AnandTech Storage Bench - The Destroyer

The Destroyer is an extremely long test replicating the access patterns of very IO-intensive desktop usage. A detailed breakdown can be found in this article. Like real-world usage, the drives do get the occasional break that allows for some background garbage collection and flushing caches, but those idle times are limited to 25ms so that it doesn't take all week to run the test. These AnandTech Storage Bench (ATSB) tests do not involve running the actual applications that generated the workloads, so the scores are relatively insensitive to changes in CPU performance and RAM from our new testbed, but the jump to a newer version of Windows and the newer storage drivers can have an impact.

We quantify performance on this test by reporting the drive's average data throughput, the average latency of the I/O operations, and the total energy used by the drive over the course of the test.

ATSB - The Destroyer (Data Rate)

The average data rate from the Crucial P1 on The Destroyer is comparable to other entry-level NVMe drives like the Phison E8-based Kingston A1000 and the Intel 660p. The P1 also roughly matches the average data rate of the Crucial MX500 SATA SSD, while several high-end NVMe drives deliver more than twice the performance.

ATSB - The Destroyer (Average Latency)ATSB - The Destroyer (99th Percentile Latency)

While the average data rate from the Crucial P1 may have been similar to the MX500, the average latency is about twice that of the MX500, and slightly higher than the Intel 660p. The situation is better for the 99th percentile latency, where the Crucial P1 comes close to the MX500 and shows half the 99th percentile latency of the Intel 660p.

ATSB - The Destroyer (Average Read Latency)ATSB - The Destroyer (Average Write Latency)

The average read and write latencies for the Crucial P1 on The Destroyer are both slightly worse than what the Intel 660p provides, and the Crucial MX500 provides a much better average write latency.

ATSB - The Destroyer (99th Percentile Read Latency)ATSB - The Destroyer (99th Percentile Write Latency)

The 99th percentile read latency from the Crucial P1 is very good, competitive with several high-end NVMe SSDs. However, the 99th percentile write latency is quite poor compared to almost any other NVMe SSD or mainstream SATA SSD. This is a huge difference in behavior compared to the Intel 660p. The Crucial P1 is optimized much more for reliable read latency, at significant cost to worst-case write latency under heavy workloads.

ATSB - The Destroyer (Power)

The Crucial P1 requires significantly more energy to complete The Destroyer than the Intel 660p, despite the near-identical hardware and very similar overall performance numbers. The Crucial P1's relatively poor efficiency on this test isn't a serious issue given that the drive is intended for less demanding use cases, but combined with the high 99th percentile write latency this points to the P1 possibly experiencing higher write amplification on The Destroyer than the Intel 660p experiences.

Introduction AnandTech Storage Bench - Heavy
POST A COMMENT

68 Comments

View All Comments

  • Oxford Guy - Thursday, November 8, 2018 - link

    "That's what I want ... not another race to the bottom."

    That's what consumers want: value.

    That's not what companies want. They want the opposite. Their wish is to sell the least for the most.
    Reply
  • Mikewind Dale - Thursday, November 8, 2018 - link

    "[Companies] want the opposite. Their wish is to sell the least for the most."

    Not true. Companies want to maximize net revenue, i.e. total revenue minus cost.

    Depending on the elasticity of demand (i.e. price sensitivity), that might mean increasing quantity and decreasing price.

    A reduction in quantity and an increase in price will increase net revenue only if demand is elastic.

    But given the existence of HDDs, it makes sense that demand for SSDs is elastic, i.e. price-sensitive. These aren't captive consumers with zero choice.

    Of course, nothing stops a company from catering to BOTH markets, i.e. high performance AND low cost markets.
    Reply
  • Mikewind Dale - Thursday, November 8, 2018 - link

    Sic:

    "A reduction in quantity and an increase in price will increase net revenue only if demand is elastic."

    That should be "inelastic."
    Reply
  • limitedaccess - Thursday, November 8, 2018 - link

    The transition to TLC drives was also shortly followed with the transition to 3D NAND using higher process (larger) from smaller planar litho process. While smaller litho allowed more density it also came with the trade off of worse endurance/higher decay. So the transition to 3D NAND effectively offset the issues of MLC->TLC which is where we are today. What's the equivalent for TLC->QLC?

    Low litho planar TLC drives were the ones that were poorly received and performed worse then they reviewed in reality due to decay. And decay is the real issue here with QLC since no reviewer tests for it (it isn't the same as poor write endurance). Is that file I don't regularly access going to maintain the same read speeds or have massively higher latency to access due to the need for ECC to kick in?
    Reply
  • 0ldman79 - Monday, November 12, 2018 - link

    I may not be correct on the exact numbers, but I think the NAND lithography has stopped at 22nm as they were having issues retaining data on 14nm, just no real benefit going to a smaller lithography.

    They may tune that in a couple of years, but the only way I can see that working with my rudimentary understanding of the system is to keep everything the same size as the 22nm (gates, gaps, fences, chains, roads, whatever, it's too late/early for me to remember the correct terms), same gaps only on a smaller process. They'd have no reduction in cost as they'd be using the same amount of each wafer, might have a reduction in power consumption.

    I'm eager to see how they address the problem but it really looks like QLC may be a dead end. Eventually we're going to hit walls where lithography can't improve and we're going to have to come at the problem (cpu speed, memory speeds, NAND speeds, etc) from an entirely different angle than what we've been doing. For what, 40 years, we've been doing major design changes every 5 years or so and just relying on lithography to improve clock speeds.

    I think that is about to cease entirely. They can probably go farther than what we're seeing but not economically.
    Reply
  • Lolimaster - Friday, November 9, 2018 - link

    Youre not specting a drive limited to 500MB to be as fast as a PCI-E 4x SSD with full support for it...

    TLC vs MLC all goes to endurance and degraded performance when the drive is full or the cache is exhausted.
    Reply
  • Lolimaster - Friday, November 9, 2018 - link

    Random performance seems the land of Optane and similar. Even the 16GB optane M10 absoluletely murders even the top of the line NVME Samsung MLC SSD. Reply
  • PaoDeTech - Thursday, November 8, 2018 - link

    Yes, price is still too high. But it will come down. I think that the conclusions fail to highlight the main strength of this SSD: top performance / power. For portable devices, this is the key metric to consider. In this regard is far ahead any SATA SSD and almost all PCIe out there. Reply
  • Lolimaster - Friday, November 9, 2018 - link

    Exactly. QLC should stick to big multiterabyte drives for avrg user or HEDT.

    Like 4TB+.
    Reply
  • 0ldman79 - Monday, November 12, 2018 - link

    I think that's where they need to place QLC.

    Massive "read mostly" storage. xx layer TLC for a performance drive, QLC for massive data storage, ie; all of my Steam games installed on a 10 cent per gig "read mostly" drive while the OS and my general use is on a 22 cent per gig TLC.

    That's what they're trying to do with that SLC cache, but I think they need to push it a lot farther, throw a 500GB TLC cache on a 4 terabyte QLC drive. That might be able to have it fit into the mainstream NVME lineup.
    Reply

Log in

Don't have an account? Sign up now