Originally posted by: Czar
looks just like the Alienware implimitation
Originally posted by: Ryoga
The effective results -- each card renders only half the screen -- are the same, but it's not clear what kind of performance difference there is between each method.
Originally posted by: Insomniac
Seems kind of odd they are load balancing with software. If they used 3dfx's way of rendering odd or even lines, there wouldn't be much need to load balance since the odd and even lines are next to each other and wouldn'tbe much different. They bought all the 3dfx tech, and I though they were finally putting it to use. I guess not. :/
Who cares? With so much integrated goodness on today's new motherboards how many people actually use all their expansion slots? I use one for my Leadtek TV2000XP Deluxe.Originally posted by: Carbonadium4
Alienware requires a thrid card... so either way, you're gonna lose slots left and right..
Originally posted by: Tostada
3dfx's implementation was obviously less efficient than having one card render the top half of the screen and the other card render the bottom. If you interleave, each card is rendering the full screen at half the resolution. If you split the screen, each card only renders half of the vertices.
Ok. I HAVE to ask: If you only render every other line, wouldn't that lead to each card rendering 1/2 the vertices??? It doesn't matter if you divide the screen into two large chunks or 1000 small chucks and have each card render 1/2 of them, either way each card is rendering 1/2 the display. That "obviously less efficent" comment was particularly amusing.
I have a feeling that the reason that 3dfx and NV implemented the dual video card method differently has more to do with HOW it was implemented: 3dfx was very much a hardware solution and interleaving the lines was very easy and led to natural load balancing. The drivers really didn't have a whole lot to do with it. NV is going for a strong software approach. It won't even work at all with out the right drivers. But as Tostada did mention, some effects do require knowledge of the adjacent pixel results to complete the calculations for the pixel you are working on. If they did SLI the way 3dfx did it, there would be a HUGE amount of traffic between each card that would hurt performance. When 3dfx came out with SLI, those effects did not really exist.
Originally posted by: Tostada
Originally posted by: Ryoga
The effective results -- each card renders only half the screen -- are the same, but it's not clear what kind of performance difference there is between each method.
Originally posted by: Insomniac
Seems kind of odd they are load balancing with software. If they used 3dfx's way of rendering odd or even lines, there wouldn't be much need to load balance since the odd and even lines are next to each other and wouldn'tbe much different. They bought all the 3dfx tech, and I though they were finally putting it to use. I guess not. :/
3dfx's implementation was obviously less efficient than having one card render the top half of the screen and the other card render the bottom. If you interleave, each card is rendering the full screen at half the resolution. If you split the screen, each card only renders half of the vertices.
FSAA is also pretty much impossible with 3dfx's SLI.
Originally posted by: Megatomic
Who cares? With so much integrated goodness on today's new motherboards how many people actually use all their expansion slots? I use one for my Leadtek TV2000XP Deluxe.Originally posted by: Carbonadium4
Alienware requires a thrid card... so either way, you're gonna lose slots left and right..
Wait, wait, wait... I can already see a handful of people listing all their PCI cards and saying how they wish they had more. To this I say, welcome to the rare group of <1% of computer users who use many/all of their expansion slots. You do not represent the majority (nor a large minority) of users.