Looks like ASUS decided to make a daughterboard that will fan out two x16 PCI-E slots into four, using two NF200 chips:
[url=“ASUS ROG Xpander might be the craziest four-way SLI solution you've ever seen - Mainboard - News - HEXUS.net”]ASUS ROG Xpander might be the craziest four-way SLI solution you've ever seen - Mainboard - News - HEXUS.net
Good luck getting this to work in most computer cases. :)
I believe we hashed this topic to death last night in this thread.
[url=“The Official NVIDIA Forums | NVIDIA”]The Official NVIDIA Forums | NVIDIA
But I am more than interested in yalls opinion on this as well. looking forward to the replies.
EG
From reading the reviews, it seemed this was just an SLI solution and had nothing to do with CUDA.
sure it does, 32 lanes across two slots in, 32 lanes across four slots out
It’s basically an option to give you full bandwidth to four GPUs individually if your motherboard does not already use some kind of PCI-Express switch already to do it. This could be handy in cases where you don’t expect all GPUs to transfer at the same time.
Although I am finding it hard to verify (the Asus web site returns to the home page after clicking on the “Rampage III Extreme” link), I believe that the original motherboard does the same thing without the benefit of the NF200s. Presumably a CUDA user would have paid the $20 extra for the P6T7 which includes the two NF200s on the motherboard and therefore wouldn’t need this funky “daughter card”.
Found the page on the Asus site:
http://usa.asus.com/product.aspx?P_ID=jy0u…&templete=2
The PCI-Express slots do the usual splitting of x16 slots to x8 slots as you install more cards. Apparently this daughterboard was a tactic to avoid having to pay for NF200 switches on a motherboard that people might not use.
But yeah, for the price of this motherboard, it would make a lot more sense to just buy the P6T7.