Performance Consistency

Our performance consistency test explores the extent to which a drive can reliably sustain performance during a long-duration random write test. Specifications for consumer drives typically list peak performance numbers only attainable in ideal conditions. The performance in a worst-case scenario can be drastically different as over the course of a long test drives can run out of spare area, have to start performing garbage collection, and sometimes even reach power or thermal limits.

In addition to an overall decline in performance, a long test can show patterns in how performance varies on shorter timescales. Some drives will exhibit very little variance in performance from second to second, while others will show massive drops in performance during each garbage collection cycle but otherwise maintain good performance, and others show constantly wide variance. If a drive periodically slows to hard drive levels of performance, it may feel slow to use even if its overall average performance is very high.

To maximally stress the drive's controller and force it to perform garbage collection and wear leveling, this test conducts 4kB random writes with a queue depth of 32. The drive is filled before the start of the test, and the test duration is one hour. Any spare area will be exhausted early in the test and by the end of the hour even the largest drives with the most overprovisioning will have reached a steady state. We use the last 400 seconds of the test to score the drive both on steady-state average writes per second and on its performance divided by the standard deviation.

Steady-State 4KB Random Write Performance

The heatsink makes a huge difference in the sustained random write performance of the M8Pe. With the heatsink, the M8Pe is competitive with the Intel 750 and the Samsung 960 Pro and EVO. Without the heatsink, the M8Pe is slower by almost 10k IOPS, but still above average for consumer SSDs, including some of the NVMe competition.

Steady-State 4KB Random Write Consistency

The M8Pe's consistency scores are quite low, indicating that it lacks the tight regulation of Samsung and Intel's best drives that have similar average performance.

IOPS over time
Default
25% Over-Provisioning

The second by second performance plot reveals that the apparently low consistency is on the M8PeY due entirely to occasional outliers of extremely high performance; most of the time the M8PeY is operating at a very consistent ~22k IOPS, and none of the outliers fall below that level. The M8PeGN without a heatsink or heatspreader suffers from frequent periods of thermal throttling and slightly lower performance when not throttled. The lower performance also causes the M8PeGN to take slightly longer to reach steady state.

Steady-State IOPS over time
Default
25% Over-Provisioning

With a heatsink, the M8Pe's steady state hovers around 22k IOPS, but without the heatsink it alternates between 18k IOPS and a thermally throttled 2400 IOPS. Fortunately, it spends more time in the high-performance state than the low performance state, and there are still a few of the extreme outliers above 100k IOPS.

The M8Pe responds very well to extra overprovisioning. The variability is increased slightly, but the performance with the heatsink jumps to 90-100k IOPS sustained. Without a heatsink, the steady state only increases to 60-65k IOPS, but the drive spends much less time thermally throttled and the performance in that state increases to about 6800 IOPS. It's no surprise that the Marvell 88SS1093 controller has been a popular choice for enterprise M.2 PCIe SSDs.

Introduction AnandTech Storage Bench - The Destroyer
Comments Locked

64 Comments

View All Comments

  • GoMoeJoe - Friday, December 16, 2016 - link

    Nand shortages ... L o L ...
    So much fake news. So much WoW !
  • frenchy_2001 - Friday, December 16, 2016 - link

    "The M8Pe's consistency scores are quite low, indicating that it lacks the tight regulation of Samsung and Intel's best drives that have similar average performance."

    This is rather misleading and shows again the little alue your consistency metric brings.
    In the case of the Plextor, it has a well defined floor around 25kops and bursts of speed above that. Although it makes it inconsistent, it has a high minimum.
    This would be like complaining that a core i7 with turbo boost is inconsistent, because although it has a 4GHz floor, it sometimes boosts to 4.5GHz when possible and as such is worse than a constant i3 at 2GHz...
  • MrSpadge - Monday, December 19, 2016 - link

    Billy is saying just that in the next paragraph. If you read both together it's like "graph A indicates... however, graph B show that it's really..."
  • helvete - Wednesday, January 11, 2017 - link

    Who would buy 1TB variant w/ a heatsink for $650 when one can take 1TB bare plus 128GB w/ a heatsink for $616 ($516 + $100). Then transfer the heatsink and the drive is still $34 cheaper and the 128GB remains remains as a free bonus:-)

Log in

Don't have an account? Sign up now