Originally Posted by Mand12
What happens if I put my fancy shiny graphics card into the x16 slot, expecting it to use all 16 lanes, and then also install one of these into an x4 slot?
And as these proliferate, are we going to be in a situation where the PCIe lane restrictions of platforms like Z87 and Z97 are going to be a downside compared to things like x79 and X99?
What about PLX chips? How will they impact lane availability for GPUs alongside PCIe storage?
Your GPU will be bumped down to 8x connection, assuming the 4x slot is connected directly to the CPU. Assuming that you're using a platform with only 16 lanes, like LGA1150 and FM2+.
These products will initially be very niche, and mostly reserved for those that are already buying enthusiast/prosumer parts (LGA2011). Down the line as it becomes more mainstream (won't happen for at least 1-2 years), Intel has the choice of either keeping their mainstream platform at 16 lanes (as far as I know, Skylake will be 16 lanes, thus Skylake's successor will be 16 lanes as well, meaning 16 lanes for mainstream Intel until at least end of 2017) to push people towards their enthusiast platform, increasing the number of lanes available from the southbridge (if they don't move to complete SoC), or increasing the number of lanes direct from the CPU.
In a PLX mainstream Intel board, the PLX chip allows 32 lanes of connectivity from 16. However, there still is only 16 lanes for communication from the CPU. Assuming you have two GPUs and a 4x PCI-E SSD, the first GPU would run at 16x, second at 8x, and 4x for the SSD. If GPU traffic to the CPU does not exceed 12x, the SSD won't be affected. If it does, and the SSD is attempting to use 4x traffic, you'll run into a bottleneck as the GPUs and SSD now have to share bandwidth.