ASRock's 5Gbit Journey and the Z270 SuperCarrier (Page 1)

ASRock's 5Gbit Journey

Some of you might have read our first sponsored article, a Z270 build guide, but today we are bringing you something different. This sponsored article will focus on a brand new technology, the new 5Gbit Ethernet standard, and two motherboards that carry it; the Z270 SuperCarrier and Fatal1ty Z270 Gaming i7.

Both of these ASRock motherboards carry one of the latest NBase-T controllers, the Aquantia AQC108. ASRock is the only vendor I know of that is offering 5Gbit networking to consumers, and they also made sure that these two motherboards offered three M.2 slots and extra storage options. Modern storage protocols and devices using NAND provide very high transfer speeds. While back in the day, 1Gbit made sense with HDDs, nowadays with SSDs the system can take advantage of faster networking, if only locally.

Why 5Gbit? The story began when vendors started to see a need for network speeds faster than 1Gbps, but couldn't justify the cost for 10Gbit infrastructure. Newer generation Wireless-AC cards were starting to exceed 1Gbps in throughput and 10Gbps over Ethernet requires Cat6a cabling, while most small offices and homes are wired with Cat5e. The NBase-T alliance was formed in 2014 and includes multiple OEMs such as Cisco and HPE and chip companies such as Intel and Aquantia. In fact, in 2015, IEEE decided to use the NBase-T specification to produce the 802.3bz standard that would make 2.5Gbps and 5Gbps speeds over 100m Cat5e and Cat6 possible.

In September 2016, 802.3bz was standardized and approved. Come early 2017, and ASRock is the first consumer motherboard vendor to integrate 2.5Gbps/5Gbps NICs into their Z270 and even X370 motherboards. The benefit of the new Aquantia 2.5/5Gbit NICs is that they will work with older Cat5e cabling, and will work with many of the NBase-T routers, switches, and access points shipped in 2015.

The Modern Bottleneck

A byte is equal to 8 bits, and it's important to differentiate Megabytes per second (MB/s) and Megabits per second (Mbps or Mbits). You should be able to divide numbers in Mbps by eight and then see how many MB/s you would get, however, we also have overhead, so in reality, we never get a clean cut bits/8 to get throughput in bytes.

For instance, the DMI in modern Intel chipsets operates at 32Gbps (x4 PCI-E 3.0), which would give us 4GB/s throughput, but because of the 128b/130b encoding we lose 2% in theory, and in reality after all overhead, we are limited to roughly 3.5GB/s. When we test three fast NVMe drives in RAID 0, we are stuck with this maximum 3.5GB/s limited by the DMI.

SATA (realistically around 550 MB/s limit) and LAN protocols work in the same way, so we don't get 625 MB/s, but closer to 500 MB/s in real-world testing of the 5Gbit protocol. Keep this in mind with the chart and connection protocols, since overhead can be found in all connections. The one outcome that should be simple to see is that there is a need for a long distance local transfer protocol that is faster than the 1Gbit NIC, can reach the same distances, and doesn't cost as much as 10Gbit infrastructure.

For some, the 5Gbit NICs on ASRock motherboards teamed with other 5Gbit systems and NBase-T switches could be the perfect options, especially if the building is already wired with Cat5e.