VirtualLarry
No Lifer
I just had the most off-the-wall idea, after reading *yet another* ATI vs. NV comparison in one of the many threads posted here.
Just thinking about the multi-GPU-on-a-single-card designs, from Gigabyte and others... ok, what if it were possible... to put both an NV, and and ATI GPU, on the same card?
(I did say that it was a crazy idea, didn't I?)
I would assume that the memory would be shared (or perhaps not, but that would double the cost if not), possibly with some sort of third-party high-speed memory-crossbar chip, to allow attachment to both GPUs, and possibly a similar host interface crossbar, or bridge chip, to allow a host, running custom 3rd-party software, to select the use of one GPU or another on the card, or potentially even allow them both to run at the same time, doing different tasks (in the case of a fast enough host interface, say PCI-E, and with independent GPU memory banks on the card).
For games that "prefer" NV, you would use the NV GPU, and likewise for ATI-optimized games. The RAMDAC would be offboard, and shared by both GPUs, such that you wouldn't have to switch monitor cables or anything.
Granted, even if this is technically feasable (possibly not, due to memory latency/trace-length issues), the cost would be exhorbitant. But for those money-is-no-object to a video-card purchase people, this could be the very thing for those that "want it all" in terms of GPU support/performance.
Wish I had thought of this on the 1st, would have made a nice idea for a photoshop gag. 🙂
(Btw, this idea is not unprecidented. There have been graphics accelerator chips, that were not themselves VGA-compatible, so they had to include a seperate 3rd-party VGA-compatible video controller core onboard as well, to handle DOS games and boot-up, and then when the Windows' driver loaded for the accellerator chip, the DOS-mode VGA core was disabled. The original Weitek-based Diamond Viper cards were like this - that was one of the reasons that there were problems writing drivers for it too. Ditto for the 3DLabs Glint-based Accel 500TX CAD card that I once used.)
Just thinking about the multi-GPU-on-a-single-card designs, from Gigabyte and others... ok, what if it were possible... to put both an NV, and and ATI GPU, on the same card?
(I did say that it was a crazy idea, didn't I?)
I would assume that the memory would be shared (or perhaps not, but that would double the cost if not), possibly with some sort of third-party high-speed memory-crossbar chip, to allow attachment to both GPUs, and possibly a similar host interface crossbar, or bridge chip, to allow a host, running custom 3rd-party software, to select the use of one GPU or another on the card, or potentially even allow them both to run at the same time, doing different tasks (in the case of a fast enough host interface, say PCI-E, and with independent GPU memory banks on the card).
For games that "prefer" NV, you would use the NV GPU, and likewise for ATI-optimized games. The RAMDAC would be offboard, and shared by both GPUs, such that you wouldn't have to switch monitor cables or anything.
Granted, even if this is technically feasable (possibly not, due to memory latency/trace-length issues), the cost would be exhorbitant. But for those money-is-no-object to a video-card purchase people, this could be the very thing for those that "want it all" in terms of GPU support/performance.
Wish I had thought of this on the 1st, would have made a nice idea for a photoshop gag. 🙂
(Btw, this idea is not unprecidented. There have been graphics accelerator chips, that were not themselves VGA-compatible, so they had to include a seperate 3rd-party VGA-compatible video controller core onboard as well, to handle DOS games and boot-up, and then when the Windows' driver loaded for the accellerator chip, the DOS-mode VGA core was disabled. The original Weitek-based Diamond Viper cards were like this - that was one of the reasons that there were problems writing drivers for it too. Ditto for the 3DLabs Glint-based Accel 500TX CAD card that I once used.)
