A purely hypothetical question: If I have two GPUs in my computer can I create an OpenGL context for each of them and render with both of them?
In theory, yes. But it cannot be done without vendor-specific extensions: NV, AMD. LWJGL supports neither of the two at the moment, but they’re both in the TODO list for the 3.0 API redesign.
LWJGL 3.0 is gonna have an API redesign? How major will this be? Could you provide more info about this please? 
The static nature of the AL/GL/CL APIs will not change, so the bulk of your code will work without changes. But we’ll have to do some incompatible changes to the window and context management APIs. It’s necessary in order to satisfy certain needs that have arisen recently. Some examples:
- Better integration with 3rd party contexts (e.g. sharing contexts with SWT GLCanvas contexts)
- Support for multiple windows/Displays.
- Support for multi-GPU solutions.
- Display-mode switching without window destruction.
It’s also hard to support additional window features with the current code. A refactoring will make our work less painful and allow for better integration with 3rd party windowing systems (e.g. Prism when it hopefully becomes open-source).
Wow, I love all the hard work you guys put into this. Good luck!
HOLY MOTHER OF GOD HAVE MY BABIES
Errrrr…
You will understand when you’re older
I think I’ve returned to my senses again…
- WHEEEEEEEN?!?!?! (Okay, I haven’t.)
- What do you mean by “support for multi-GPU solutions”? The ability to tell the driver to use AFR or SFR (unlikely) or the ability to specify which GPU to use for which contexts (what I asked for in the opening post)?
- WHEEEEEEEEEEEN?!?!?!?!
- Random question: Is a 750W Corsair power supply enough to drive a GTX 295 plus a new Kepler card?
- WHEEEEEEEEEEEEEEEEEN?!?!?!?!?!?!??!
- Random question 2: Is it possible to connect two different NVidia cards with an SLI-bridge and utilize it for data transfer between the cards?
- WHEEEEEE-- Oh, wait, I think you get it by now…
I have a Cosair 750W for my GTX 580 and i7 2600k so I do believe you’re very fine
I meant having both a GTX 7xx card AND a GTX 295 plugged into the same computer at the same time…
I have no idea how much power a GTX 295 draws but it’s a dual GPU card…I would get a bit more than a 750W…
The best place for hardware questions is the #hardware channel on Freenode. Those guys are awesome
I believe SLI works only when it’s the same type of card. I know that’s the case with my 2 560’s.
Some cards do support “mixed SLI”, but only for minor revisions within the same family. The 570 and 580 could probably do SLI together for example. But a 780 and 295, forget it.
Plus, you could probably keep a hot air balloon aloft with what you’ll have blowing out the back of a machine with both of those cards running full blast.
when you want 1 render context you need 2 exactly equal GPUs
,but if u want to use 2 render contexts or want to render on one card and do calculations on the other you can have any 2 GPUs you want(and they needn’t be connected by a SLI bridge).
I think you can’t share or exchange data on both cards if they aren’t connected with SLI or Crossfire.
For example you need two GPUs for debugging CUDA(or a notebook with a Nvidia Optimus card, which all recent are), but no SLI. You can even debugg over LAN if both PC have a CUDA capable GPU.
I know.
This is why I asked about multiple contexts in the first place. I know the GPUs are different, I just wonder if I can render with them all in parallel by manually dividing the workload between them, and send the data over an SLI-bridge to avoid a PCI/CPU bottleneck.