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 Light test allows the flash-based SSDs to make the most of their write caching throughout the test, so even the 900p is surpassed by the Samsung NVMe SSDs while the 800p ranks with the budget NVMe drives. When the drives are full and the flash-based SSDs get bogged down with garbage collection, the 900p comes out ahead but the 800p still trails behind the Samsung 960 PRO. The VROC configuration look especially poor in terms of average data rate, and the RAID-5 performance is surprisingly low.

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

The Optane SSD 800p scores in the middle tier of SSDs for average latency on the Light test, and VROC RAID is no help here. VROC does help with the 99th percentile latencies, but without it the 800p looks like a low-end drive that merely doesn't have garbage collection problems.

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

The average read latencies from the 800p are almost twice as high as those from the 900p, and VROC's overhead only makes it worse. The average write latencies of the 900p aren't as good as the best flash-based SSDs, and the write latency of the 800p is well over twice that of the 900p.

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

The 99th percentile latency scores from the 800p are not great, but at least the drives perform just as well when full. Small flash-based drives are the most heavily affected when constant garbage collection becomes necessary.

ATSB - Light (Power)

The Light test is a fairly short run with any of these drives, but the 800p still manages to return extremely good power usage numbers that are well ahead of any flash-based NVMe SSD.

AnandTech Storage Bench - Heavy Random Performance
Comments Locked

116 Comments

View All Comments

  • eddman - Monday, March 12, 2018 - link

    90? It is stated as 20nm in that table up there.
  • Nottheface - Monday, March 12, 2018 - link

    I was told these are not related in a previous article's posts: https://www.anandtech.com/comments/12136/the-intel...
  • Ewitte12 - Monday, April 30, 2018 - link

    They had difficulty keeping the enterprise drives in stock.

    The 2X quote was for RAM. low queue depth obliterates NAND. Most other speeds are on par with NAND (with sustained a bit behind) but this is direct access to the storage. Most NAND drives have sophisticated RAM caching it can be writing way after the bar disappears off your screen.

    The biggest issue with pricing. Optane has high early adopter fees (which come with a few extra bugs usually). Also anything under the 900p is kinda pointless. 3.0x2 and low capacities??? Not worth it.
  • Gothmoth - Friday, March 9, 2018 - link

    intel hyped this like crazy and after reading the paper i was hyped too.

    but this seems like just another way for intel to push it´s stock market value with redicolous claims.
  • hescominsoon - Friday, March 9, 2018 - link

    Semiaccurate had 3d x-point pegged from the beginning:

    https://www.semiaccurate.com/?s=point
  • Ashinjuka - Saturday, March 10, 2018 - link

    Optanic.
  • DanNeely - Thursday, March 8, 2018 - link

    Could we see results from Optane as cache + budget SSD and Optane as cache + high end SSD?

    I'm not sure it'd be worthwhile with a fast SSD since it only beats them in a subset of benches, but it looks capable of giving a decent boost to budget flash. Cost effectiveness vs just buying better flash'd be the harder question.
  • iter - Thursday, March 8, 2018 - link

    Cache only makes sense for a HDD. It would make no difference combining it with an SSD. Not in terms of real world application performance anyway.

    Spending on 118 gb of optane is pointless when you can get a decent 512 gb ssd for the same money. Over 200% higher the capacity at 99% of the performance. It is a no brainer. Intel will have to resort to bribing OEMs once again if they are to score any design wins.
  • patrickjp93 - Saturday, March 10, 2018 - link

    Uh, think again on big data where the indices for the databases you're running are way too big to fit in memory. AWS is just one cloud provider making extensive use of Optane, especially in DynamoDB, RDS, Memcached, and Lambda where multi-tenant container environments definitely benefit in rapid spinup thanks to the much lower latency 3DXP.
  • Billy Tallis - Thursday, March 8, 2018 - link

    All of our usual SSD tests are for the drive acting as a secondary drive, but Intel's Optane-specific cache software only supports the boot volume, so it's rather awkward to test.

Log in

Don't have an account? Sign up now