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

205 Comments

View All Comments

  • investlite - Monday, November 6, 2017 - link

    I bet you're still pissed we don't have flying cars. OMG, do you go to every new car unveiling and talk about how crappy each new car is because we were supposed to have flying cars by now? Reply
  • Gothmoth - Friday, October 27, 2017 - link

    when you do huge particle simulations you will want the fastest SSD you can get. Reply
  • ddriver - Friday, October 27, 2017 - link

    I've been doing that, for VFX as well as multiphysics simulations for over a decade. It has always been an in-memory thing. It doesn't seem they simulated it, it seems they read baked simulation data, and stored in some insanely inefficient manner at that.

    As I implied, this has got to be a new record in rigged benchmarks. Shame!
    Reply
  • extide - Friday, October 27, 2017 - link

    Sigh, the Optane drive didn't improve the performance of reading the data into the simulation. The simulation required (significantly?) more RAM than the system had. They put a big swap file on either a 960 PRO or on the Optane drive. It probably doesn't even matter where the simulation data was stored. Reply
  • ddriver - Friday, October 27, 2017 - link

    Well, they should have bought moar ram then. Then maybe they could have kept that CPU busy at 100% and get much better time.

    I mean it is not like hypetane offers terabytes of capacity. Topping at 480 GB - that's entirely doable in RAM. More expressive - sure, but nonetheless a perfectly sensible investment if you are doing such simulations. It will pay for itself, as RAM is tremendously faster, and also doesn't wear, at least nowhere nearly as much as xpoint does.
    Reply
  • extide - Friday, October 27, 2017 - link

    For that particular case, sure, maybe more ram is the way to go, but there are plenty of cases where the drive is better, like several of the ones I mentioned above. Most of those rely on the non-volatile aspect, which obviously RAM doesn't have. Reply
  • ddriver - Friday, October 27, 2017 - link

    As I said - it has its advantages and uses. I also said I might even buy it.

    And the only reason I call it hypetane is because intel shamelessly lied about it, and continues to cheat in order to make it look good even after it became evident that it is not anywhere nearly as good as they initially claimed, and call me old-fashioned, but I have a problem with that.

    It boggles the mind that people around here have such a problem with me just because I don't have my tongue up intel's rectum...
    Reply
  • ddriver - Friday, October 27, 2017 - link

    BTW, 1.1 billion particles, presuming the simulation is ran in FP64 mode, with x y and z coordinates for each particle would only require about 24 gigs of ram.

    Which raises the question, did they allocate each of those points on the heap or something?
    Reply
  • ddriver - Friday, October 27, 2017 - link

    Opsie, silly me, it would take another 24gb for the vector of force for each particle. Now it is a little more plausible that 64gb might not be enough. Reply
  • CaedenV - Sunday, October 29, 2017 - link

    Sure, in a perfect world you buy more RAM. But if you are in a situation where you don't have infinate cash and you can buy more RAM at $7/GB or an Optane SSD for ~$1-2/GB then Optane begins to look a bit more appealing. A 480GB drive running as cache for $600 vs 480GB of DDR4 at $3350.... that would make almost anyone thing twice.

    Or in the case of my work, we have a bunch of clustered servers, and we are maxxed out on ram but not yet ready to do a server upgrade (hoping to get 2 more years out of them), but we need more fast cache for a bunch of different applicaitons. The idea of running those caches on this kind of SSD sounds a lot more appealing than running on traditional SSDs.

    But yes, when we upgrade servers, we will simply have more RAM on board. That is the obvious solution. But when a motherboard can only hold 256GB of RAM and you need more... life is often about compromises, and Optane tech sounds like a good compromise. But what you would use this for in daily life or in a normal computer? Man, that totally beats me! This product is almost too cheap for what it is good for (business class SSDs typically cost more than $1.25/GB still and are far slower than consumer SSDs), and completely useless and overpriced for that they are advertising it for.
    Reply

Log in

Don't have an account? Sign up now