2 GPU's using both card's cuda cores without SLI possible?

459pm

Honorable
Dec 3, 2012
84
0
10,660
In my Nvidia control it has a section to select what nvidia GPU's cuda cores can be used to play a game. Does that I mean that I can put my old GTX 650 back into my rig in addition to my GTX 770 and get performance increase by using the CUDA cores from the GTX 650 too?

I thought this was the same idea around SLI...but I'm still rather confused about how cuda cores, SLI, ect works. Please help!
 
Solution
If you want both cards to work on the same task you must SLI them, this makes them seem as a single GPU and work as a single GPU. If you were to put your 650 in beside your 770 they cannot SLI so it won't boost your gaming performance but you could use it to offload PhysX calculations to for games that support GPU based PhysX acceleration. If you wanted to use them for a GPCompute workload you could use both of them and have them both working on different datasets, that would work fine and is how the Tesla cards are used.

In short, if you want multiple GPUs to drive a single screen they must SLI, if you want the second one to do something different you don't need SLI.
No, you even said it yourself. It lets you select what GPU to use to play a game, not both. Both is SLI, that's what SLI is all about, combining two cards together, but they ahve to be the same card. The best you could would be put in your 650 and try running it for PhysX and the few games that support it.
 
As said above you could if your PSU is good enough and your motherboard supports it put the GTX 650 back in and us it as a dedicated Physx card but you would not be able to SLI them at all. And really with the power of the GTX 770 there is really very little point in it to tell you the truth. I would just keep it out and use your GTX 770.
 
If you want both cards to work on the same task you must SLI them, this makes them seem as a single GPU and work as a single GPU. If you were to put your 650 in beside your 770 they cannot SLI so it won't boost your gaming performance but you could use it to offload PhysX calculations to for games that support GPU based PhysX acceleration. If you wanted to use them for a GPCompute workload you could use both of them and have them both working on different datasets, that would work fine and is how the Tesla cards are used.

In short, if you want multiple GPUs to drive a single screen they must SLI, if you want the second one to do something different you don't need SLI.
 
Solution

smartpc

Reputable
Sep 12, 2014
5
0
4,510
hey there...does it matter which card is used to power the screen??
i have an nvidaia quadro fx3800 and an nvidia grforce 9800gt...i also have an onboard gpu...can i view the vscreen from onboard and use the other 2 cards to help with extra processing...i use cs6 on win7 and do alot of video editing/color correction/rendering...
what would be the best way to use these..??
thanks