CAD card + Geforce card goes to which displays?

[Kaido]

This is probably a dumb question, but I don't know what the answer is. I've built a couple engineering computers recently where they needed a CAD card for dual monitors (say a Quadro FX 5800), plus a couple extra monitors for email & such. I slapped in some cheap cards like an 8400GS for the extra monitors. How does CAD rendering work across the monitors?

Does the application pull from the CAD GPU no matter which monitor it's on? Or does it default to the CAD card-connected LCD's? If you drag it halfway across a CAD-connected monitor to a cheap card-connected monitor, how does that work? I know with the newer laptops that have the iGPU's, you can toggle which applications actually use the dedicated GPU so you don't burn up the battery life. On a desktop without that setup, how does it handle it under Windows typically?

 

[i7Baby]

You should run all your monitors off the one card.

 

[Kaido]

You should run all your monitors off the one card.

There are 4 monitors; most of the CAD cards, even the higher-end ones, only support 2 monitors (despite having 3 physical plugs). An extra Geforce card was put in to add the extra 2 monitors, bringing the total up to 4. Works great on all machines I've put it on (Quadro 4000, 5800, 6000, etc.), but I don't have a clear idea of how the rendering is distributed. It doesn't make much sense financially to buy a pair of Quadro cards when the extra 2 monitors are only going to be running email or PDF files, but I am curious as to how the OS & drivers distribute displaying CAD files.

 

[nvgpu]

You can't use Quadro and GeForce cards simultaneously because the driver INFs don't support each other.

You'll have to get a newer Quadro M4000 or K1200, both with 4 DP 1.2 outputs for quad-monitor support.

http://www.nvidia.com/object/quadro-desktop-gpus.html

Another alternative if those systems have Intel integrated GPUs is to plug those monitors into the motherboard display outputs.

 

[zir_blazer]

You can't use Quadro and GeForce cards simultaneously because the driver INFs don't support each other.

He said that it is already working that way.


Can't you do something like keeping two GPU-Z instances open at the same time with the Status tab for each card and check for yourself if the other one does something? Based on my experience back at BitCoin mining, coordination between multiple Video Cards is nonexistant. The most that you can expect from an application, is that it allows you to choose the GPU that you want to use, and to open multiple sessions per each GPU akin if you were using a Single Threaded application and needed to open one and pin it to each CPU Core to get to Full Load. Basically, the rendering itself should possibily be done by the Quadro alone.
A thing which I never got is if there is a single big framebuffer in a master card that distributes it to the video output of the other GPUs, if each one has a local framebuffer for the Monitors its supposed to drive, or if it instead uses system RAM and distributing the framebuffer to the respective cards/video outputs is done purely by Software. I don't know if moving stuff around the Desktop surface of multiple Monitors is actually moving the data from card to card or how it is supposed to work from the point of view of 2D acceleration. This also applies since it is possible to use totally difference Video Cards (Example: Using the Intel IGP and a discrete Video Card and get output from both at the same time).

 

[Kaido]

He said that it is already working that way.


Can't you do something like keeping two GPU-Z instances open at the same time with the Status tab for each card and check for yourself if the other one does something? Based on my experience back at BitCoin mining, coordination between multiple Video Cards is nonexistant. The most that you can expect from an application, is that it allows you to choose the GPU that you want to use, and to open multiple sessions per each GPU akin if you were using a Single Threaded application and needed to open one and pin it to each CPU Core to get to Full Load. Basically, the rendering itself should possibily be done by the Quadro alone.
A thing which I never got is if there is a single big framebuffer in a master card that distributes it to the video output of the other GPUs, if each one has a local framebuffer for the Monitors its supposed to drive, or if it instead uses system RAM and distributing the framebuffer to the respective cards/video outputs is done purely by Software. I don't know if moving stuff around the Desktop surface of multiple Monitors is actually moving the data from card to card or how it is supposed to work from the point of view of 2D acceleration. This also applies since it is possible to use totally difference Video Cards (Example: Using the Intel IGP and a discrete Video Card and get output from both at the same time).

Yeah, that's what I'm trying to dig into. I haven't played with the CAD stuff extensively as I install it rather than use it myself, but next time I visit a user I'm going to check it out. The models seem to rotate fine on the other monitors, so I dunno. Some are older (Quadro 4000) and some are newer (Quadro 6000) with various GeForce cards installed as secondaries, zero issues, had some of them running in that config for over a year. I'd just like to have a clear idea of how everything is being piped, you know?

 

[tweakboy]

FireGL from google it. gl