Random Read Performance

One of the major changes in our 2015 test suite is the synthetic Iometer tests we run. In the past we used to test just one or two queue depths, but real world workloads always contain a mix of different queue depths as shown by our Storage Bench traces. To get the full scope in performance, I'm now testing various queue depths starting from one and going all the way to up to 32. I'm not testing every single queue depth, but merely how the throughput scales with the queue depth. I'm using exponential scaling, meaning that the tested queue depths increase in powers of two (i.e. 1, 2, 4, 8...). 

Read tests are conducted on a full drive because that is the only way to ensure that the results are valid (testing with an empty drive can substantially inflate the results and in reality the data you are reading is always valid rather than full of zeros). Each queue depth is tested for three minutes and there is no idle time between the tests. 

I'm also reporting two metrics now. For the bar graph, I've taken the average of QD1, QD2 and QD4 data rates, which are the most relevant queue depths for client workloads. This allows for easy and quick comparison between drives. In addition to the bar graph, I'm including a line graph, which shows the performance scaling across all queue depths. To keep the line graphs readable, each drive has its own graph, which can be selected from the drop-down menu.

I'm also plotting power for SATA drives and will be doing the same for PCIe drives as soon as I have the system set up properly. Our datalogging multimeter logs power consumption every second, so I report the average for every queue depth to see how the power scales with the queue depth and performance.

Iometer - 4KB Random Read

While the other SSDs hover at 60-90MB/s for random reads, the SM951 provides a rather noticeable upgrade at 108MB/s. 

Samsung SM951 512GB

Looking at the performance more closely reveals that the SM951 delivers better performance at all queue depths, although obviously the difference is at high queue depths where the SM951 can take advantage of the faster PCIe interface. The SM951 actually does over 150K IOPS when the MB/s is translated into throughput.




Iometer - 4KB Random Write

Random write performance is equally strong. The line graphs shows how the SM951 shifts the whole curve up, implying a performance increase at all queue depths. Especially the performance at queue depths of 1 and 2 are noticeably better than on other drives.

Samsung SM951 512GB
AnandTech Storage Bench - Light Sequential Performance


View All Comments

  • Kevin G - Tuesday, February 24, 2015 - link

    "I also verified that the SM951 is bootable in tower Mac Pros (2012 and earlier)."

    Excellent. The old 2010/2012 towers continue to show that being expandable provides long term benefit. I'm glad that I picked up my tower Mac Pro when I did.

    Now to find a carrier that'll convert the 4x PCIe 3.0 link of the M.2 connector to an 8x PCIe 2.0 link for a Mac Pro. (Two two M.2s to a single 16x PCIe 2.0 link.)
  • extide - Tuesday, February 24, 2015 - link

    You will need a PLX chip to do that, you can't just put 2 x4 devices into an x8 slot... Reply
  • jimjamjamie - Wednesday, February 25, 2015 - link

    It's pretty hilarious how many people drink the shiny plastic trash bin kool-aid. Reply
  • Tunnah - Tuesday, February 24, 2015 - link

    I'm not super knowledgeable on the whole thing, but isn't NVMe really only a big deal for enterprise, as it's more a benefit for multi drive setups ? Reply
  • Kristian Vättö - Tuesday, February 24, 2015 - link

    It's of course a bigger deal for enterprises because the need for performance is higher. However, NVMe isn't just a buzzword for the client space because it reduced the protocol latency, which in turn results in higher performance at low queue depths that are common for client workloads. Reply
  • knweiss - Sunday, March 1, 2015 - link

    Kristian, did you ever test how much influence the filesystem has? I would love to see a filesystem comparison on the various platforms with NVMe drivers (Windows, Linux, FreeBSD, etc). Reply
  • The_Assimilator - Tuesday, February 24, 2015 - link

    Hopefully NVMe will be standard on SSDs by the time Skylake and 100-series chipsets arrive. Reply
  • sna1970 - Tuesday, February 24, 2015 - link

    What is the point of this expensive drive when you can have the same numbers using 2 SSD in Raid 0 ?

    and please no one says to me risk of Data Loss .. SSD are not mechanical and the chance of loosing 1 SSD is the same of 2 of them.
  • Kristian Vättö - Tuesday, February 24, 2015 - link

    RAID only tends to increase high QD and large IO transfers where the IO load can easily be distributed between two or more drives. Low QD performance at small IO sizes can actually be worse due to additional overhead from the RAID drivers. Reply
  • dzezik - Tuesday, February 24, 2015 - link

    Hi sna1970. You misses Bernouli's "introduced the principle of the maximum product of the probabilities of a system of concurrent errors" it is quite old 1782 but is is still valid. Have You ever been in school. Do You have mathematics classes? Reply

Log in

Don't have an account? Sign up now