Whole-Drive Fill

This test starts with a freshly-erased drive and fills it with 128kB sequential writes at queue depth 32, recording the write speed for each 1GB segment. This test is not representative of any ordinary client/consumer usage pattern, but it does allow us to observe transitions in the drive's behavior as it fills up. This can allow us to estimate the size of any SLC write cache, and get a sense for how much performance remains on the rare occasions where real-world usage keeps writing data after filling the cache.

The Sabrent Rocket Q takes the strategy of providing the largest practical SLC cache size, which in this case is a whopping 2TB. The Samsung 870 QVO takes the opposite (and less common for QLC drives) approach of limiting the SLC cache to just 78GB, the same as on the 2TB and 4TB models.

Sustained 128kB Sequential Write (Power Efficiency)
Average Throughput for last 16 GB Overall Average Throughput

Both drives maintain fairly steady write performance after their caches run out, but the Sabrent Rocket Q's post-cache write speed is twice as high. The post-cache write speed of the Rocket Q is still a bit slower than a TLC SATA drive, and is just a fraction of what's typical for TLC NVMe SSDs.

On paper, Samsung's 92L QLC is capable of a program throughput of 18MB/s per die, and the 8TB 870 QVO has 64 of those dies, for an aggregate theoretical write throughput of over 1GB/s. SLC caching can account for some of the performance loss, but the lack of performance scaling beyond the 2TB model is a controller limitation rather than a NAND limitation. The Rocket Q is affected by a similar limitation, but also benefits from QLC NAND with a considerably higher program throughput of 30MB/s per die.

Working Set Size

Most mainstream SSDs have enough DRAM to store the entire mapping table that translates logical block addresses into physical flash memory addresses. DRAMless drives only have small buffers to cache a portion of this mapping information. Some NVMe SSDs support the Host Memory Buffer feature and can borrow a piece of the host system's DRAM for this cache rather needing lots of on-controller memory.

When accessing a logical block whose mapping is not cached, the drive needs to read the mapping from the full table stored on the flash memory before it can read the user data stored at that logical block. This adds extra latency to read operations and in the worst case may double random read latency.

We can see the effects of the size of any mapping buffer by performing random reads from different sized portions of the drive. When performing random reads from a small slice of the drive, we expect the mappings to all fit in the cache, and when performing random reads from the entire drive, we expect mostly cache misses.

When performing this test on mainstream drives with a full-sized DRAM cache, we expect performance to be generally constant regardless of the working set size, or for performance to drop only slightly as the working set size increases.

The Sabrent Rocket Q's random read performance is unusually unsteady at small working set sizes, but levels out at a bit over 8k IOPS for working set sizes of at least 16GB. Reads scattered across the entire drive do show a substantial drop in performance, due to the limited size of the DRAM buffer on this drive.

The Samsung drive has the full 8GB of DRAM and can keep the entire drive's address mapping mapping table in RAM, so its random read performance does not vary with working set size. However, it's clearly slower than the smaller capacities of the 870 QVO; there's some extra overhead in connecting this much flash to a 4-channel controller.

Introduction AnandTech Storage Bench
POST A COMMENT

152 Comments

View All Comments

  • TheinsanegamerN - Monday, December 7, 2020 - link

    "50% of people know what QLC means? Is that a joke? 50% of ATers don't understand what that means"

    Care to back up your staement with evidence? AT is mostly perused by techie people who understand the difference between SLC, MLC, TLC, and QLC.
    Reply
  • at_clucks - Wednesday, December 9, 2020 - link

    @TheinsanegamerN, yeah, you're swimming in the evidence. Check out the comment section carefully and you'll see how well the average ATer understands this. Some may know something about "bits per cell, whatever that means", some may know it's less reliable because "it wears out faster whatever that means", so they know the marketing concepts but not what lies underneath them. Most will blindly assume SLC > MLC > TLC > QLC not why or how, not what the cell is, how it works, how many levels of charge it can have, how it's read or how it's written, how they're organized, not the impact of the implementation, controller, firmware, OS, not why exactly wear is a thing, not why writes wear the cell but reads aren't an issue, not what planar/2D vs. 3D means, etc. Being a "techie" today means you *buy* a lot of tech and gloss over some articles with bar charts of which product is faster. That's it.

    If you want me to give "evidence" of every statement I make prepare to provide answers that have enough references in the footnote to look like a PhD thesis.

    In the meantime it's all but guaranteed that a regular consumer has no clue what QLC means or that the product name is a reference to QLC. They see an SSD that fits their computer, has a certain capacity, and costs a certain price. Maybe the manufacturer on the label alleviates their concerns.

    Knowing QLC has less endurance than SLC ("wears out") or that this is slower than that doesn't mean you understand the tech more than knowing some cars drive faster than others or have lower safety ratings makes you a piston head or mechanic.
    Reply
  • ripbeefbone - Friday, December 11, 2020 - link

    you're way too online Reply
  • Oxford Guy - Sunday, December 6, 2020 - link

    In large part because product pushers like slickdeals don't list the type of NAND in the listing title.

    This is the opposite of how manufacturers wanted to use LED to push TV sales so LED was always listed in product listings.

    People become aware of what manufacturers want them to become aware of. That's why we have so many marketing programs generating graduates all over the world.
    Reply
  • Samus - Sunday, December 6, 2020 - link

    Fortunately we know, and we know to stay away from this crap at this price. An 8TB 870 EVO is "worth" $600 to me and that's all I'm willing to pay for a drive that should logically cost much less than 8x1TB SSD's, not the SAME EXACT PRICE at 8x1TB SSD's (the 870 QVO 1TB regularly sells for $80-$90, and is currently $90 at Best Buy.

    Using Samsung's metric to scale, an 8TB hard drive should cost $400. The controller, DRAM and overall package are the same between drives. The only difference is platters\NAND.
    Reply
  • Oxford Guy - Tuesday, December 8, 2020 - link

    We knowing is irrelevant because consumer ignorance working in manufacturers’ favor is about the bulk of consumer demand not a small number of people who make extra effort to learn specs manufacturers don’t want us to know about and therefore choose to not push. Reply
  • Oxford Guy - Sunday, December 6, 2020 - link

    "Every one of those deals has people saying no to QLC."

    Apples and oranges. The listing titles don't list the type of NAND.
    Reply
  • Oxford Guy - Sunday, December 6, 2020 - link

    There is also the trick of them calling TLC and QLC "MLC". Technically, it is multi-layer NAND so they can get away with it, even though it is completely shady. Reply
  • shabby - Sunday, December 6, 2020 - link

    First company to do that will be stoned to death. Reply
  • Samus - Monday, December 7, 2020 - link

    Do you really think the 860 EVO is MLC like it is advertised as? No, “3-bit” VNAND or more commonly known as TLC. Samsung has been calling TLC [MLC] for years and has it been stoned to death yet. Reply

Log in

Don't have an account? Sign up now