Drive and Audio Performance Results

I recently updated my testing for HDD performance. We now test using four different interfaces: SATA 6 Gb/s ports, NMVe M.2 plugs, USB 3.0 front panel plugs, and USB 3.1 Type-C ports.

USB 3.1 Type-C ports are the new small, reversible plugs found on the back panel of most recently released motherboards that look similar to the Thunderbolt-type plugs found on many devices these days. USB 3.0 plugs are tested using the header closest to the front panel since some motherboards have multiple ports. SATA 6 Gb/s ports are tested using internal cabling connected the board's PCH/chipset/FCH. NMVe is tested using the uppermost port since some boards have multiple ports.

The numbers reported below are the maximum MB/s HDTune's basic read test of the entire drive reports. The NMVe test uses a Samsung 950 PRO 256 GB with no data present on the device, while the other tests use a Crucial BX200 256 GB drive with about 60 GB of data present (all of my testing tools and applications are stored on this drive and are then copied to the OS drive for usage). Once set for testing, neither drive gets any data written to it and is freshly trimmed. USB 3.1 Type-C and USB 3.0 testing is done using the same external dock, a StarTech S251BPU313. The cable between the dock and the tested PC is swapped out to the cable for the tested interface.

M.2 devices are connected to the system in different ways on different motherboards (some directly to the CPU, some via the chipset/FCH/PCH), and USB devices are connected using different controllers depending on which motherboard is used. I try to take pictures of those USB controllers when possible, but cooling sometimes gets in the way and not all motherboard makers declare how their M.2 ports are connected to the system. A direct connection to the CPU's root PCIe complex is obviously the fastest solution available, but based on which platform is used, such isn't always possible.

In the past few months, I have tested well over 100 different drive connections to the system and have come to the conclusion that reporting their maximum bandwidth is the best option since some drives feature dynamic thermal throttling, which can affect their average bandwidth. With both USB tests, the maximum number reported usually occurs within the first 25% of the drive's read, and the average bandwidth is usually about 93% of the maximum, but doesn't occur until after the maximum peak. The NVMe Samsung 950 PRO 256 GB drive has a very straight line on CPU-connected ports and a quite jagged line on PCH connections. The Crucial BX200 initially performs a bit slower compared to maximums and averages at the beginning of the test, and the maximum spike levels out shortly after, while the average occurs for the remainder of the drive's read test (due to the data recorded to it).

Numbers reported in such a way show how much effort a board maker puts into circuit optimization for HDD data, which is the slowest part of a system these days, and being the slowest, should in my eyes be the most crucial and most optimized.

HDTune Pro (SATA 6Gb/s)

HDTune Pro (USB3.0 Front Panel)

HDTune Pro (USB 3.1 Type-C)

HDTune Pro (NVME M.2)

RightMark Audio Analyzer

The RMAA test results from the Gigabyte X99-Ultra Gaming are as expected; Very Good. The choice of a Realtek ALC1150 here instead of a Creative CODEC certainly makes me happy as I tend to use digital output to a proper HT amp, and Creative chips aren't so good with that. The Realtek ALC1150, though, works great!