AnandTech Storage Bench - The Destroyer

The Destroyer has been an essential part of our SSD test suite for nearly two years now. It was crafted to provide a benchmark for very IO intensive workloads, which is where you most often notice the difference between drives. It's not necessarily the most relevant test to an average user, but for anyone with a heavier IO workload The Destroyer should do a good job at characterizing performance. For full details of this test, please refer to this article.

AnandTech Storage Bench - The Destroyer (Data Rate)

Despite the lack of IO consistency, the BX100 does very well in our heaviest The Destroyer trace. It's easily faster than the MX100 and quite close to the higher-end SSDs as well in both data rate and latency.

AnandTech Storage Bench - The Destroyer (Latency)

AnandTech Storage Bench - The Destroyer (Latency)

The share of high latency IOs is very reasonable too, suggesting that the consistency is fine under real-world workloads.

AnandTech Storage Bench - The Destroyer (Latency)

AnandTech Storage Bench - The Destroyer (Power)

And finally the power consumption where the BX100 shines in. Even though it's not the fastest drive on the market, it's by far the most power efficient and the difference to the MX100 is nearly twofold. 

Performance Consistency AnandTech Storage Bench - Heavy
Comments Locked


View All Comments

  • mczak - Monday, April 13, 2015 - link

    Note though the M500, M550 and MX100 all suffer from a very serious issue wrt queued trim (possibly related to link power management). It is imho far more serious than the 840 EVO issues as it kills your data, but much fewer people are affected by it (only these using linux and of those most won't ever see it because the kernel was patched to blacklist the feature). This problem which was known for one and a half year or so is now finally acknowledged and fixed in firmware for the M550 and MX100 (still unfixed on the M500 where it was discovered, the MX200 had it fixed from start and the BX100 doesn't support queued trim in the first place apparently and has a different controller anyway).
    That does not exactly inspire confidence - Crucial claiming "hard to reproduce" or something along these lines, because apparently both Windows and Mac OS only use non-queued variant of trim (well I don't know for sure about Mac OS because this one doesn't use trim at all by default for non-apple built ssds), but it was very easy to produce failures with linux.
    If windows were to support queued trim tomorrow you'd see return rates soar to levels never seen before... (or probably not, because if that would be enabled now surely the feature would be blacklisted for these drives too).
    Not saying queued trim is an essential feature (it's clearly not), but if Crucial wasn't willing to actually test it with the only OS which supports it they probably shouldn't have enabled it in the first place...
    I have to say though this is not really enough to steer me away from Crucial SSDs (they indeed provide very good value overall), but keep that in mind if you think Samsung are the only ones with Firmware issues.
  • FunBunny2 - Thursday, April 16, 2015 - link

    Since AT went into the OCZ madness, back then, how about some investigation into this new firmware? From following the comments, it's something of a clusterfluff.
  • GregGritton - Monday, April 13, 2015 - link

    I hope that Anandtech returns to posting the log-based I/O consistency graphs.

    What you really want in the graph is an indication of how likely and severe of slowdowns you will experience, which means you want to graph the time per I/O operation (averaged over a short period like a second) rather than the number of I/O operations per second. Then, any outliers have significance. This means all of the useful information in the ops/second graphs are scrunched down at the bottom, where it is hard to see what the actual value is. (For example, the Crucial BX seems to generally have 2000 I/O ops/second, but it is hard to tell as the first line on the 1st graph is 10,000, and 5,000 on the 2nd.)

    The logarithmic graphs spread out the lower I/O ops/second values enough so that you could tell farily easily where there lower values were. Thus, they were a good compromise between a ops/second graph and a (milli/micro)seconds/op graph.
  • jamesnieves - Monday, April 13, 2015 - link

    my Aunty Isabella recently got a superb Dodge Challenger SRT8 by working part time off of a macbook air.
  • Laststop311 - Monday, April 13, 2015 - link

    Well if you want to put an ssd in your laptop this is the drive to get.
  • soccerharms - Tuesday, April 14, 2015 - link

    The 850 EVO just dropped to 179 for the 500GB on amazon with the bx100 at 187. I will be using this in a laptop. What does everyone think?
  • CknSalad - Tuesday, April 14, 2015 - link

    Samsung 850 Pro 256GB is $130 on Ebay just today! Just ordered mine!
  • JackF - Wednesday, April 15, 2015 - link

    I was surprised that the Mushkin Reactor 1T was not in the table comparisons. It was just reviewed by Kristian back in February and received an Anandtech Recommendation. It looks to me to fin right in this performance category and they have been running the 1T versions at $339.

    After deliberating, I just upgraded to a Samsung 850 EVO 1TB (at $350). It is a noticeable upgrade from my older Crucial M4 256GB.
  • Walkeer - Friday, April 17, 2015 - link

    perhaps I am blind, but the Samsung EVO seems either more power efficient or equal compared to BX100 from the idle power consumption graph, is that correct? That invalidates the final words.
  • leexgx - Monday, April 20, 2015 - link

    The bx100 is the most power efficient ssd at this time (I have the bx100 120gb soon)

    Only interesting thing here was devsleep used or was this just slumber (dipm+hipm only)

Log in

Don't have an account? Sign up now