I currently have a HD7850 which I use daily, but I'm about to do a project that uses CUDA so I borrowed an NVidia card.
Is it possible to just stick the NVidia card next to my AMD card and be able to use the CUDA capabilities while still using my AMD card for everything else? Or is such a hybrid not possible? Will the drivers interfere if I install them concurrently? I'm on windows 7.
>>57551087
I'd say it depends on the motherboard and chipset. Normally on a desktop with Intel graphics and Dedicated graphics you set the display device through the BIOS, or it auto-detects whichever has a monitor plugged in. If both GPUs have a monitor plugged it often assumes one as default. I'm guessing if yours has two PCI-E slots, it will run whatever's in the first slot and ignore the 2nd one.
>>57551087
Yes you can.
>>57551161
Wrong. You can use multi monitor from different GPUs, including the iGPU. You can use a GPU for compute only while using another GPU for your monitor. How else do you think nvidia teslas work when they don't even have video ports?
Never mind, my power supply won't be able to handle both cards at the same time anyway :^)
Guess I'll just switch the cards for now.
>>57551223
Maybe i'm using an older motherboard because i dont have an option like that. Compute sure, but i've never seen two or more GPUs as display devices, especially from different brands...
>>57551254
retard
you can use all of them at the same time, you dont need a BIOS option to pick one
>>57551254
>Compute sure
>>57551087
>and be able to use the CUDA capabilities
Should work for CUDA stuff.
I believe Nvidia don't let you use their GPUs for PhysX in games if you have an AMD card or something like that.
>>57551087
yes you can, but you will have to gogle for unlocked drivers. Nvidia is going out of their way to block driver installation if amd gpu is detected, ever since people started getting cheaper nvidia gpus for dedicated physx acceleration.
They also disliked Lucid hydra solution, or perhaps it was just crap.
>>57551322
Why is Nvidia such an anti-consumer shit company. Damn monopolists.
>>57551487
They're just following Apple's lead
Just for the record it does work, the CUDA APIs expose a way to choose which GPUs to use.
You can also use multiple nvidia GPUs that don't support SLI for concurrent CUDA computations