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 burst random read performance of the Samsung PM981 is great by the standards of TLC SSDs, but is surpassed by several MLC-based drives, including the Phison E7-based Patriot Hellfire with planar MLC NAND.

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

On the longer random read test that includes some higher queue depths, the PM981 comes a bit closer to the standard set by Samsung's MLC drives, and it outperforms all the non-Samsung drives.

Both capacities of the PM981 show performance scaling with queue depth in the typical manner for a high-performance drive, though the 512GB model has passed an inflection point by QD32 and is approaching saturation.

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)

There are a few MLC-based SSDs that offer substantially higher burst random write performance than the Samsung PM981, but it is on par with most high-end drives including the Samsung 960 PRO.

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

On the longer random write test, the 1TB PM981 stands out with clearly higher performance than the Samsung 960 series could manage. The 512GB PM981 is slower but still definitely performing like a high-end drive.

The random write performance of the 1TB PM981 scales very well with increasing queue depth. As compared to the Samsung 960 series, it also reaches its plateau around QD8, but is providing much higher throughput by that point. The 512GB model runs out of SLC cache during portions of this test so its performance is much lower and less steady.

AnandTech Storage Bench - Light Sequential Performance
Comments Locked

53 Comments

View All Comments

  • romrunning - Thursday, November 30, 2017 - link

    @sleeplessclassics - "Today, even TLC can perform better than MLC/SLC just 2-3 generations ago due to better controllers."

    Well, the MLC-based 950/960 Pro still is basically beating all of the newer TLC drives. Even in SATA, my Sandisk Extreme Pro still beats all of the TLC drives.
  • sleeplessclassics - Friday, December 1, 2017 - link

    @romrunning, well you should have paid more attention in your high school English class.

    I am comparing present-gen TLC NAND with SLC/MLC that is two generations old.
    Of course, current gen 950/960 Pro MLC NAND with Polaris controllers will beat TLC NAND with Polaris controllers.

    I suggest you begin one of the simpler ones like Aesop's fables or maybe those illustrated children books will more your level. And while you are at it, try getting an IQ test as well
  • romrunning - Friday, December 1, 2017 - link

    You mad, bro?! ;)

    Lighten up a little, and act a little more objectively. Try clarifying the original statement or submitting more information to support your point without resorting to childish insults.

    For example, I will submit that you would have had to define "generations" and product lines before you put out generic statements like "Today, even TLC can perform better than MLC/SLC just 2-3 generations ago due to better controllers.". It's also hard to compare since the interface can change (like from SATA to NVMe).

    For my example, I will say again that my SATA Sandisk Extreme Pro will still beat newer TLC-based SATA drives from Sandisk. Also, I believe older Intel enterprise controllers (like a DC P3700) can still beat a terrible newer drive like the Intel 600p. There are even specialized drives from several "generations" of product lines ago that can beat some of the "newer" TLC-based drives in the same product line-up.

    However, obviously this is changing with NVMe-based drives, although it would harder to find a mfg with two "generations" of controllers on NVMe drives. Plus, we're getting into different tech like 3d XPoint, so TLC likely won't be around anymore a number of years in the future.

    I personally am looking forward to retail releases of Samsung's Z-NAND-based drives. It will be very interesting to see how they measure up performance-wise to Optane.
  • treecrab - Thursday, November 30, 2017 - link

    Sequential write numbers are off.
    1TB drive has fooled you - it has HUGE write buffer. Like 50GB huge. You need to check write speed second by second and on a much larger span (100 GB?)
  • mapesdhs - Thursday, November 30, 2017 - link

    Toms posted a review and pointed this out with their sustained tests.
  • Drumsticks - Thursday, November 30, 2017 - link

    It feels like we haven't seen new high end drives from Samsung in a while (not that they're really heavily in need given the performance on tap already). It'll be nice to see another round of products coming out of them. Thanks for the review!

    Side note: Would it be possible to, in future SSD reviews, add those buttons that change the graphs based on capacity for the different storage metrics? Perhaps a button for "All SSDs," "250GB," "500GB," "1TB+" or something. Performance can vary wildly across capacities, and it would be a nice way to sort through all of the 500GB class drives that you've reviewed for example. The only outlier would be Optane since it isn't quite as dominated by the amount of parallel dies you can add.
  • mapesdhs - Thursday, November 30, 2017 - link

    Good idea; it's bizarre how sometimes the 960 EVO looks terrible compared to the 1TB version, and sometimes the other way round. Steady state is particularly bad, it's why I stuck with hunting for 950 Pros instead, which also have their own boot ROM and thus work ok on older mbds.
  • Kastriot - Thursday, November 30, 2017 - link

    200$ for 512GB i have intel 530 480GB which i bought on ebay for 80 euros 3 years ago and still working great so ssd prices are insane like memory+gpu+intel cpu-s prices together, dark times for PC desktop owners unless you have dosh :)
  • mapesdhs - Thursday, November 30, 2017 - link

    GN has commented on this quite a lot recently, it is indeed a sucky time to build a new PC. Hence why I make the most of used parts (or new ones via normal auction) until the need for something better really is paramount. Bagged another 840 Pro 256GB recently for a good price; pity old models like this are not included in newer product reviews, I bet they'd put newer products to shame. For a while the old Vector was retained in newer reviews, but then it vanished, probably because it just looked too good compared to the latest tech. The same likely applies to the Neutron GTX, Vertex4 and various other models, at least in the SATA world anyway. If one can though, it's better to go NVMe, the SM951 and SM961 are rather good.
  • bcronce - Thursday, November 30, 2017 - link

    It looked mostly on par until the "mixed" results. Nice!

Log in

Don't have an account? Sign up now