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 short-burst QD1 random read speed of the ADATA XPG SX950 is good, though Samsung still has a clear lead. The SX950 is 16% faster than the Crucial BX300 that uses the same NAND and controller.

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

With a longer test and higher queue depths, the ADATA SX950's standing drops somewhat, but it still outperforms most TLC SSDs and the Crucial BX300.

Sustained 4kB Random Read (Power Efficiency)

The power efficiency of the SX950 during random reads is decent but not outstanding. The planar MLC-based PNY CS2211 beats the SX950 on both performance and power efficiency, but otherwise the SX950 is only beat by a few other 3D NAND SSDs.

For an MLC-based SSD, the ADATA SX950 does a poor job of scaling up random read performance with higher queue depths. Its QD1 performance is fine, but by QD8 it's lagging behind many of its competitors. The Crucial BX300 is marginally slower until QD32, where the SX950's performance starts to level off while the BX300 gets very close to the peak speeds reached by the best SATA SSDs.

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)

After seeing clear indicators from the ATSB tests that the ADATA SX950 is very aggressive with SLC write caching, it's good to see that the QD1 burst random write performance is near the top of the charts.

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 with higher queue depths included, the ADATA XPG SX950's random write score is unimpressive: barely faster than the TLC-based ADATA SU800 and well behind the Crucial and Samsung drives.

Sustained 4kB Random Write (Power Efficiency)

In spite of mediocre performance on the longer random write test, the SX950 delivers great power efficiency, in second place only slightly behind the Crucial MX300.

The SX950's random write speed mostly levels off in the second half of the test as queue depths grow beyond 4, and even at QD32 the SX950 is not really close to saturating the SATA bus. But at QD1 and QD2, its performance is as good as any SATA drive and its power consumption is unbeatable.

AnandTech Storage Bench - Light Sequential Performance
Comments Locked


View All Comments

  • menthol1979 - Monday, October 9, 2017 - link

    Oh dear God, another SSD that has absolutely no reason of existence. Really bored to see another SSD that gets pwned by 850 EVO (leave the PRO). I wonder if manufacturers actually test and benchmark their products before driving them to market.
  • Stochastic - Monday, October 9, 2017 - link

  • ddriver - Monday, October 9, 2017 - link

    Sadly, very little of what humans do is because it is necessary or it makes sense.
  • Reflex - Monday, October 9, 2017 - link

    @ddriver And yet you continue posting...
  • Samus - Monday, October 9, 2017 - link

  • ddriver - Monday, October 9, 2017 - link

    Moot point, as I don't identify with the human herd. Cattle mentality and the accompanying irrational behavioral patterns don't sit well with me. Which is also why I refer to humans in third person, a subtle nuance an intelligent person would have read into.

    But not you though, you perfectly fit the profile, seeing how once again you fail at getting stuff or making sense ;) But still, an understandable effort, you are probably still hurting by that chain of pwnage. And it's only parroting cliches because you really cannot do better.

    You humans, sometimes I am amazed you made it this far. And since you wouldn't get the nuance, there are two contexts to that, the first being that you still haven't succumb to your stupidity, and the second being "this far into devolution". I suppose that's why you cherish the establishment and its mediocrity so much, even if it is what pushes you to regress into cattle, you still get to survive, suckling at its toxic tit. It's your mommy, that's what your infant mind can identify it as, not as what it really is.
  • ddriver - Monday, October 9, 2017 - link

    And just in case you are perplexed how me responding to your post is something that makes sense, since you obviously can't get all this, it is quite simple - you are not the intended audience, just the means of making a point for the occasional few that can get it ;)
  • vgray35@hotmail.com - Tuesday, October 10, 2017 - link

    This quote "Sadly, very little of what humans do is because it is necessary or it makes sense," is a telling feature reveal of this AI Cyborg miscreant, who apparently has a deep rooted need for focusing on humans, describing humans, engaging humans, belittling humans; and it's apparent its existence and glorified self aggrandizement is defined solely on the existence lowly humans, as evidenced by the closing statement "you are not the intended audience ...".

    Sadly, very little of what this AI cyborg does makes sense. Prattle over product reviews is merely pretense of know how . Sadly no one has yet found the power down switch for this AI cyborg. For as much as it exudes disdain for humans, yet its very reason for being relies entirely on the necessity for engaging with them, to establish meaning in its miserable existence. These posts are its food, and a belittlement posture its means of self aggrandizement compensating for its low class software programming. The prattle is evidence that surely this really is no human (as it itself claims). It needs a firmware upgrade and an implant to put it out of its misery. I wish scientists would stop creating such experimental specimens for their own misguided research.
  • mapesdhs - Wednesday, October 11, 2017 - link

    vgray, that was awesome. 8)
  • svan1971 - Wednesday, October 11, 2017 - link

    Bravo !

Log in

Don't have an account? Sign up now