USB 3.0 Performance
We measure the speed of the USB 3.0 interface using CrystalDiskMark 3.0 and an external Buffalo DriveStation HD-HX1.0TU3 drive. The diagram below shows the results of all mainboards with USB 3.0 we have tested so far and we can see them all split up into two groups. One group delivers higher USB 3.0 performance. These are the Socket AM3 models which USB 3.0 controller is connected to a PCI Express 2.0 bus and the models with Intel chipsets where the USB 3.0 controller is connected via a PLX bridge. The Intel-based mainboards without that bridge connect the controller to a first-generation PCI Express bus which limits their USB 3.0 bandwidth. The results are sorted by read speed but we can see the same thing at writing. The ASRock H55 Extreme3 is the only exception, but it implements USB 3.0 with a Fresco Logic FL1000G controller rather than with a NEC D720200F1 chip as in the rest of the mainboards, which must be the reason for the difference.
The small GA-H55N-USB3 mainboard doesn’t have a PLX bridge, so it can be seen in the bottom part of the diagram among those mainboards which USB 3.0 performance is limited by the bandwidth of the PCI Express 1.0 bus. Its USB 3.0 interface is as fast as 120 and 95 MBps at reading and writing, respectively. You may have noticed, however, that the GA-H55N-USB3 appears twice in the diagram and its performance in the second case is as high as that of the Socket AM3 mainboards or of the Intel-based mainboards with a PLX bridge: about 130 MBps at reading and 120 MBps at writing. What does it mean?
The user manual from Gigabyte contains a component layout chart of the mainboard. It shows that the onboard USB 3.0 controller can work either with the PCI Express 1.0 bus provided by the chipset or with the PCI Express 2.0 bus which controller resides in the CPU.
So, when you install an external graphics card, it is supported by the CPU’s PCI Express 2.0 controller while USB 3.0 devices are connected via PCI Express 1.0. But if there is no discrete graphics card in the system and the CPU-integrated graphics core is used instead, the second-generation PCI Express bus can be utilized to provide the full USB 3.0 bandwidth. Our tests confirm that this implementation is effective. We can also suppose that the Gigabyte GA-H55M-USB3 and GA-H57M-USB3 mainboards which use the same dual connection of the USB 3.0 controller can deliver higher USB 3.0 performance if benchmarked without an external graphics card.
The lack of a PLX bridge on Gigabyte’s H55/H57-based mainboards can be justified as their users are likely to rely on the integrated graphics and thus will have the highest USB 3.0 bandwidth possible. Most importantly, this makes such mainboards less complex and, consequently, cheaper. And even if the user installs a discrete graphics card, the mainboard’s USB 3.0 performance is going to lower but a little and will still be much higher than the speed of USB 2.0.