Performance Consistency

We've been looking at performance consistency since the Intel SSD DC S3700 review in late 2012 and it has become one of the cornerstones of our SSD reviews. Back in the days many SSD vendors were only focusing on high peak performance, which unfortunately came at the cost of sustained performance. In other words, the drives would push high IOPS in certain synthetic scenarios to provide nice marketing numbers, but as soon as you pushed the drive for more than a few minutes you could easily run into hiccups caused by poor performance consistency. 

Once we started exploring IO consistency, nearly all SSD manufacturers made a move to improve consistency and for the 2015 suite, I haven't made any significant changes to the methodology we use to test IO consistency. The biggest change is the move from VDBench to Iometer 1.1.0 as the benchmarking software and I've also extended the test from 2000 seconds to a full hour to ensure that all drives hit steady-state during the test.

For better readability, I now provide bar graphs with the first one being an average IOPS of the last 400 seconds and the second graph displaying the IOPS divided by standard deviation during the same period. Average IOPS provides a quick look into overall performance, but it can easily hide bad consistency, so looking at standard deviation is necessary for a complete look into consistency.

I'm still providing the same scatter graphs too, of course. However, I decided to dump the logarithmic graphs and go linear-only since logarithmic graphs aren't as accurate and can be hard to interpret for those who aren't familiar with them. I provide two graphs: one that includes the whole duration of the test and another that focuses on the last 400 seconds of the test to get a better scope into steady-state performance. These results are for all drives 240GB and up.

Steady-State 4KB Random Write Performance

Steady-state performance is fairly good, although given that the SX930 employes 12% over-provisioning it's expected to perform better than drives with only 7%. 

Steady-State 4KB Random Write Consistency

Consistency, on the other hand, isn't that good. The good news is that the JMF670H is more consistent than the SM2246EN in the BX100, but there's a long way to achieve Samsung-level consistency. 

ADATA XPG SX930
Default
25% Over-Provisioning

As the SX930 isn't very consistent, we saw performance dropping below 1,000 IOPS. It seems that consistency is one of the areas that truly separate Samsung and Marvell drives from the rest because JMicron, Silicon Motion and Phison based drives all have trouble sustaining steady performance. Basically, the baseline performance is about 1,000 IOPS for all controllers with frequent peaks of +5,000 IOPS, whereas Samsung and Marvell drives have very little variation in performance. I suspect the controller design itself has something to do with this because Samsung and Marvell controllers are all multi-core, but JMicron and Silicon Motion rely on single-core designs for higher cost efficiency. 

ADATA XPG SX930
Default
25% Over-Provisioning
Introduction, The Drives & The Test AnandTech Storage Bench - The Destroyer
POST A COMMENT

67 Comments

View All Comments

  • Impulses - Thursday, July 16, 2015 - link

    The 500GB EVO has actually been $164 at Amazon for like a week, making it the cheapest one on the chart lol... Not sure why they list it at $178.

    The 1TB just went back down to $340 too... I want two in the long run, trying to decide whether to grab just one now and wait for better BF pricing or just grab both now.
    Reply
  • MrSpadge - Thursday, July 16, 2015 - link

    Yes, it's called the BX100. ARC100 is pretty decent, too. Reply
  • zodiacfml - Friday, July 17, 2015 - link

    Nothing till other manufacturers release their 3D NAND versions. Don't forget about pricing too. ADATA couldn't compete because they don't design controllers or manufacture chips. They buy parts and assemble them, making it costly. Reply
  • Samus - Saturday, July 18, 2015 - link

    Considering the MX100 is cheaper and more reliable with about the same performance, I think I just answered your question.

    Samsung TLC drives can not be trusted, specifically the 840 EVO and 850 EVO.
    Reply
  • Flunk - Thursday, July 16, 2015 - link

    When it comes to budget drives right now, if it's not cheaper than a Samsung 850 EVO I'm not even going to look twice. The EVO is much faster in most reasonable workloads than any of the other discount drives so I can't see paying more for less. Reply
  • MrSpadge - Thursday, July 16, 2015 - link

    I don't see the EVO being "much faster" than a BX100 or ARC100. I'd rather say they're about on par. Reply
  • Uplink10 - Thursday, July 16, 2015 - link

    "initial firmware mostly focused on optimizing performance for benchmarks such as CrystalDiskMark and AS-SSD, which typically use higher IO sizes and queue depths"

    I do not believe what am I reading. They are optimizing firmware so it will look good on charts but are not optimizing the firmware for all uses that the users will perform.

    Components for a PC are used in a lot of ways and optimizing the component for just a few uses that are similar as known benchmark software is wrong. The way I see it they are trying to sell the drive by optimizing the performance that will look good in known benchmark software. That is wrong!!!
    Reply
  • Kristian Vättö - Thursday, July 16, 2015 - link

    It certainly is sad and unfortunately that is more or less the state of the whole component industry (smartphones having "Turbo" modes to exceed the TDP for improved benchmark scores). It's surprising how some SSD companies have very little understanding of real world client workloads and all they see is the few benchmark scores, which frankly are not even relevant to end-users.

    We are trying to fight the fight and convince manufacturers to focus on areas that matter. Getting JMicron to develop a new firmware just based on our feedback (they honestly had no idea why we are even testing low queue depth performance before I explained it to them) is a sign that companies value our input and expertise, which I think is ultimately a win-win for everyone.
    Reply
  • Impulses - Thursday, July 16, 2015 - link

    And we users thank ya for it. AT has been instrumental in providing valuable feedback to SSD manufacturers since the early days, regardless of who was doing the writing.

    Maybe it's time for Kristian's own "SSD Anthology"... :p Seems people keep asking everywhere whether M2/PCI-E or NVMe drives are worth it for them, I know the stock answer is "if you gotta ask..." but still.

    Maybe a one time revisit of some real world testing? HDD vs SATA SSD vs etc one last time to reassess how far the gaps have grown and whatnot.
    Reply
  • ggathagan - Thursday, July 16, 2015 - link

    " (they honestly had no idea why we are even testing low queue depth performance before I explained it to them)"
    That's a scary confession. Makes you wonder if they should find another area of endeavor and leave this market to those who *do* understand.
    Reply

Log in

Don't have an account? Sign up now