Random Read Performance

The random read test requests 4kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test spans the entire drive, which is filled before the test starts. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.

Iometer - 4KB Random Read

It is unsurprising to see that the TLC-based 960 EVO has slower random read speeds than the MLC-based 950 Pro and 960 Pro, but the 960 EVO still manages to be faster than all the non-Samsung drives.

Iometer - 4KB Random Read (Power)

The 960 EVO's power consumption is essentially the same as Samsung's other drives, which puts it at an efficiency disadvantage to their MLC PCIe SSDs but more efficient than all the lower-performing drives.

As with Samsung's other SSDs, random read speed scales with queue depth until hitting a limit at QD16.

Random Write Performance

The random write test writes 4kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test is limited to a 16GB portion of the drive, and the drive is empty save for the 16GB test file. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.

Iometer - 4KB Random Write

The Samsung 960 EVO's random write speed is essentially tied with the 960 Pro and the OCZ RD400A, while the Intel 750 holds on to a comfortable lead.

Iometer - 4KB Random Write (Power)

The 960 EVO is not as power efficient as the 960 Pro, but it is still far better than everything else.

The scaling behavior of the 960 EVO is essentially the same as the 960 Pro: full performance is reached at QD4, and there's no indication of any severe thermal throttling.

AnandTech Storage Bench - Light Sequential Performance
Comments Locked

87 Comments

View All Comments

  • TheinsanegamerN - Tuesday, November 15, 2016 - link

    Is there currently any consumer software that can change these settings?
  • Billy Tallis - Tuesday, November 15, 2016 - link

    I use nvme-cli on Linux to manually test these settings, and there's a patch working its way toward a stable kernel release that will let supporting drives automatically make use of their various idle states. I'm not aware of any Windows tools that give the same degree of fine-grained control, but Intel's tools for enterprise SSDs and the 750 have some power management options.
  • TheinsanegamerN - Tuesday, November 15, 2016 - link

    That would be nice. I'd love to be able to set my 950 pro to a max of, say 2 watt on battery, or set the maximum speed to sata III speed (on battery), if it meant better battery life, and enable full speed when on mains power.
  • philehidiot - Tuesday, November 15, 2016 - link

    Yeh I noted the thermal limits kicking in and I did wonder if there's any major benefit from hacking these things, putting on some thermal goop, a big heat sink and a fan. Obviously it'll probably not help in the majority of cases but the mentalist part of me was pondering just how much performance there is to be unlocked.
  • TheinsanegamerN - Tuesday, November 15, 2016 - link

    The fan would be a bit of overkill, but the 950 pro showed some gains with a passive heatsink installed. The 960 would probably benefit even more.
  • philehidiot - Tuesday, November 22, 2016 - link

    Time to break out the liquid nitrogen. If anyone at work asks where it has all gone I'll just say I was remove the mother of all warts from a patient.
  • philehidiot - Tuesday, November 22, 2016 - link

    *removing. Bloody spelling.
  • nagi603 - Tuesday, November 15, 2016 - link

    A shame they dropped the write endurance to half of the 950!
  • TheinsanegamerN - Tuesday, November 15, 2016 - link

    TLC vs MLC
  • bull2760 - Tuesday, November 15, 2016 - link

    Why do you not show consistent results. In some tests you include the Intel 750 and on other you exclude it in the graph. You excluded the Intel 750 in the power consumption both times. Are you not able to calculate the power being used when the Intel drive is being used?

Log in

Don't have an account? Sign up now