AnandTech Storage Bench - Light

Our Light storage test has relatively more sequential accesses and lower queue depths than The Destroyer or the Heavy test, and it's by far the shortest test overall. It's based largely on applications that aren't highly dependent on storage performance, so this is a test more of application launch times and file load times. This test can be seen as the sum of all the little delays in daily usage, but with the idle times trimmed to 25ms it takes less than half an hour to run. Details of the Light test can be found here. As with the ATSB Heavy test, this test is run with the drive both freshly erased and empty, and after filling the drive with sequential writes.

ATSB - Light (Data Rate)

The Intel Optane SSD 900P doesn't come in first place for overall data rate on the Light test, until the drives are filled and the average data rate of all the flash-based SSDs takes a big hit.

ATSB - Light (Average Latency)ATSB - Light (99th Percentile Latency)

The average and 99th percentile latencies of the Optane SSD on the Light test are on par with the top flash-based SSDs when the test is run on an empty drive. When the drives are filled before the test, the flash-based SSDs slow down enough that the Optane SSD takes first place easily, with an especially wide margin on the 99th percentile latency.

ATSB - Light (Average Read Latency)ATSB - Light (Average Write Latency)

The average read latency of the Optane SSD 900P on the Light test is merely tied for first place, when the test is run on an empty drive. When the drives are filled, the Optane SSD has half the average read latency of anything else. The write latency situation is quite different; whether or not the drives are filled, most of the top flash-based SSDs are able to fit the bursts of writes in their caches and deliver better latency than the uncached writes of the Optane SSD.

ATSB - Light (99th Percentile Read Latency)ATSB - Light (99th Percentile Write Latency)

The 99th percentile read latency of the Optane SSD 900P on the Light test is tied for first place when the test is run on an empty drive, and leads by more than 60% when the drives are filled before the test. The 99th percentile write latency lags behind the top flash-based SSDs a bit, but nowhere near enough to be noticeable: the latency is still an order of magnitude lower than SATA SSDs.

AnandTech Storage Bench - Heavy Random Performance
Comments Locked

205 Comments

View All Comments

  • ddriver - Friday, October 27, 2017 - link

    You probably missed the slide where intel positions this product at "eSports" - believe it or not, it will make a world of a difference. You may not see even 1% of improvement, but it will definitely make you much better at eSports.

    But then again, you can get exactly the same benefit from just putting a sticker on your box, without actually paying for the product. It is enough for others to think you have it.

    Sarcasm aside, at this price it is a good purchase for database usage. But only if very low latency is required, meaning a local server on a very fast network, and the server is being lightly loaded. For an internet server the advantages will be diminished by the internet connection bottleneck, and for higher loads, as evident from the test results, SSDs do catch up as the QD increases, and still offer significantly lower cost per GB.
  • lmcd - Friday, October 27, 2017 - link

    While the endurance shut-off seems pretty ridiculous, I'd imagine as this technology ramps up its density and cost-efficiency it will supplant traditional NAND in the data center before traditional NAND even garners much of a foothold.
  • iwod - Saturday, October 28, 2017 - link

    I think the reality is 90% of consumers are never going to hit that endurance number, ever. Statiscally speaking, they are much more likely to get a Memory Error, Capacitor malfunction, Power Supply issues or likely their CPU heat cooling system ( Increasingly a problem ) messed up before that number ever arrived.

    And I think another problem, is Intel cant figure out the best endurance time and method on this new tech. And they are playing it safe then sorry. Running a 10 PB test takes time.
  • "Bullwinkle J Moose" - Friday, October 27, 2017 - link

    Thank you ddriver

    I was going to let the My Banhammer stand until I remembered why I love this site so much...

    It's not the articles, it's the comments

    You have learned well my friend!
  • eddman - Sunday, October 29, 2017 - link

    Horses attached to the same carriage. Keep pulling.
  • peevee - Friday, October 27, 2017 - link

    Looks like this thing is fast enough to be used as swap drive for memory-intensive tasks.
  • Reflex - Friday, October 27, 2017 - link

    This is a good product for those who can benefit from it, and a great start for the first major advance in solid state storage we have seen in many years. Hopefully the price continues to come down, power consumption declines and capacity rises. I will be very interested in seeing whether or not they can achieve the projections that were made when the technology was first announced.
  • Huayra - Friday, October 27, 2017 - link

    Is it possible to install macOS on it using thunderbolt enclosure connected to a Macbook?
  • Ippokratis - Saturday, October 28, 2017 - link

    Hi, nice article.
    1 - Can Optane be used as a scratch disk for Ryzen / Threadripper, OSX /Linux ?
    2 - Could you please a compiler (gcc, javascript) benchmark ? Random IOPS are very important in compiling and making compile times shorter is a good reason to buy tech - consider it as a real life test.
    Thanks
  • peevee - Monday, October 30, 2017 - link

    Agree with compile test. On 16+ core machine though, not 4 measly cores. And source code which does not all fit in RAM.

Log in

Don't have an account? Sign up now