What's the difference between pci 2x8 and 1x16?

[rec_laratta]

I have no clue what it is and when I google it it's just ram this and ram that

 

[kanewolf]

This is a description of how PCIe lanes from the CPU are distributed. If you have one graphics card you get (1x16) -- one slot at x16 interface If you have two graphics cards connected, then you get (2x8) -- two cards each with an x8 interface.

 

[rhysiam]

A more detailed explanation in case it's helpful...

If you're after a bit more detail about how PCIe slots work and what the x1, x4, x16, etc mean, check out this article: http://blog.duropc.com/2013/the-difference-between-pcie-x1-x4-x8-x16-and-x32

In terms of your specific question, almost all modern desktops provide at least one full bandwidth x16 PCIe slot. This allows one device (usually a graphics card) to be connected into a PCIe slot that is wired directly to the CPU for the best possible performance. Most boards then provide a range of other PCIe slots (x4, x1, etc) for devices that don't require ultra high bandwidth. Those PCIe slots are (usually) connected to a chipset before the reach the CPU, so they're quite a bit slower, but still plenty fast enough for like sound cards, network cards and (most) PCIe SSDs. That all works fine as long as you only require 1 ultra high bandwidth device.

What if you want to install two graphics cards? Or a high end storage controller or a single card housing multiple PCIe SSDs? The performance of those sorts of cards can be seriously impacted by a slow x4 PCIe slot that has to go through the chipset before reaching the CPU.

One solution to this problem is getting a high end platform that offers more PCIe lanes from the CPU (X299 for Intel or AMD's Threadripper, for example). But these are usually much more expensive.

A much cheaper workaround is to allow the 16 lanes from the CPU to be split between two PCIe slots - so instead of a single x16 slot, you can provide 2 x8 slots. Obviously the bandwidth is still halved with only 8 lanes instead of 16, but that's often still plenty of bandwidth even for high end GPUs and you still have the advantage that the slots are directly connected to the CPU. This does add a little complexity to the motherboards, so entry level chipsets (B350, B360, etc) don't usually offer this feature. On those cheaper boards the 16 lanes are directly wired to a single slot and cannot be split. More expensive boards (X470, Z370, etc) on the other hand usually do offer this feature, so their PCIe slots will be labelled with something like 1x16, 2x8. What this means is that two particular slots are paired together. If only the top slot is populated it will get the full 16 lanes, however once the second slot is populated the board will automatically detect this and split the lanes between the slots, providing x8 to each.