Topic: Nvidia Quadro Cards

[tug]

Hi all

As you know I'm usisng several graphics cards based in Nvidia Quadro Fermi chipset.

Quadro cards are very specialized one, for technical works and "CPU" rendering, even havig GPU cores and being more expensive than Geforce Series.

We have tested in our work Quadro FX 3700M (1Gb), Quadro 4000 (2Gb) and Quadro 5000M (2Gb) in large screens (Dell U2711) and large resolution (2560x1440), also we heve tested Nvidia GTX 5800 (3Gb) from Gainward and reach some conclusions maybe could be useful for Quadro users and Lumion.

The first thing we've understood is that Quadro are not the best choice for Lumion, GTX 580 are faster and cheaper. But if we have those cards and try to get the best performance with them we must reduce the screen size (not full screen) to work and reduce the "stars quality" inside Lumion to 2 or 3 stars, with these simple tips we've we managed to optimize the performance of the Quadro and Lumion, getting to work faster and more smoothly.

From all those we are using now, Gainward GTX 580 Phantom (3Gb) is the best one.

I hope you can solve, as we, how working with Quadro cards with this simple tip.

[Michael Betke]

This is the downside working with realtime graphics and 3d too.
GForce cards are for gamers and cut down in driver support for 3D works. Wireframe display is not supported so your GForce card will idle around 2-8% while trying to rotate a full scene within 3dsmax in DX9 mode.
But it delivers great performance in real-time tools because its made for this task.

On the other hand Quadro cards deliver great performance within 3dsmax but are poorly optimized for realtime (only CUDA computing).

You would go best with two cards in a workstation. I also tried the 3dmax 2012 demo yesterday and this new Nitrous stuff accelerated my scenes a lot. in DX9 max is as sluggish as in my current 2010 version.
I got improvements from 2fps to 30+ with a complex DWG file for example. I will keep my 3GB Gainward but upgrade to max 2013 next year to have best of two worlds.

[tug]

that's right Michael, this is the worst side for designers whom want to use this software in realtime rendering.

We opted for buy a new workstation dedicated exclusively to Lumion with the Gainward 3GB and in a short future when we can find a accurate Power Supply for this workstation install another "Gainward Phantom" together, finally we'll be using two GTX SLI with 6 Gb for Lumion.

Be sure this is the best solution, we design with Quadro Cards and rendering in Lumion with GTX, expensive? maybe but better solution, we reduce rendering time and designing time.

[Michael Betke]

You wont have 6GB available if you use two 3GB cards because the data gets feed into both cards evenly.

Its the same for GTX590. It has 2x 1,5GB and you only can use 1,5GB not 3GB.

There is a TI 550 now from POV. It has 4GB Vram. But its not that fast. Maybe two of them for "okay" speed with 4GB usage would be a good choose. Over 4GB you would have to go with a Quadro again. :-/

[tug]

Mmmmmm must test it then, we tested Quadro 6000 with a huge investment and decided change it for 5000M or 5010M

[ilwedritschel]

hello,

i make some test with diferent gtx and quadro. i have a question i can anderstand this:
wath will happend if the gpu ram ist full?
 i work with a scene and the gpu usage is about 99.9% is still working. (gtx570 1250gb) 25 fbs
on quadro 600 1000 gb the same scene is also working. not so fast only with 6 fps
on gtx 450 1000 gb the same scene is also working . 12 fps

the differnenz of the fbs is the differenz analog to the streaming proz 96-196-420 the same differenz for movie renderings.

what woud happen if i work with 3000 gb and scene wich used 99% and then change to the system with 1000 gb?

it crasy, i wont buy a gtx 590 for fast renderin. SLI is not the right because, can use only on monitor.

we must make a tabelle wich show the real results of the gpu with one referenz scene.

[Michael Betke]

For GPU Rendering like "Octane Renderer, iRay, etc.." the scene wont start rendering with too much data in the vram.

For Games the data will be swapped to the computers ram making the scene very slow and sluggish. I think Lumion scales down textures automatically then after re-import for complex scenes if I I observed this correctly.

SLI needs to be supported by the tool you use. For example 3dsmax doesnt support a SLI profile so you wont have a performance benefit. For rendering you will benefit of the cards a lot.
I dont know how Lumion supports two graphic cards and if it scales well. One or two TFT screesn shouldnt be a problem with SLI.