Originally posted by: ddarko
Originally posted by: praesto
Originally posted by: ginfest
Is that GTX280 performance across the board? All games and no problems? I thought I've read that when CF works it's great but that there were compatibility problems with some games? Please clarify because I can get the 2 4850 for $300.00 and have been offered $250.00 for my card.
Or is the 4870 the way to go?
No it's not gtx280 performance across the board. Nobody can promise you that there won't be any problems in any games, neither can they promise gtx 280 performance in all games. CF doesn't work AT ALL in some games, which leaves you with one hd 4850. Some games might work with CF, but only let you gain a 30% performance increase over a single hd 4850. Hell, some games will give you 95% performance increase over a single hd 4850. That's how multi-GPU works.
My personal opinion: Multi-gpu is not reliable, be it SLI or CF.
The thing that is overlooked is that GTX280 isn't always GTX280 performance across the board. Even single GPU solutions produce results that fall short of where they should. Look at Anand's review of the GTX280/260 and examine the Assassin's Creed and Bioshock results. The 280/260 are producing far below where they should. In the other games, the 280/260 give the highest performance of any solution, even exceeding SLI/Crossfire. But in some games like Assassin's Creed and Bioshock, the 280/260 comes is below where they should, hardly exceeding a 9800GTX.
My point is, performance will vary from game to game for BOTH dual AND single GPU solutions. Single GPUs do suffer from the same performance variance that dual GPU solutions do yet, for some reason, while SLI/Crossfire are criticized for this, when it happens to single GPU, it gets excused.
I am not saying variance is not a problem; just that it occurs with single GPUs as well. I think the way to deal with this is to focus ultimately on the results and ask: even when the solution doesn't work as well as it should, does it still give acceptable results? In some cases, the answer is no, the frame rates are too low. This was the result of 4850 in Crossfire in the Witcher benchmark where you only got frame rate around 21 fps. But in many other cases, the frame rates, even though they're coming in below where they should, are still giving great results. In Assassin's Creed, the 280/260 are still producing results in the high 30s to mid 40s. In Bioshock, in the 50-60 fps range. The same goes for 4850 in Crossfire mode for Bioshock, where even though Crossfire didn't scale at all, it still gave a frame of 55 fps at 2560x1600 resolution.
Don't get hung up the efficiency of scaling. Efficiency is a useful criteria for evaluation but efficiency for efficiency's sake is missing the forest for the trees. In the end, look at the frame rates you're getting. Even if you're getting terrible efficiency, why does it matter from a gaming perspective if you still get high playable frame rates.