Each PCI-E slot only has 16 pipelines available. Therefore, each GPU is gettingo nly 8 in quad SLi. If you have jsut the one card, each GPU is still only getting 8.

Okay, I just checked my motherboard manual, as I thought, "In SLI mode, the PCI Express x16 slots work at the full bandwidth of PCI Express x16 on each slot for a combined bandwidth of x32":-roll

Exactly. Your TOTAL bandwidth is 32 pipelines. Trust me dude, I have the same board. You have 4 GPU's (2 per card.) 32/4=8 Each GPU gets 8 pipelines to work with, where as on a normal SLi setup, each GPU gets 16 to work with...

It's a moot point anyway from an FS perspective, as SLI ain't taken advantage of!Gary

#### man... It works great! :D

I'm sure it works, it just doesn't work any better with FS than if you didn't have it at all. It is a wasted expense if done purely for FS performance.Gary

:-wave That's dudette

From my discussion with NV devrel engineers, this just isnt true.Reading the NV GPU Programming Guide, Chapter 8, pg 59:When the two graphics boards are then linked via an external SLI bridge connector, the driver recognizes the configuration and allows you to enter SLI Multi-GPU mode. In SLI Multi-GPU mode the driver configures both boards as a single device: the two GPUs look like a single logical device to all graphics applications.The main issue is when does the fillrate start to show up in the cost equation so that the GPU cost starts to overcome the CPU bound nature of the app. Somewhere north of 16x12, there should start to be enough fillrate to show some advantage. Depending on CPU and memory config, it might take 19x12 or more before there is any advantage.Another thing to remember is to disable the frame rate limiter when testing this, and to disable VSync. These 2 pieces of advice are in the NV documents.I dont see any advice in the NV docs that FSX violates, and its pretty clear from the 1st part that its all done in the driver. So some have reported here that SLI is showing some benefit.