Benefit to 2 GPU's in Mac Pro?

[violst]

Is there a benefit to having 2 GPU's in my Mac Pro? I replaced a working 8800GT with a 5770. I have two monitors hooked up the the 5770 a Dell u2711 and a u2312.

What I would like to do is increase my open GL performance. Should I hook up the u2711 to the 5770 and the u2312 to the 8800GT.

Any advice will be greatly appreciated.

[jav6454]

Depends on your usage... what do want to achieve?

[violst]

I would like to increase my open GL performance for applications like Cinema 4D.

[MacVidCards]

You have a car.

You have 2 @ 4 cylinder engines of 2.5 L each.

You have a V8 of 5.0 litres.

While it may be possible to come up with a complicated set of transmissions and transfer cases and have the engines drive separate ends of the car, it would be a whole lot easier to just drop the V8 in and use a regular transmission and have a faster car.

Same deal here. For the 2 mid grade cards to help each other and create speed of larger, faster card would require some hi-tech OpenCl goodness. It doesn't exist. Even in Windows world it requires 2 cards of identical chips connected via cables and software.

We don't have the SLI/Crossfire software and your cards don't match anyway. You are back to having 2 @ 4 cylinders and no way to have them help each other.

A "V8" in this instance would be a 5870 or GTX580, fastest Mac cards with boot screens today. In a week or two that might change. Right now you have a couple 4 bangers that don't like each other.

[ClassObject]

for two monitors, really no benefit. if you want to run four+ monitors, multiple cards are your friend

[jav6454]

Originally Posted by violst

I would like to increase my open GL performance for applications like Cinema 4D.

More cores usually signifies better performance, but there are other factors such as the type of GPU architecture you are putting in. Surely there are better options than 2009's HD 5000 GPU series.

[Tesselator]

Originally Posted by violst

I would like to increase my open GL performance for applications like Cinema 4D.

The speed you're probably talking about is selecting, creating, and transforming vertices, faces, edges, sub-patches, and so on. Just a stagnant model rotates pretty fast on the cards you have now right?

If this is the case the speed is determined two things primarily - way above the abilities or speed of the GPU you're running. Those two things are main-board memory throughput speed (bus+ram specs), and the CPU clock rate. Get a main-board that accepts the fastest RAM of the day and which can overclock the CPU to like 4.5 or 5.0 GHz (two cores is fine for modeling and scene editing - more cores will NOT help you here) and watch that 8800GT frigging fly! Something like a GTX 570 will help a little bit but you'll still be wondering where all the speed is if the other components mentioned aren't there first. The RAM and CPU speeds are 80 to 85% of it - the card believe it not is only about a 15% to 20% contributed - for all things 3D editing (LW3d, C4D, Max, Maya, Houdini, Modo, ZBrush, and so on...).

Rendering is of course a different story - as well as game graphics and video too...

[echoout]

As Nick over at Greyscale Gorilla points out in his optimization tutorial, some of the biggest resource sucks in Cinema are global illumination, ambient occlusion, antialiasing, blurry reflections and area shadows. I've found my video card to have very little affect on the performance of just getting around in Cinema. I enjoyed having 24 buckets during my renders before I sold my Mac Pro but nothing beats a really really fast processor and RAM as others have pointed out.

I've not used anything out there that's made Cinema feel snappy when the aforementioned features get cranked up.

[Tesselator]

Originally Posted by echoout

As Nick over at Greyscale Gorilla points out in his optimization tutorial, some of the biggest resource sucks in Cinema are global illumination, ambient occlusion, antialiasing, blurry reflections and area shadows. I've found my video card to have very little affect on the performance of just getting around in Cinema. I enjoyed having 24 buckets during my renders before I sold my Mac Pro but nothing beats a really really fast processor and RAM as others have pointed out.

I've not used anything out there that's made Cinema feel snappy when the aforementioned features get cranked up.

Nods to the CPU+RAM speed again. If he didn't however, Nick should point out that all of those features (Display modes) should be only momentarily turned on when adjusting and then immediately turned off again right after. Most CG packages will not try and use DX, GL or another fast card functions for those attributes as they then will not match up in the rendering engine. Typically the most one can hope for is GLSL textures. But even though most CG apps do not take advantage of the same accelerations used in/by game rendering engines, some will allow you to use specific game rendering engines directly in a viewport for the purpose of game development.

So if it's those are the things that are slowing you down Violst, just turn them off like normal.

[echoout]

Originally Posted by Tesselator

Nods to the CPU+RAM speed again. If he didn't however, Nick should point out that all of those features should be only momentarily turned on when adjusting and then immediately turned off again right after. Most CG packages will not try and use DX, GL or another fast card functions for those attributes as they then will not match up in the rendering engine. Typically the most one can hope for is GLSL textures. But even though most CG apps do not take advantage of the same accelerations used in/by game rendering engines, some will allow you to use specific game rendering engines directly in a viewport for the purpose of game development.

So if it's those are the things that are slowing you down Violst, just turn them off like normal.

Yeah, the whole tutorial is about how little you can get away with. It's just a process that is so tedious that it doesn't feel very creative to me. Heaven forbid a designer can see what they're doing. I can dream.

[Tesselator]

Mmmm, it's not too bad... IRay and others can be implemented to speed up workflow without having to test render all the time. It works. Ya just need to keep your interactive viewports simple - even wireframe. Here's two samples of IRay in use:

.

[violst]

Thanks for all the replies they are very helpful.

I have been using a lot of mograph cloning lately, large scenes with many clones and the viewport is sloooowwwww. I know a better GPU will help but I was just wondering if the 2nd GPU would help any, and as I suspected it will not.

I have been hoping the new Mac Pro would come out already so I can replace my old box. Im just trying to optimize it the best I can with what I have right now.

[Tesselator]

Originally Posted by violst

Thanks for all the replies they are very helpful.

I have been using a lot of mograph cloning lately, large scenes with many clones and the viewport is sloooowwwww. I know a better GPU will help but I was just wondering if the 2nd GPU would help any, and as I suspected it will not.

I have been hoping the new Mac Pro would come out already so I can replace my old box. Im just trying to optimize it the best I can with what I have right now.

In that case it's what MacVidCards said. And hey, even if it turns out that it doesn't, at least you'll have the killer card for the stuff it will do faster...

[nigelbb]

Hard core colour graders using DaVinci Resolve will have multiple graphics cards in a Cubix PCIe expander box so it works for CUDA.