Introduction

It's been one year since our last PCI-Express scaling article in which we used GeForce GTX 980s. Not much has happened since then in terms of slot technology and bandwidth. Just like in recent years, almost everybody is using PCI-Express x16 3.0, and while Intel's Enthusiast Platform (Haswell-E) has plenty of lanes for even multiple VGA cards, it's the exception as every other motherboard still splits the x16 3.0 slot into two x8 3.0 slots for a dual-GPU setup.

PCI-Express 3.0 saw its debut on Intel's Sandy Bridge-E processors, but was adopted into the mainstream when the Ivy Bridge architecture was released in 2012. Since then, PCI-E 3.0 has been used on Haswell and now Skylake.



While PCI-Express 1.0 pushes 250 MB/s per direction, PCI-Express 2.0 pushes 500 MB/s, and PCI-Express 3.0 doubles that to 1 GB/s. The resulting absolute bandwidth of PCI-Express 3.0 x16, 32 GB/s, might seem like overkill, but the ability to transfer that much data per lane could come to the rescue of 8-lanes (x8) and 4-lanes (x4) configurations.

In this review, we test the impact of running AMD's Radeon Fury X on our new Windows 10 test bench using electrical Intel Skylake PCI-Express x16, x8, and x4 slots. We made a point of testing all three generations of the PCI-Express interface: 1.1, 2.0, and 3.0.



To adjust the number of lanes available to the graphics card, we used common plastic adhesive tape to cover up and disable lanes, mimicking the different slot widths.

For the graphs on following pages, we've colored same-bandwith configurations in the same color, as an easy visual reference. PCIe x8 3.0, for example, offers just as much bandwidth as PCIe x16 2.0, which is why both are of the same color.

We also tested PCI-Express x4 3.0 bandwidth via the chipset separately, which usually powers the third PCIe slot on most motherboards.