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)

Like the Optane Memory M.2, the Optane SSD 800p has extremely high random read performance even at QD1. The M.2 drives even have a substantial lead over the much larger and more power-hungry 900p and its enterprise counterpart P4800X. Even the best flash-based SSDs are almost an order of magnitude slower.

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

The Optane SSDs continue to dominate on the longer random read test, though the addition of higher queue depths allows the 900p to pull ahead of the 800p.

Sustained 4kB Random Read (Power Efficiency)

With extremely high performance but lacking the high power draw of the enterprise-class 900p, the Optane SSD 800p is by far the most power efficient at performing random reads.

The Optane SSD 800p starts out in the lead at QD1, but its performance is overtaken by the 900p at all higher queue depths. The flash-based SSDs have power consumption that is comparable to the 800p, but even at QD32 Samsung's 960 PRO hasn't caught up to the 800p's random read performance.

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)

Flash-based SSDs can cache and combine write operations, so they are able to offer random write performance close to that of the Optane SSDs, which do not perform any significant caching. Where the 32GB Optane Memory offered relatively poor burst random write performance, the 800p is at least as fast as the best flash-based SSDs.

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

When higher queue depths come into play, the write caching ability of Samsung's high-end NVMe SSDs allows them to exceed the Optane SSD 800p's random write speed, though the 900p still holds on to the lead. The 800p's improvement over the Optane Memory is even more apparent with this longer test.

Sustained 4kB Random Write (Power Efficiency)

The power efficiency of the 800p during random writes is pretty good, though Samsung's top drives are better still. The Optane Memory lags behind on account of its poor performance, and the 900p ranks below that because it draws so much power in the process of delivering top performance.

The Samsung 960 PRO and the Intel Optane SSD 900p show off at high queue depths thanks to the high channel counts of their controllers. The Optane SSD 800p doesn't have much room for performance to scale beyond QD2.

AnandTech Storage Bench - Light Sequential Performance
POST A COMMENT

116 Comments

View All Comments

  • Alexvrb - Saturday, March 10, 2018 - link

    Calin you are obviously too young to remember some of the early "affordable" consumer NAND SSDs. Hammer them a bit and they stalled... producing worse results than a lot of fast HDDs... especially in random writes. Sequential speeds were never a major issue that I can recall. Reply
  • The_Assimilator - Friday, March 09, 2018 - link

    Trying to equate a NAND-to-Optane transition to the mechanical-HDD-to-SSD transition is laughable. Reply
  • wumpus - Friday, March 09, 2018 - link

    The moment pseudo-SLC in TLC showed up, Optane was pretty much dead in the SSD market. They would presumably compete with SLC (does anybody still make it?), but TLC is the coffin nail in consumer markets.

    From the moment the 3d-xpoint hypetrane started, it was clear that it would try to wedge itself into the memory hierarchy, presumably between flash and DRAM, and hopes for replacing flash.

    Flash isn't going anywhere, and 3d-xpoint hasn't shown the endurance needed for a fast-paging DRAM replacement. It certainly wouldn't replace *all* DRAM, but anyone who's seen a 4GB machine actually function (slow, but they do work) knows that nearly all that expensive (hopefully DDR4 will fall back to Earth) DRAM could be replaced by something sufficiently fast, but neither flash nor 3d-xpoint is quite there.

    To compound the problems, Intel decided that "Optane in a DDR4 slot" would be strictly proprietary. So there are marketing/political problems trying to get manufacturers to support it as well as technical issues to make the stuff.
    Reply
  • name99 - Friday, March 09, 2018 - link

    Consumer NAND launched in an environment where it had SOME spaces where it was optimal, and so had the chance to grow. It started in phones and DAPs, then grew to ultra-laptops, and finally the desktop. Point is --- there were niches that could pay for on-going improvement.

    Octane is different because there is NO obvious niche that justifies continuing to pump money into it. The niche that was SUPPOSED to justify it (NV-DIMMs) is STILL MIA years after it was promised...
    Reply
  • iwod - Friday, March 09, 2018 - link

    I am all for super fast QD1 results. But so far none of the application seems to benefits from it. At least not according to test results. I am wonder, we are either testing it wrong, looking at the wrong thing, or the benefits of QD1 is over thought and bottleneck is somewhere else.

    And NAND continues to get bigger better and faster. We may be looking at below $100 250GB SSD this year.
    Reply
  • iter - Friday, March 09, 2018 - link

    Exactly. It is hilarious how them fanboys keep claiming that we overlook the advantages, when I explicitly state them almost every time.

    There are very little and far in between workloads where those advantages can translate into tangible improvement of real world performance.

    When your bottleneck is a human being interacting via input devices, discrete savings of several dozens of microseconds are simply not perceivable.

    Even cumulative savings are in fact not, because most of the time that data has to also be processed by the cpu, which is why synthetics aside, raw real world applications snow minuscule going from a decent ssd to a crazy fast nvme device.
    Reply
  • sor - Friday, March 09, 2018 - link

    Probably has something to do with your name calling and “it keeps getting worse and worse” when that objectively isn’t true. You come off as having an axe to grind.

    It is not true that this is worse and worse. The power improvements shown here are quite impressive. Low QD performance is still better than NAND by an order of magnitude, and looks to have gotten a roughly 20% improvement. Sequential read now even beats NAND.

    You and others are falling over yourselves to crap on it for some strange reason, and clearly are ignoring the upsides. It’s just a product.
    Reply
  • iter - Friday, March 09, 2018 - link

    "when that objectively isn’t true"

    It absolutely is. It is slower than the 900p. They improved power a bit - big whoop, especially considering it came at the cost of gutting the interface by 50%.

    118 GB? I bet enthusiasts all over the planet are drooling about that crazy capacity. Not to mention the smaller model...

    Nobody denies the strong points, it is just that they are way too little to make this a good product.

    Instead of getting bigger and faster it gets smaller and slower.

    And somehow the price per GB increases.

    Truly impressive.
    Reply
  • nevcairiel - Friday, March 09, 2018 - link

    If you want to go down that road, at current consumer SSD speeds (say Samsung 960 Pro), I doubt any normal user would even notice if the performance suddenly doubled (or halfed, for that matter).

    Does that mean we should not innovate? Perhaps consumer work-load isn't the main goal, but if you have the hardware, why not try to make a consumer product, anyway.
    Reply
  • MrSpadge - Friday, March 09, 2018 - link

    With decently fast SATA SSDs the bottleneck is almsot entirely the CPU already, unless you've got purely I/O load. Reply

Log in

Don't have an account? Sign up now