Networking and Storage Performance

We have recently started devoting a separate section to analyze the storage and networking credentials of the units under review. On the storage side, one option would be repetition of our strenuous SSD review tests on the drive(s) in the PC. Fortunately, to avoid that overkill, PCMark 8 has a storage bench where certain common workloads such as loading games and document processing are replayed on the target drive. Results are presented in two forms, one being a benchmark number and the other, a bandwidth figure. We ran the PCMark 8 storage bench on selected PCs and the results are presented below.

Futuremark PCMark 8 Storage Bench - Score

Futuremark PCMark 8 Storage Bench - Bandwidth

The MZVPV256 is the first NVMe SSD in the M.2 form factor. Given its PCIe 2.0 x4 link (the controller supports PCIe 3.0, but the unit connects to the PCIe 2.0 lanes in the Broadwell-U package) and the advantages of NVMe over AHCI, we expected our configuration to lead the storage charts. Our unexpected results made us run a couple of other artificial benchmarks (reproduced below).

The results from ATTO and CrystalDiskMark matches the specifications (accounting for the downgrade of the host link from PCIe 3.0 to PCIe 2.0). After discussion with Samsung, it turned out that the performance difference was due to the Microsoft NVMe driver creating FUA (Force Unit Access) I/O write commands. These FUA commands bypass the DRAM cache on the SSD and directly write to the flash, increasing the response time and also lowering bandwidth. For the same access traces, this situation does not happen with the Microsoft AHCI driver.

We observed something similar with the Intel SSD P3700 NVMe PCIe drive. With the Microsoft NVMe driver, the benchmark reported storage bandwidth around 320 MBps, while the Intel NVMe driver bumped that upwards of 500 MBps. So, we can conclude with a high degree of confidence that the Microsoft NVMe driver needs some fixes.

On the networking side, we restricted ourselves to the evaluation of the WLAN component. Our standard test router is the Netgear R7000 Nighthawk configured with both 2.4 GHz and 5 GHz networks. The router is placed approximately 20 ft. away, separated by a drywall (as in a typical US building). A wired client (Zotac ID89-Plus) is connected to the R7000 and serves as one endpoint for iPerf evaluation. The PC under test is made to connect to either the 5 GHz (preferred) or 2.4 GHz SSID and iPerf tests are conducted for both TCP and UDP transfers. It is ensured that the PC under test is the only wireless client for the Netgear R7000. We evaluate total throughput for up to 32 simultaneous TCP connections using iPerf and present the highest number in the graph below.

Wi-Fi TCP Throughput

In the UDP case, we try to transfer data at the highest rate possible for which we get less than 1% packet loss.

Wi-Fi UDP Throughput (< 1% Packet Loss)

Recent driver updates have enabled the Intel AC726x solutions to perform up to their advertised potential. Users facing issues with them are urged to update their drivers as well as router firmware in order to take advantage of the capabilities. The Broadwell-U NUC (NUC5i5RYK) uses the same Intel AC-7265 2x2 802.11ac soldered M.2 Wi-Fi card as the NUC5i7RYH. So, it is not surprising that the TCP and UDP throughputs are similar for the two systems.

Gaming Benchmarks HTPC Credentials
Comments Locked

66 Comments

View All Comments

  • ShieTar - Monday, April 20, 2015 - link

    Depends on what you mean with "the M.2 spec". I have a XP941 on a MSI Z97 Board, its measurably faster than the 840Pro I came from.

    I figure in this case the problem should be with the NUC board rather than the interface spec. The Plextor itself is not fast enough to profit from the interface, but it should be fast enough to work without noticable stuttering.
  • nutternatter34 - Tuesday, April 21, 2015 - link

    In my case we're talking about an X99 board, and an Intel 530 series M.2 drive (180gb). I compared that to each SSD I own. An Intel 320 series, a Samsung 830 series and the Crucial MX100 series. (120gb/256gb/512gb). M.2 was a disaster, what's worse there's barely anything to configure, it should just work.
  • meacupla - Monday, April 20, 2015 - link

    Oh, so this is limited to 45W?

    I wonder what the 3rd party makers could do with this chip if they expanded the power envelope and cooling capabilities.
  • Qwertilot - Monday, April 20, 2015 - link

    28w for the CPU/GPU, 17 for the rest even ;)

    I guess we'll find out what is possible when Broadwell K finally makes its much delayed appearance....
  • charea - Monday, April 20, 2015 - link

    So why use a 65W brick for a system limited at 45W? It doesn't make sense.
  • close - Monday, April 20, 2015 - link

    If you're going to power some more devices from the system then it helps to have a power source that's slightly oversized.
  • charea - Monday, April 20, 2015 - link

    Are you saying that the monitor is excluded from this limit? Was the test done without a screen included?
  • dave_the_nerd - Monday, April 20, 2015 - link

    USB devices pull up to 5w each. The test doesn't include a monitor. 30-40w draw for a 24" monitor is typical.
  • ganeshts - Monday, April 20, 2015 - link

    Right - 4x USB 3.0 ports need 20W. Add that to the 45W, and you are already at the 65W limit.

    Our stress test only loads up the CPU and GPU - it doesn't even do the internal storage stressing or WLAN stressing - these are bound to increase the power consumption a bit. That said, stressing those might actually result in the CPU not getting loaded as much as it does in our Prime 95 test.
  • ShieTar - Monday, April 20, 2015 - link

    And thats nicely keeping to the spec, there are ports and charger cables out there working with 2A => 10W.

Log in

Don't have an account? Sign up now