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.

Burst 4kB Random Read (Queue Depth 1)

The QD1 burst random read speeds of the Phison E7 drives are all over the chart. The Corsair Neutron NX500 ranks just above the middle, while the Patriot Hellfire is the second-fastest drive in the bunch. With the exception of the underperforming Zotac SONIX and the SATA-based 850 PRO, this chart shows clear separation between the MLC and TLC drives.

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 a longer test that includes some moderately higher queue depths, the NX500 comes out on top among the Phison E7 drives, but it is a bit below average compared to the market as a whole.

Sustained 4kB Random Read (Power Efficiency)

The power efficiency of the NX500 is not quite competitive with the Patriot Hellfire M.2, and nowhere close to most of the 3D NAND SSDs.

The Corsair Neutron NX500's random read performance scales at higher queue depths much better than the other Phison E7 SSDs, but even at QD32 it is slower than Toshiba's own OCZ RD400 using the same NAND.

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 burst random write speed of the Corsair Neutron NX500 is far below what it should be. Both of the other Phison E7 drives are substantially faster and yet still slower than even the Intel SSD 600p.

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

With a longer test and bringing in some higher queue depths, the Intel 600P and WD Black fall below the NX500, but the NX500 is still performing much worse than the other Phison E7 SSDs or any other MLC NVMe SSD. The Patriot Hellfire is basically tied for second place behind Samsung.

Sustained 4kB Random Write (Power Efficiency)

The NX500's power efficiency during random writes isn't great, but it's only slightly worse than some of the competing NVMe MLC drives and is substantially better than the entry-level NVMe TLC drives.

The NX500's random write speed does reach decent levels at queue depths of 8 or higher, but at the low queue depths that matter for real-world desktop performance, the NX500 is way behind.

AnandTech Storage Bench - Light Sequential Performance
Comments Locked

45 Comments

View All Comments

  • Billy Tallis - Thursday, August 17, 2017 - link

    The ATSB Heavy and Light tests include data from runs on a full drive, and The Destroyer writes more than enough data to put this drive into steady-state. Synthetic benchmarks of steady-state performance would not be more representative of real-world usage. Client drives do not get hammered with constant writes. I will eventually add some steady-state tests back into the test suite, but they will not be and never have been the most important aspect of a client drive review. They're useful to study how the drive handles garbage collection under pressure, but the impact that has on real-world performance is minimal.
  • qlum - Friday, August 18, 2017 - link

    The only place I wouldn't go for samsung is when you want to use a cheap 120gb ssd. At that point the cheapest samsung drives are just too expensive.
  • Vorl - Wednesday, August 16, 2017 - link

    did I miss something big, besides the card? This while a good review, is a very uninteresting product that just wastes space compared to a 4x m.2 slot.
  • DanNeely - Wednesday, August 16, 2017 - link

    You can put a card form factor drive in an older board without m.2 slots. Unfortunately the underlying Phision controller isn't much faster than older SATA models; making it another underwhelming product.
  • DigitalFreak - Wednesday, August 16, 2017 - link

    Even then you can buy a cheap PCI-E x4 to m.2 adapter for like $15. There's no reason for this card to exist at these capacities. If it was 2 or 4TB, maybe, but not 400/800GB
  • mapesdhs - Thursday, August 17, 2017 - link

    Yup, I have a 960 Pro 512GB on an Akasa card on an X79 board, does about 3.5GB/sec in CDM.

    Pity the review didn't mention the cheaper SM951/SM961, and they really need to get a 960 Pro to round out the data, the one I bought wasn't that much more than the EVO and it's a far better product. I don't like the 960 EVO, it's slower than the 950 Pro most of the time.
  • r3loaded - Wednesday, August 16, 2017 - link

    > Skip to the graphs.
    > Another SSD that gets pwned by a 960 Evo, nevermind the Pro.
    > Write this comment, ignore the rest of the review and close the tab.
  • creed3020 - Tuesday, August 22, 2017 - link

    +1

    Unfortunately so, wish it wasn't......very disappointing Corsair.
  • timchen - Wednesday, August 16, 2017 - link

    If I am not mistaken, 960 EVO 1 TB can perform quite differently to 500GB... so using the 1TB performance per dollar does not seem very fair to the 400 GB...
  • Billy Tallis - Wednesday, August 16, 2017 - link

    I do wish I had a sample of the 500GB 960 EVO, because performance does generally scale with capacity. But it's pretty safe to assume that at low queue depths and while the SLC cache isn't full, the 500GB 960 EVO will perform similarly enough to the 1TB that it still beats the Phison E7 drives.

Log in

Don't have an account? Sign up now