AnandTech Storage Bench - The Destroyer

The Destroyer is an extremely long test replicating the access patterns of very IO-intensive desktop usage. A detailed breakdown can be found in this article. Like real-world usage, the drives do get the occasional break that allows for some background garbage collection and flushing caches, but those idle times are limited to 25ms so that it doesn't take all week to run the test. These AnandTech Storage Bench (ATSB) tests do not involve running the actual applications that generated the workloads, so the scores are relatively insensitive to changes in CPU performance and RAM from our new testbed, but the jump to a newer version of Windows and the newer storage drivers can have an impact.

We quantify performance on this test by reporting the drive's average data throughput, the average latency of the I/O operations, and the total energy used by the drive over the course of the test.

ATSB - The Destroyer (Data Rate)

The Intel SSD 660p manages an average data rate on The Destroyer that is only slightly slower than the Crucial MX500 mainstream SATA SSD and the Kingston A1000 entry-level NVMe SSD. It's a step up from the performance of the 512GB Intel SSD 600p, and more than three times faster than the DRAMless Toshiba RC100.

ATSB - The Destroyer (Average Latency)ATSB - The Destroyer (99th Percentile Latency)

The average and 99th percentile latency scores for the Intel SSD 660p are quite poor by NVMe standards and significantly worse than the Crucial MX500, but the latency isn't completely out of control like it is for the Toshiba RC100.

ATSB - The Destroyer (Average Read Latency)ATSB - The Destroyer (Average Write Latency)

The average read latency from the 660p during The Destroyer is comparable to other low-end NVMe SSDs and better than the 600p or Crucial MX500. The average write latency is more than twice that of the MX500 but lower than the 600p and RC100.

ATSB - The Destroyer (99th Percentile Read Latency)ATSB - The Destroyer (99th Percentile Write Latency)

The 99th percentile read latency from the Intel SSD 660p on The Destroyer is significantly worse than any other NVMe SSD or the Crucial MX500 SATA SSD, but the 99th percentile write latency is an improvement over the 600p and does not show the extreme outliers that the Toshiba RC100 suffers from.

ATSB - The Destroyer (Power)

The energy usage of the 660p during The Destroyer is a bit better than average for NVMe SSDs, though still quite a bit higher than is typical for SATA SSDs. The 660p is less power hungry than most NVMe drives and slower, but not enough to drag out the test for so long that the power advantage disappears.

Introduction AnandTech Storage Bench - Heavy
POST A COMMENT

92 Comments

View All Comments

  • Spunjji - Wednesday, August 8, 2018 - link

    It was pretty obvious that the 840 was affected because it used the same NAND as the 840 Evo, just without the caching mode. It was also pretty obvious that Samsung didn't care because it was "old" so they never properly fixed it. Reply
  • OwCH - Wednesday, August 8, 2018 - link

    Ryan, I love that you will. It is not easy for the user to find real world data on these things and it is, at least to me, information that I want before making the decision to buy a drive.

    Looking forward to it!

    Thanks!
    Reply
  • Solid State Brain - Tuesday, August 7, 2018 - link

    The stated write endurance should already factor data retention, if it follows JEDEC specifications (JESD219A). For consumer drives, it should be be when the retention time for freshly stored data drops below 1 year after the SSD is powered off, at 30°C. Reply
  • BurntMyBacon - Wednesday, August 8, 2018 - link

    The Samsung 840 EVO would like to have a word with you. Reply
  • eastcoast_pete - Wednesday, August 8, 2018 - link

    Yes, it should factor data retention, and it should follow JEDEC specs. The problem is the "should". That doesn't mean it or they do. I found that "Trust but verify" is as important in IT as it is in life. Even the biggest names screw up, at least occasionally. Reply
  • IntenvidiAMD - Tuesday, August 7, 2018 - link

    Are there any reviewers that do test that? Reply
  • DanNeely - Tuesday, August 7, 2018 - link

    Over 18 months between 2013 and 2015 Tech Report tortured a set of early generation SSDs to death via continuous writing until they failed. I'm not aware of anyone else doing the same more recently. Power off retention testing is probably beyond anyone without major OEM sponsorship because each time you power a drive on to see if it's still good you've given its firmware a chance to start running a refresh cycle if needed. As a result to look beyond really short time spans, you'd need an entire stack of each model of drive tested.

    https://techreport.com/review/27909/the-ssd-endura...
    Reply
  • Oxford Guy - Tuesday, August 7, 2018 - link

    Torture tests don't test voltage fading from disuse, though. Reply
  • StrangerGuy - Tuesday, August 7, 2018 - link

    And audiophiles always claim no tests are ever enough to disprove their supernatural hearing claims, so... Reply
  • Oxford Guy - Tuesday, August 7, 2018 - link

    SSD defects have been found in a variety of models, such as the 840 and the OCZ Vertex 2. Reply

Log in

Don't have an account? Sign up now