Random Read Performance

Our first test of random read performance uses very short bursts of operations issued one at a time with no queuing. The drives are given enough idle time between bursts to yield an overall duty cycle of 20%, so thermal throttling is impossible. Each burst consists of a total of 32MB of 4kB random reads, from a 16GB span of the disk. The total data read is 1GB.

Burst 4kB Random Read (Queue Depth 1)

The Optane SSD 900P doesn't break the record for QD1 random reads, but only because we've also tested the 32GB Optane Memory M.2, which is about two microseconds faster on average for each 4kB read. The Optane SSD 900P is still about 7 times faster than any flash-based SSD.

Our sustained random read performance is similar to the random read test from our 2015 test suite: queue depths from 1 to 32 are tested, and the average performance and power efficiency across QD1, QD2 and QD4 are reported as the primary scores. Each queue depth is tested for one minute or 32GB of data transferred, whichever is shorter. After each queue depth is tested, the drive is given up to one minute to cool off so that the higher queue depths are unlikely to be affected by accumulated heat build-up. The individual read operations are again 4kB, and cover a 64GB span of the drive.

Sustained 4kB Random Read

When longer transfers and higher queue depths come into play, the Optane SSD 900P passes the Optane Memory M.2 and remains more than 6 times faster for random reads than any flash-based SSD.

Both Optane devices more or less level off at queue depths of 8 or higher. The Optane SSD 900P saturates at about 1800 MB/s while the Optane Memory tops out around 1300 MB/s. The Samsung 960 PRO 2TB hasn't caught up by QD32, and doesn't surpass the QD1 random read performance of the Optane SSD until the Samsung reaches a queue depth of about 8.

Random Write Performance

Our test of random write burst performance is structured similarly to the random read burst test, but each burst is only 4MB and the total test length is 128MB. The 4kB random write operations are distributed over a 16GB span of the drive, and the operations are issued one at a time with no queuing.

Burst 4kB Random Write (Queue Depth 1)

The burst random write performance of the Optane SSD 900P is slightly higher than the Intel SSD 750 1.2TB, and about 14% faster than Samsung's fastest.

As with the sustained random read test, our sustained 4kB random write test runs for up to one minute or 32GB per queue depth, covering a 64GB span of the drive and giving the drive up to 1 minute of idle time between queue depths to allow for write caches to be flushed and for the drive to cool down.

Sustained 4kB Random Write

With higher queue depths in play, the Optane SSD 900P scales up faster than the Intel SSD 750 1.2TB, leaving the Optane SSD with a 7-10% lead over the Samsung 960s and Intel 750.

Samsung's 960 PROs and the larger 960 EVO all trail slightly behind the Optane SSD's random write performance for queue depths 1 to 4, then the Samsung drives level off and leave the Optane SSD with a substantial performance advantage at high queue depths. The Intel 750 is slightly faster at QD1 and QD2, but saturates at an even lower performance level than the Samsung 960s.

AnandTech Storage Bench - Light Sequential Performance
POST A COMMENT

204 Comments

View All Comments

  • ddriver - Friday, October 27, 2017 - link

    "MLC/TLC NAND treated like SLC"

    That sounds like "a snail treated as a cheetah". I bet feeding antelopes to a snail will make it as fast as a cheetah.

    There is already a huge gap in access performance between MLC and TLC. TLC drives turn pathetic the moment they run out of cache. It is physically impossible to store multiple bits and access as if it is a single bit. Neither in terms of performance, nor in terms of endurance.

    We haven't even seen what SLC is truly capable of, only the very early SSDs had SLC, and back then they were crippled by the primitive and tremendously under-powered controllers. At the medium level, SLC is insanely fast.
    Reply
  • MFinn3333 - Friday, October 27, 2017 - link

    OK, no that is simply not true.

    Fujtisu made a drive in 2014 entirely of Intel 25nm SLC and a Sandforce 2281 controller. It was called Fujitsu FSX 240GB (And 120GB). It kicked ass for it's class and time but it is nowhere and I mean nowhere near the speeds that are here or what you are claiming.

    I set a couple of them up in RAIeD-0 through both hardware and software RAID and while it did often saturat the bandwidth but only with sequential transfers. The only number that it could ever come close to matching anything here is when you are talking random writes. I could easily hit 220MB/s at 4K Random Write but it's Random Read speed was around 50MB/s at the best of times.

    SLC was and is awesome and I feel comfortable with doing horrible things to those drives (I defragged them for no reason, compiled code, did multiple virtual machines including Windows 3.11, and more benchmarking than any person ever should) but it's time is over.

    Get over it.
    Reply
  • ddriver - Friday, October 27, 2017 - link

    Oh wow, you are some kind of a tech genius, expecting raid to boost something other than sequential access.

    2014 is ancient history in the world of SSD controllers. SLC on the physical level is capable of 500-600 times better performance than what the "best" SLC controller could squeeze out of it.

    I am not saying this isn't the case for xpoint as well, I am just saying SLC is far more capable than what people can imagine.
    Reply
  • MFinn3333 - Friday, October 27, 2017 - link

    "We haven't even seen what SLC is truly capable of, only the very early SSDs had SLC, and back then they were crippled by the primitive and tremendously under-powered controllers. "

    That is what you wrote, not about SSD's a few years ago but the very early ones. You moved your own goalposts.

    SLC is about 4x faster than MLC. You are claiming it to be 500x while showing little to no evidence and roasting Intel for their claims. So either put up or shut up with your evidence.
    Reply
  • Reflex - Friday, October 27, 2017 - link

    SLC does not obey physics. Gotta take that into account. ;) Reply
  • chrnochime - Wednesday, November 1, 2017 - link

    No comment when the other guy mentions 2014 is not "very early SSD"? Come on now. Your prejudice against anything other than SLC is showing. We all know early SSDs go back much earlier than even 2007. Reply
  • edzieba - Saturday, October 28, 2017 - link

    If SLC were truly better and cheaper than PCM, then companies would be using it and undercutting the competition with their cheaper, faster drives with lower production costs. Reply
  • jospoortvliet - Friday, November 3, 2017 - link

    Indeed: https://xkcd.com/808/ Reply
  • extide - Friday, October 27, 2017 - link

    Regardless of the tests used here this thing is one of the fastest storage devices available PERIOD, and it's honestly priced pretty well. It's cheaper/GB than the first SSD I bought, in fact. I could see this being used for large swaps on servers that need a huge memory footprint for a lot cheaper than a shitload of RAM, or as a ZFS L2ARC or ZIL, or for hosting a ton of VM's or for running databases off of, etc.

    BUT you are ddriver, the king of cynicism, so I can at least say "I got what I expected."
    Reply
  • ddriver - Friday, October 27, 2017 - link

    I am also the king of "1000 times better means 1000 times better" ;)

    I too got what I expected, because I expected that "1000 times better" to be a lie.

    If you look at my comments, I am actually 100% objective about acknowledging the benefits of hypetane. Which is where my true biggest fault lies. How dare I be objective rather than expressing nothing short of complete awe and admiration?
    Reply

Log in

Don't have an account? Sign up now