection. Maybe you could plug the monitors in the 1070 and use the 1080 for rendering ?
Any more thought into this? I just have a hard time understanding why this works well for Redshift/Fstorm but not for us?
I’m also curious about this technology where the GPU can utilize RAM memory when you go over the board’s. Is that something that are being considered? It would seem that paging against slower memory is better than not being able to start the scene at all?
