Random Read Performance

Our first test of random read performance uses very short bursts of operations issued one at a time with no queuing. The drives are given enough idle time between bursts to yield an overall duty cycle of 20%, so thermal throttling is impossible. Each burst consists of a total of 32MB of 4kB random reads, from a 16GB span of the disk. The total data read is 1GB.

The Patriot Hellfire, in blue, is highlighted as an example of a last-generation Phison E7 drive. Although we didn't test it at the time, the MP500 was based on the same controller and memory.

Burst 4kB Random Read (Queue Depth 1)

The Corsair Force MP510 can't match the burst random read performance of a Silicon Motion controller paired with IMFT 64L 3D TLC, but the MP510 has the fastest random reads of any drive using Toshiba/SanDisk BiCS TLC and it also beats the Samsung 970 EVO.

Our sustained random read performance is similar to the random read test from our 2015 test suite: queue depths from 1 to 32 are tested, and the average performance and power efficiency across QD1, QD2 and QD4 are reported as the primary scores. Each queue depth is tested for one minute or 32GB of data transferred, whichever is shorter. After each queue depth is tested, the drive is given up to one minute to cool off so that the higher queue depths are unlikely to be affected by accumulated heat build-up. The individual read operations are again 4kB, and cover a 64GB span of the drive.

Sustained 4kB Random Read

On the longer random read test that adds some higher queue depths, the MP510's performance standing falls somewhat as the Samsung 970 EVO and a few other drives with BiCS TLC overtake it, while Silicon Motion drives retain a commanding lead.

Sustained 4kB Random Read (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

The power efficiency of the MP510 during random reads is reasonable but is about 15% worse than what WD and Toshiba can do using their own controllers with this NAND.

The MP510 may not provide the best random read performance at low queue depths, but its performance does scale up nicely when the queue depth continues to grow. By QD32 it is delivering over 800MB/s with a little more than 3W and showing no sign of approaching a performance ceiling; Phison's plans for enterprise drives based on this controller seem to have merit.

Random Write Performance

Our test of random write burst performance is structured similarly to the random read burst test, but each burst is only 4MB and the total test length is 128MB. The 4kB random write operations are distributed over a 16GB span of the drive, and the operations are issued one at a time with no queuing.

Burst 4kB Random Write (Queue Depth 1)

The Corsair Force MP510 is essentially tied for first place for burst random write performance: its SLC write cache has a very low latency.

As with the sustained random read test, our sustained 4kB random write test runs for up to one minute or 32GB per queue depth, covering a 64GB span of the drive and giving the drive up to 1 minute of idle time between queue depths to allow for write caches to be flushed and for the drive to cool down.

Sustained 4kB Random Write

On the longer random write test, the Corsair MP510 loses the lead but stays in the top tier of high-performing drives.

Sustained 4kB Random Write (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

The power efficiency of the Corsair Force MP510 during random writes is a bit better than average for NVMe drives, but significantly worse than what Toshiba and WD can do by pairing the same NAND with their own controllers. The WD Black manages this substantial efficiency advantage wil also slightly outperforming the MP510.

Like many of its closest competitors, the Corsair Force MP510's random write speed is saturated by QD4, but it plateaus well below the limit of the WD Black, which doesn't require any more power than the MP510.

AnandTech Storage Bench - Light Sequential Performance
Comments Locked

42 Comments

View All Comments

  • ATC9001 - Friday, October 19, 2018 - link

    Not bad...competition is good to drive prices down, but if I were in the market for an nvme drive I'd take the HP EX920 1TB for 199!
  • euler007 - Friday, October 19, 2018 - link

    I'm really liking these prices. If RAM comes down in price a new PC is in my future.
  • enzotiger - Friday, October 19, 2018 - link

    Please check your numbers. Random read IOPS of 610K is not only by far the highest IOPS among M.2, it actually beats Optane 905P. Highly suspicious.
  • Billy Tallis - Monday, October 22, 2018 - link

    The 610k IOPS for random reads is the advertised specifications from Corsair, not my own measurements. I don't test consumer drives at queue depths high enough to determine whether it can actually hit 610k IOPS, because that doesn't come close to representing any real consumer workload.
  • Hxx - Friday, October 19, 2018 - link

    those prices are wrong right? I see the 480 gb model for 240+ at amazon unless amazon is price gouging.
  • eek2121 - Saturday, October 20, 2018 - link

    All the big retailers have algorithms to automatically shift pricing based on supply vs demand. Anandtech lists the MSRPs, but if everyone rushes out to buy the drive at once, Amazon, Newegg, etc. want to make as much money as possible while still balancing supply vs demand, so the price automatically shifts up. I'm surprised people haven't figured this out yet. That's why you wait for demand to drop before buying a product.
  • ballsystemlord - Sunday, October 21, 2018 - link

    Tallis, where are the 4k sequential read and write tests? I have a use case for them!
  • Billy Tallis - Monday, October 22, 2018 - link

    I doubt it. Whatever OS and filesystem you are using is likely to have a prefetch mechanism that make your small block sequential reads into mostly large block reads, and write caching that will batch up small block sequential writes. If you're trying to bypass the write cache for small block writes, then you probably need to be shopping for an enterprise SSD.
  • ballsystemlord - Monday, October 22, 2018 - link

    Ok. Thanks!
  • Violet Giraffe - Tuesday, November 13, 2018 - link

    I'm keen to think a lot of real-life use cases are bound on small block reading speed. E. g. databases.

Log in

Don't have an account? Sign up now