How is performance with SB in crossfire/SLI setups?

[Jhatfie]

I am currently running a i5-750 @ 4Ghz and crossfired 6950's in eyefinity. Looking around I am not seeing much if any information on Crossfire/SLI scaling with the new SB 2500k/2600k cpu's and how they compare to the previous generation.

Are there any reviews that I just must be missing or what kind of performance increase might I expect with multi GPU with the potential upgrade to the Sandy Bridge Platform?

 

[createdAmonster]

I gather it all depends on the resolution you are playing at.
The very few numbers I have come across regarding gaming, show a very negligible increase in performance as far as gaming goes.
I would venture to say that you would not see an increase in performance in gaming that would warrant upgrading to SB.
I will try to find that review link.

 

[Axon]

It's the same or better than the P55 chipset. I run 8x/8x lanes on my 1gb 460 SLI.

 

[Castiel]

P55's lanes are about half the speed of P67's

 

[VirtualLarry]

P55's lanes are about half the speed of P67's

The chipset lanes, yes, but not the GPU lanes. They are both PCI-E 2.0 on both the 1156 and 1155. Therefore, there's no performance difference due to PCI-E.

 

[createdAmonster]

Did the crew over at Hardocp not conduct a test to measure the performance difference between x16x/16x and 8x/8x? [Edit] and even x16/x16 and x4/x4.

[Edit:] Ah here it is; http://hardocp.com/article/2010/08/23/gtx_480_sli_pcie_bandwidth_perf_x16/x16_vs_x8x8/
And as it turns out the performance penality is almost nil.

 

[T101]

I ran a test with 2 x 470GTX in sli on my rig. There was only a few games where the difference was enough to make it worth it. So I decided to put the second card back in the pc where it belong. I did not save any of my results, unfortunately. But on average I saw either 100% load on one card and 20% on the second, or 50% load on both cards (and nothing more), or the cpu was actually the bottleneck as the game was poorly threaded and ran only on one core. I tried every game I have on my steam account, which includes; Crysis, Alien Vs Predator, Operation Flashpoint: Dragon Rising, GTA IV, and many more.

I kind of regret not keeping the results after I was done testing that. But then, it was just to make the decision if I was going to run sli or not. Power consumption 24/7 vs performance gain was what I was looking at mostly.

 

[moonbogg]

Well, I can tell you what I experience. I had an E8400 and gtx 260-216 sli setup, and the cards were bottlenecked, especially in BFBC2. Ok, so that was off topic, but thought i'd spit it out there anyway.
I upgraded to i7-2600k @ 4.3 with 2x GTX570SLI. The cards do not appear bottlenecked in any way at all. I didn't run them on a nehalem rig so I can't compare, butI can say they seem to be all freed up and not bottlenecked at all. BFBC2 @1920x1200 all high 8xaa floats around 100-150 with occasional dips into the double digits, around 75 lowest with smoke and gunfire and all kinds of crap with 32 people in my face.
Too bad my rig will be trash for BF3. We will all need dual Keplers for that game I just know it.