Burst IO Performance

Our burst IO tests run at a queue depth of one and the amount of data transferred is limited to ensure that SLC write buffers don't fill up and controllers don't overheat. In between each burst there's enough idle time to keep the drive averaging a 20% duty cycle, allowing for some buffered writes and deferred garbage collection to be completed. The random read and write tests use 4kB operations and the sequential tests use 128kB operations. All the burst tests are confined to a 16GB portion of the drive, so DRAMless SSDs are not disadvantaged as much as they are for larger tests.

QD1 Burst IO Performance
Random Read Random Write
Sequential Read Sequential Write

The aggressive SLC caching strategy used by the Corsair MP400 and most other QLC NVMe SSDs allows them to provide best-case random IO performance that is competitive with many high-end PCIe 3.0 drives. However, despite the 8-channel controller, the burst sequential IO performance of the Corsiar MP400 is still fairly low by NVMe standards. The Sabrent Rocket Q 8TB's results indicate that at least some of the higher-capacity MP400 models should also be able to provide better burst sequential write speeds by virtue of having larger and faster SLC caches.

Sustained IO Performance

Our sustained IO tests measure performance on queue depths up to 32, but the scores reported here are only the averages for the low queue depths (1,2,4) that are most representative of real-world consumer workloads. Each queue depth is tested for up to one minute or 32GB, and the tests are confined to a 64GB span of the drive.

Sustained IO Performance
Random Read Random Write
Sequential Read Sequential Write

On the longer synthetic IO tests, the Corsair MP400's best results are for random writes, where its SLC cache is sufficient to keep it competitive against high-end PCIe 3 drives. The sequential write and random read performance scores are both constrained to the entry-level NVMe performance ranges, but are competitive for that market segment. The sequential read performance is relatively poor even for entry-level NVMe drives, though still significantly better than Samsung's QLC SATA alternative.

Sustained IO Performance
Random Read Random Write
Sequential Read Sequential Write

With QLC NAND and an aging 8-channel controller, it's no surprise that the Corsair MP400's power efficiency scores are generally unimpressive, especially compared to what the 4-channel NVMe drives score when they are performing well. However, the only particularly poor efficiency score from the MP400 is for the sequential read test that it did not perform well on.

Performance at a glance
Random Read Random Write
Sequential Read Sequential Write

Plotting power and performance against our entire library of benchmark results shows that—for better or for worse—the Corsair MP400 doesn't stand out from the crowd or break new ground. The random read performance stays entirely within the range of SATA drives. Random and sequential writes make some use of PCIe performance, but don't come close to saturating the PCIe 3 x4 interface. The sequential read performance does almost make it to 3GB/s at higher queue depths, but it isn't able to fully saturate the PCIe interface the way the 8TB Sabrent Rocket Q can.

Random Read
Random Write
Sequential Read
Sequential Write

Digging into how performance and power scale with increasing queue depths reveals no particular surprises for the Corsair MP400. The biggest discrepancies with the 8TB Sabrent Rocket Q are for random reads and sequential writes: more flash allows the 8TB drive to continue scaling up random read performance after the 1TB MP400 is starting to reach saturation, and the larger SLC cache for the 8TB drive allows higher and more consistent sequential write performance.

AnandTech Storage Bench Mixed Read/Write Performance And Power Management
Comments Locked

75 Comments

View All Comments

  • ballsystemlord - Friday, December 11, 2020 - link

    I have seen normal people burn through 150TBW in about 2m. One of them used folding@home to help with the pandemic (IDK if that program was the cause). I forget what the other was doing.

    Therefore, I also support the demand for higher endurance flash.
  • Ithaqua - Friday, December 11, 2020 - link

    150TB in 2 minutes? That is some massive write speeds.
  • Unashamed_unoriginal_username_x86 - Saturday, December 12, 2020 - link

    1250GB/s or 157 PCIe 6.0 lanes' bandwidth
  • Samus - Sunday, December 13, 2020 - link

    I just hopped in my DeLorean to check and this is the norm in 2029!
  • ballsystemlord - Saturday, December 12, 2020 - link

    2 months, silly. ;)
  • Gigaplex - Wednesday, December 16, 2020 - link

    Lower case m as a unit of time signifies minutes, not months, silly.
  • Beaver M. - Saturday, December 12, 2020 - link

    That sure as hell is not normal.
    My drive is 5 years old and only has 62 TBW. And its the only drive in my PC, so everything is done on it, 12 to 16 hours a day.
  • Spunjji - Monday, December 14, 2020 - link

    Seconding this; whatever they were doing was not "normal".
  • joesiv - Monday, December 14, 2020 - link

    Or the firmware is at fault. Bugs in SSD firmwares happen and can cause very bad write efficiencies. I saw the same thing at work. Same workload given to 5 different brand/model drives, one drive ran through 1% of it's life expectancy every day, others were fine.
  • rrinker - Monday, December 14, 2020 - link

    I have an 850 EVO that was the C drive in my old computer, now in as the D drive in my new one - I looked it up, I've had it closer to 6 years than 5, bought early 2015. As the D drive in this computer, it is my download drive, torrent temp storage, and basically all around junk collector drive - and it only has 16.2 TBW. I typically download to this drive, then copy things to my server, but really the most changing stuff besides the OS are on this drive. I don't do video production sort of tasks.

Log in

Don't have an account? Sign up now