Are we looking at the same graph? >_>
Vega 56 gets about 75 FPS average, while Vega 56 CF gets 126 FPS, a 67% improvement. The minimum frame rates only show a 28% improvement, but that is likely due in part to the CPU limiting performance in those cases. Likewise, an RX 580 averages around 50 FPS, while RX 580 CF gets 96 FPS, a 91% improvement, and minimums show about a 54% improvement for that card.
In fact,
the article that image comes from includes the line...
If you have a second matched GPU available, both CrossFire and SLI work, with CrossFire in particular putting up some impressive scaling results.
And at 4K, these two cards see 92% and 98% average frame rate increases respectively, with their minimums showing 71% and 86% improvements. So, that's not a particularly good "case in point", at least for CrossFire. The SLI scaling in that game is certainly less impressive, but makes at least some notable difference, which is more than can be said for many other games. In general, I agree that multi-card setups are typically not the best option, unless perhaps someone is not satisfied with the performance of the fastest cards, and is fine with spending a lot of money on something that will only work in a limited number of games, but you could have probably found a better example of a game that doesn't support multi-card setups well. : P
buzznut47 :
And why can they not put the cmos battery in a better place? If I have 2 video cards I'll never be able to get to it. If they're water cooled, I need to take apart my whole loop if I need access to the battery.
Would you even need to remove the battery though? If it's for resetting motherboard settings, I imagine that the board includes a button or jumper to perform the same task. And I really doubt that the battery would be likely to fail for a number of years, during which time anyone with liquid cooled graphics cards would probably need to tear their loop apart for maintenance more than once anyway. And of course, cards with standard air coolers should take only a matter of seconds to remove.