QLC Goes To 8TB: Samsung 870 QVO and Sabrent Rocket Q 8TB SSDs Reviewed

Flash memory prices have been on a downward trajectory for years. A decade ago, this trend was helping SSDs establish a foothold in the consumer market—largely for enthusiasts. Now, SSDs have taken over as the default storage medium for consumer PCs and further advances in flash memory are no longer pushing consumer SSDs into new product segments. Instead, cheaper flash is driving an increase in SSD capacity.

That growth in drive capacity has not been steady. For both technical and marketing reasons, consumer SSD capacities stick close to powers of two. The first 2TB consumer SSDs started to show up in 2015, and now 2TB options are common across all the SSD market segments. 4TB drives started to show up in 2018 but are still quite rare, and this year we’ve seen the first 8TB consumer SSDs.

Today we’re looking at the first two consumer-oriented 8TB SSDs. The 8TB Samsung 870 QVO is a SATA drive from the brand that has been at the forefront of the past several capacity increases and leads the SSD market by most other measures. The other drive is the 8TB Sabrent Rocket Q, a M.2 NVMe drive from a brand that’s working to stand out from the crowd of many other Phison partners. Unsurprisingly, both of these drives use four bit per cell QLC NAND flash memory which offers the lowest cost per GB and the highest per-die capacities currently available. QLC NAND generally puts SSDs into an entry-level market segment, but due to their extreme capacities these 8TB SSDs are still some of the most expensive drives in the consumer SSD market.

Sabrent Rocket Q 8TB

Sabrent Rocket Q Specifications
Capacity 500 GB 1 TB 2 TB 4 TB 8 TB
Form Factor M.2 2280 single-sided
PCIe 3 x4
M.2 2280 double-sided
PCIe 3 x4
Controller Sabrent-branded Phison E12S
NAND Flash Micron 1Tbit 96L 3D QLC
DRAM Kingston DDR3
Sequential Read (MB/s) 2000 3200 3300
Sequential Write (MB/s) 1000 2000 3000 2900
Random Read IOPS (4kB) 95k 125k 255k 550k 550k
Random Write IOPS (4kB) 250k 500k 670k 680k
Power
Consumption (W)
Read 3.5 5.0 5.5 5.0 5.6
Write 3.5 4.5 6.0 6.5 6.5
Warranty 5 years
Write Endurance 120 TB
0.13 DWPD
260 TB
0.14 DWPD
530 TB
0.14 DWPD
940 TB
0.13 DWPD
1800 TB
0.13 DWPD
Current Retail Prices $64.99 (13¢/GB) $109.98 (11¢/GB) $219.98 (11¢/GB) $599.98 (15¢/GB) $1299.99 (16¢/GB)
 

The Sabrent Rocket Q is a bit unusual among QLC NVMe SSDs, and not just because it offers such high capacities. Most consumer QLC SSDs use fairly low-end SSD controllers rather than let the performance potential of a high-end controller be wasted on slow QLC NAND. But the Rocket Q uses the Phison E12, a familiar mainstay of the high-end NVMe market segment (but seen here in the more compact E12S packaging). This means the Rocket Q has an 8-channel controller at its disposal rather than just four channels, and that helps immensely at the higher capacities where there’s enough flash to compensate for the low performance of QLC NAND.

The Rocket Q does cut corners a bit by using just one fourth of the DRAM we usually see on mainstream SSDs. That hurts a bit at the lower capacities (though nowhere near as much as a fully DRAMless design would), but is much less of a problem for this 8TB model: 2GB of DRAM is still plenty for the SSD to handle any typical consumer workload.

The Rocket Q lineup goes from 500GB to 8TB, but we generally consider QLC drives smaller than 1TB to be a poor alternative to DRAMless TLC drives. That’s even more true for the Rocket Q, because the 500GB model can only use half of the Phison E12’s 8 channels.

Sabrent has also introduced the Rocket Q4 as a partial successor. This uses the Phison E16 controller and brings PCIe 4 support and improved performance. However, the E16 is not yet (and may never be) available in a small package size like the E12S controller, so it is not yet practical for Sabrent and Phison to squeeze 8TB of QLC onto a PCIe gen4 M.2 drive.

Samsung 870 QVO 8TB

Samsung 870 QVO Specifications
Capacity 1 TB 2 TB 4 TB 8 TB
Form Factor 2.5″ 7mm SATA
Controller Samsung MKX
NAND Flash Samsung 1Tbit 92L 3D QLC
LPDDR4 DRAM 1 GB 2 GB 4 GB 8 GB
Max SLC Cache Size 42 GB 78 GB 78 GB 78 GB
Sequential Read 560 MB/s
Sequential
Write
SLC 530 MB/s
QLC 80 MB/s 160 MB/s
Random
Read
IOPS (4kB)
QD1 11k (SLC)
5k (QLC)
11k (SLC)
5k (QLC)
QD32 98k (SLC)
45k (QLC)
98k (SLC)
74k (QLC)
Random
Write
IOPS (4kB)
QD1 35k (SLC)
22k (QLC)
35k (SLC)
34k (QLC)
QD32 88k (SLC)
22k (QLC)
88k (SLC)
42k (QLC)
Power
Consumption
Read 2.1 W 2.1 W 2.2 W 2.4 W
Write 2.2 W 3.0 W 3.2 W 3.3 W
Idle 30 mW 30 mW 35 mW 45 mW
DevSlp 3 mW 4 mW 7 mW 10 mW
Warranty 3 years
Write Endurance 360 TB
0.3 DWPD
720 TB
0.3 DWPD
1440 TB
0.3 DWPD
2880 TB
0.3 DWPD
Current Retail Prices $89.99
(9¢/GB)
$199.99
(10¢/GB)
$419.99
(10¢/GB)
$899.99
(11¢/GB)
 

We reviewed the Samsung 870 QVO when it first launched, but the 8TB capacity arrived a bit later. Other than the higher capacity, there’s not much new to say about the 8TB model of their second-generation QLC SSD. It has twice the NAND and twice the DRAM and twice the total write endurance, but the same performance ratings and SLC cache sizes as the 2TB and 4TB models. Samsung’s SATA SSD controller doesn’t offer much potential for higher performance once all channels are populated with at least two NAND dies.

Compared to the Rocket Q, the Samsung 870 QVO has higher write endurance ratings both in terms of drive writes per day and total TB written. However, the Rocket Q comes with a 5 year warranty and the 870 QVO only has a 3 year warranty. The Samsung 870 QVO is far cheaper at these high capacities; NVMe SSD controllers are only a little bit more expensive than SATA SSD controllers, but lack of competition leaves Sabrent free to charge a much higher price per GB for their QLC NVMe product. Samsung’s vertical integration probably helps them maintain decent profit margins even on their more competitively-priced drive.

Gallery: Sabrent Rocket Q 8TB and Samsung 870 QVO 8TB Teardown

QLC NAND’s Impact on the Consumer SSD Market

The introduction of QLC NAND as a cheaper alternative to three bit per cell TLC NAND has not revolutionized consumer SSD affordability, but it has made higher SSD capacities practical. QLC NAND offers just a 33% increase in theoretical storage density, but in practice most QLC NAND is manufactured as 1024Gbit dies while TLC NAND is manufactured as 256Gbit and 512Gbit dies. This means that it is easier to fit much more flash into the same form factor using QLC than with TLC NAND. Indeed, the Sabrent Rocket Q is bumping up against the practical limits for a M.2 drive.

For drives with more commonplace capacities, QLC NAND has several distinct disadvantages. Storing more bits per physical memory cell requires more precise control over the voltage of each cell, and as a result writing to QLC NAND is much slower than writing to TLC NAND (reading is also a bit slower). That sensitivity to cell voltage also reduces the usable write endurance of QLC NAND before data retention becomes a problem. Drives using QLC NAND have to be rated for fewer drive writes per day (DWPD) in order to meet industry standards for data retention of a worn-out consumer SSD.

However, almost all of those downsides of QLC NAND can be mitigated with sheer capacity. The sequential transfer speed of a single die of NAND flash memory has never been particularly impressive, regardless of how many bits are stored per cell. These 8TB QLC drives use a total of 64 NAND flash dies each, which allows for a lot of parallelism in data transfers (though the SSD’s controller becomes a bottleneck). SLC caching helps address most of the remaining performance problems, but when the SLC cache on a QLC drive runs out, the performance impact is much more severe than for TLC drives.

Write endurance ratings of 0.1 to 0.3 DWPD appear inferior to the 0.5 DWPD or more from good consumer TLC drives, but looking at endurance as a fraction of drive capacity perhaps isn’t the most useful measure for these drives. Both of these 8TB QLC drives are warrantied for over 1TB of writes per day (over 2TB per day for the Samsung drive, but its warranty is only three years rather than five). Most consumer use cases for a multi-TB drive do not involve re-writing most of that data often. Large collections of games, movies and photos can use up the capacity of these 8TB drives, but such infrequently-modified data won’t put much of a dent in the drive’s total write endurance. A mere 3% of the capacity of these drives (240GB) is plenty to hold an OS and most data that will see frequent modification. Filling the rest of the drive with relatively static data won’t hurt the drive’s lifespan.

Who Needs 8TB SSDs?

A capacity of 8TB is a bit on the large side even for mechanical hard drives. Sure, consumer-oriented hard drive product lines are starting to go beyond 14TB, but average capacity sold is much lower. Many use cases for large drives don’t require high performance. An 8TB SSD will offer significant noise and power efficiency advantages over an 8TB hard drive, but either one is adequate for storing a large movie collection. A mechanical hard drive is definitely preferable for long-term archival/backup duty, but an SSD has advantages for keeping data readily accessible.

That’s especially true in a mobile setting, which may be where these 8TB drives make the most sense. Most desktops can accommodate several drives of lower capacity, and so far these 8TB drives carry a significant price premium on a $/GB basis compared to 1TB or 2TB models. But in a notebook, it’s uncommon to have more than one drive bay/slot, and more than three is only found in machines that stretch the definitions of “notebook” and “portable”. So the most sensible or plausible use cases we can imagine for these drives are scenarios that more or less fall into mobile workstation territory. But the sustained write speed of these drives will be a problem when it comes to ingesting uncompressed video even if performance is adequate for editing a large amount of video already residing on one of the drives, so these drives definitely aren’t suitable for every scenario where multiple TBs of data are thrown around.

It’s also easy to imagine other niche use cases for these drives where cost is of no object: a small form factor nearly-silent NAS, for example. (QLC performance doesn’t matter if it sits behind a 1Gbps network bottleneck, and still isn’t much of an issue even with a 10Gbps network.) But for today, we’re going to evaluate these drives with our usual consumer SSD testing methodology.

The Competition

It’s tricky deciding what to compare these 8TB drives against. The use of QLC NAND would traditionally flag these drives as low-end options. But their extreme capacity is unmatched by consumer TLC drives, and the raw cost of 8TB of NAND makes for a high-priced drive overall. Aside from QLC drives, most other low-end consumer SSDs are DRAMless TLC designs—and those product lines mostly top out at 1TB. We’ve included the Mushkin Helix and Toshiba BG4 as representatives of the DRAMless TLC NVMe market segment.

The largest consumer SSDs we have to compare against are Samsung’s earlier 4TB SATA SSDs. We’ve included the 4TB 860 EVO. For some tests, we also have included results from a few enterprise drives: 8TB NVMe models from Intel and SK hynix, and 4TB SATA drives from Kingston and Samsung. These all use TLC NAND, but without SLC caching.

High-end consumer SSD product lines are starting to include more multi-TB capacities, but for now the largest high-end consumer NVMe drives we have on hand are a “mere” 2TB each: Samsung’s 970 EVO Plus and the HP EX950.

AnandTech 2018 Consumer SSD Testbed
CPU Intel Xeon E3 1240 v5
Motherboard ASRock Fatal1ty E3V5 Performance Gaming/OC
Chipset Intel C232
Memory 4x 8GB G.SKILL Ripjaws DDR4-2400 CL15
Graphics AMD Radeon HD 5450, 1920×1200@60Hz
Software Windows 10 x64, version 1709
Linux kernel version 4.14, fio version 3.6
Spectre/Meltdown microcode and OS patches current as of May 2018

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.

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.

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
Average Data Rate
Average Latency Average Read Latency Average Write Latency
99th Percentile Latency 99th Percentile Read Latency 99th Percentile Write Latency
Energy Usage

The Sabrent Rocket Q turns in shockingly good scores on The Destroyer, matching the Samsung 970 EVO Plus, a high-end TLC SSD. The reason why the decidedly less high-end Rocket Q can do this is due entirely to the extreme capacity. For the first time, we have a drive that can handle The Destroyer entirely in its SLC cache. That means the results here are a bit misleading, as the drive would not be able to sustain this level of performance if it was full enough to reduce the SLC cache capacity down to more typical sizes. Power efficiency is also pretty decent here, but again operating out of the SLC cache helps.

Meanwhile, the 8TB Samsung 870 QVO turns in pretty much the same performance scores as the 4TB model, as expected. However, the 8TB drive is a little bit more power-hungry due to the higher part count.

Our Heavy storage benchmark is proportionally more write-heavy than The Destroyer, but much shorter overall. The total writes in the Heavy test aren’t enough to fill the drive, so performance never drops down to steady state. This test is far more representative of a power user’s day to day usage, and is heavily influenced by the drive’s peak performance. The Heavy workload test details can be found here. This test is run twice, once on a freshly erased drive and once after filling the drive with sequential writes.

ATSB Heavy
Average Data Rate
Average Latency Average Read Latency Average Write Latency
99th Percentile Latency 99th Percentile Read Latency 99th Percentile Write Latency
Energy Usage

The Heavy test doesn’t allow the Sabrent Rocket Q a unique advantage from its massive SLC cache; the smaller high-end NVMe drives can also make good use of their caches and overtake the Rocket Q’s performance. However, it does appear that the sheer capacity of the 8TB Rocket Q continues to help significantly on the full-drive test runs. We haven’t measured it directly, but I suspect the minimum SLC cache size reached when the drive is full is still quite a bit larger than what the 2TB and smaller drives have to work with, and that’s how the Rocket Q avoids the horrible latency spikes that the other QLC drives suffer from.

As on The Destroyer, the 8TB Samsung 870 QVO shows no major differences in performance or efficiency from the 4TB model, which means it’s still clearly a bit on the slow side even by SATA standards—especially when full.

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
Average Data Rate
Average Latency Average Read Latency Average Write Latency
99th Percentile Latency 99th Percentile Read Latency 99th Percentile Write Latency
Energy Usage

The 8TB Sabrent Rocket Q offers decent performance on the Light test, even when full: it still provides a large enough SLC cache to handle all the writes from this test. A lot of smaller drives (using QLC or TLC) can’t manage that and show greatly increased write latency on the full-drive test runs.

The 8TB Samsung 870 QVO shows slightly improved latency scores on the full-drive test run compared to the 4TB model, but otherwise performance is the same as expected. As usual, the 8TB QVO is a bit more power-hungry than the smaller versions, and the Rocket Q is considerably more power-hungry than the smaller low-end NVMe drives.

At the end of 2019, UL released a major update to their popular PCMark 10 benchmark suite, adding storage performance tests that had been conspicuously absent for over two years. These new storage benchmarks are similar to our AnandTech Storage Bench (ATSB) tests in that they are composed of traces of real-world IO patterns that are replayed onto the drive under test. We’re incorporating these into our new SSD test suite, and including our first batch of results here.

PCMark 10 provides four different storage benchmarks. The Full System Drive, Quick System Drive and Data Drive benchmarks cover similar territory to our ATSB Heavy and Light tests, and all three together take about as long to run as the ATSB Heavy and Light tests combined. The Drive Performance Consistency Test is clearly meant to one-up The Destroyer and also measure the worst-case performance of a drive that is completely full. Due to time constraints, we are not yet attempting to add the Drive Performance Consistency Test to our usual test suite.

PCMark 10 Storage Tests
Test Name Data Written
Data Drive 15 GB
Quick System Drive 23 GB
Full System Drive 204 GB
Drive Performance Consistency 23 TB + 3x drive capacity

The primary subscores for the PCMark 10 Storage benchmarks are average bandwidth and average latency for read and write IOs. These are combined into an overall score by computing the geometric mean of the bandwidth score and the reciprocal of the latency score. PCMark 10 also records more detailed statistics, but we’ll dig into those in a later review. These PCMark 10 Storage test runs were conducted on our Coffee Lake testbed:

AnandTech Coffee Lake SSD Testbed
CPU Intel Core i7-8700K
Motherboard Gigabyte Aorus H370 Gaming 3 WiFi
Chipset Intel H370
Memory 2x 8GB Kingston DDR4-2666
Case In Win C583
Power Supply Cooler Master G550M
OS Windows 10 64-bit, version 2004

Data Drive Benchmark

The Data Drive Benchmark is intended to represent usage a secondary or portable drive may be subject to. This test simulates copying around files, but does not simulate the IO associated with launching and running applications from a drive.

PCMark 10 Storage - Data
Overall Score Average Bandwidth Average Latency

Starting off, the 8TB Sabrent Rocket Q leads the field thanks to its massive and fast SLC cache; it clearly outperforms even the decently high-end 2TB TLC-based HP EX920. The several capacities of the Samsung 870 QVO all performa about the same: less than half the speed of the faster NVMe drives, and slower than the slowest entry-level NVMe drives. The enterprise SATA drive with no SLC caching comes in last place.

Quick System Drive Benchmark

The Quick System Drive Benchmark is a subset of the Full System Drive Benchmark, running only 6 out of the 23 sub-tests from the Full test.

PCMark 10 Storage - Quick
Overall Score Average Bandwidth Average Latency

Moving on to the Quick test, the Sabrent Rocket Q no longer stands out ahead of the other NVMe drives, but still offers decent performance. The performance gap between the NVMe drives and the Samsung 870 QVO drives has narrowed slightly, but is still almost a factor of two.

Full System Drive Benchmark

The Full System Drive Benchmark covers a broad range of everyday tasks: booting Windows and starting applications and games, using Office and Adobe applications, and file management. The “Full” in the name does not mean that each drive is filled or that the entire capacity of the drive is tested. Rather, it only indicates that all of the PCMark 10 Storage sub-tests are included in this test.

PCMark 10 Storage - Full
Overall Score Average Bandwidth Average Latency

The Full test starts to bring the downsides of QLC NAND into focus. The Sabrent Rocket Q is now the slowest of the NVMe drives, only moderately faster than the 8TB Samsung 870 QVO. The 1TB 870 QVO is also falling behind the larger and faster models. However, the QLC-based Intel 660p manages to hold on to decent performance, possibly a result of the class-leading SLC cache performance we usually see from Silicon Motion NVMe controllers paired with Intel/Micron flash.

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)

The burst random read performance from the 8TB Samsung 870 QVO is even worse than the smaller 870s; even though these drives have the full amount of DRAM necessary to hold the logical to physical address mapping tables, there are other significant sources of overhead affecting the higher capacity models.

The Sabrent Rocket Q’s burst random read performance doesn’t quite fall at the opposite end of the spectrum, but it does clearly offer decent random read latency that is comparable to other drives using the Phison E12(S) controller and not too far behind the NVMe drives using Silicon Motion controllers.

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 QLC drives almost all fare poorly on the longer random read test. The Sabrent Rocket Q falls to be the second-slowest NVMe drive in this batch, and a bit slower than Samsung’s TLC SATA drives. The 8TB Samsung 870 QVO is no longer the slowest capacity; while it is again a bit slower than the 4TB model, the 1TB 870 QVO takes last place in this test.

Sustained 4kB Random Read (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

The power efficiency scores are mostly in line with the performance scores, with the slower drives tending to also be less efficient. The QLC drives follow this pattern quite well. The outliers are the particularly efficient Mushkin Helix DRAMless TLC drive, and the enterprise NVMe SSDs that show poor efficiency because they are underutilized by the low queue depths tested here.

The Sabrent Rocket Q shows good performance scaling as queue depths increase during the random read test. The Samsung 870 QVO seems to be approaching saturation past QD16, even though the SATA interface is capable of delivering higher performance.

Random Reads - All Drives
Sabrent Rocket Q 8TB Samsung 870 QVO 8TB

Comparing the 8TB drives against everything else we’ve tested, neither is breaking new ground. Both drives have power consumption that’s on the high side but not at all unprecedented, and random read performance that doesn’t push the limits of their respective interfaces.

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)

The two 8TB drives have opposite results for the burst random write performance test. The 8TB Sabrent Rocket Q it at the top of the chart with excellent SLC cache write latency, while the 8TB Samsung 870 QVO is a bit slower than the smaller capacities and turns in the worst score in this bunch.

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

The two 8TB drives have opposite results for the burst random write performance test. The 8TB Sabrent Rocket Q it at the top of the chart with excellent SLC cache write latency, while the 8TB Samsung 870 QVO is a bit slower than the smaller capacities and turns in the worst score in this bunch.

Sustained 4kB Random Write (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

Despite their dramatically different random write performance, the two 8TB QLC drives end up with similar power efficiency that’s fairly middle of the road: better than the enterprise drives and the slow DRAMless TLC drives, but clearly worse than the better TLC NVMe drives.

The random write performance of the Rocket Q scales a bit unevenly, but seems to saturate around QD8. Power consumption actually drops after QD4, possibly because the drive is busy enough at that point with random writes that it cuts back on background cleanup work. The Samsung 870 QVO reaches full random write performance at QD4 and steadily maintains that performance through the rest of the test.

Random Writes - All Drives
Sabrent Rocket Q 8TB Samsung 870 QVO 8TB

Unlike on the random read test, the Samsung 870 QVO comes across as having reasonably low power consumption on the random write test, especially at higher queue depths. The Sabrent Rocket Q’s power consumption is still clearly on the high side, especially the spike at QD4 where it seemed to be doing a lot of background work instead of just directing writes to the SLC cache.

Our first test of sequential read performance uses short bursts of 128MB, issued as 128kB operations with no queuing. The test averages performance across eight bursts for a total of 1GB of data transferred from a drive containing 16GB of data. Between each burst the drive is given enough idle time to keep the overall duty cycle at 20%.

Burst 128kB Sequential Read (Queue Depth 1)

Both of the 8TB QLC SSDs provide burst sequential read performance that is on par for their respective market segments. The Sabrent Rocket Q performs similarly to both the Mushkin Helix DRAMless TLC and Intel 660p QLC SSDs. The 8TB Samsung 870 QVO is just a bit slower than the other Samsung SATA SSDs.

Our test of sustained sequential reads uses queue depths from 1 to 32, with the performance and power scores computed as the average of QD1, QD2 and QD4. Each queue depth is tested for up to one minute or 32GB transferred, from a drive containing 64GB of data. This test is run twice: once with the drive prepared by sequentially writing the test data, and again after the random write test has mixed things up, causing fragmentation inside the SSD that isn’t visible to the OS. These two scores represent the two extremes of how the drive would perform under real-world usage, where wear leveling and modifications to some existing data will create some internal fragmentation that degrades performance, but usually not to the extent shown here.

Sustained 128kB Sequential Read

On the longer sequential read tests, the Sabrent Rocket Q starts to fall behind the other low-end NVMe drives, though it still offers competitive performance reading data that was written with random writes. The Samsung 870 QVO holds on to its status as only slightly slower than the other Samsung SATA drives, but due to the SATA bottleneck this is still far slower than any of the NVMe drives.

Sustained 128kB Sequential Read (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

The Sabrent Rocket Q is clearly the least efficient consumer NVMe drive in this bunch for sequential reads of contiguous data; the DRAMless TLC drives outperform it while using much less power, and the more power-hungry high-end TLC SSDs have higher performance to match. The 8TB Samsung 870 QVO again scores just a bit worse than its lower-capacity siblings, because the 8TB model is slightly slower and draws slightly more power.

Like many Phison-based NVMe SSDs, the Sabrent Rocket Q’s sequential read performance doesn’t really begin to scale up until queue depths go beyond 4, explaining its poor low-QD scores above. By QD16 it is basically saturating the PCIe 3 x4 interface. The Samsung 870 QVO saturates the SATA interface starting at QD2.

Sequential Reads - All Drives
Sabrent Rocket Q 8TB Samsung 870 QVO 8TB

While both 8TB drives saturate their respective host interfaces with sequential reads when the queue depths are sufficiently high, they also both draw more power than average among our entire collection of test results. However, neither is power-hungry enough to stand out as an outlier from that crowd.

Our test of sequential write burst performance is structured identically to the sequential read burst performance test save for the direction of the data transfer. Each burst writes 128MB as 128kB operations issued at QD1, for a total of 1GB of data written to a drive containing 16GB of data.

Burst 128kB Sequential Write (Queue Depth 1)

The burst sequential write test primarily illustrates SLC cache performance, and the Sabrent Rocket Q does quite well here, outperforming the rest of the NVMe drives in this bunch. The 8TB Samsung 870 QVO is the slowest drive, but is only slightly slower than the other SATA drives.

Our test of sustained sequential writes is structured identically to our sustained sequential read test, save for the direction of the data transfers. Queue depths range from 1 to 32 and each queue depth is tested for up to one minute or 32GB, followed by up to one minute of idle time for the drive to cool off and perform garbage collection. The test is confined to a 64GB span of the drive.

Sustained 128kB Sequential Write

On the longer sequential write test, the Rocket Q falls behind the high-end consumer NVMe drives but remains clearly faster than the other budget NVMe drives. Meanwhile, the 8TB 870 QVO stays in last place, but is not actually meaningfully slower than the other SATA drives.

Sustained 128kB Sequential Write (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

The Sabrent Rocket Q has the worst power efficiency among the consumer NVMe drives during the sequential write test, but it still offers better performance per Watt than the SATA drives. The 8TB 870 QVO has a lower efficiency score than the other consumer SATA drives, but the enterprise drives are even worse.

Both of the 8TB QLC drives hit their full sequential write speed at QD2 and maintain it for the rest of the test without the SLC cache running out. However, the performance from the Rocket Q is a somewhat variable, probably indicating that it is affected by background work the controller is doing to flush the SLC cache.

Sequential Writes - All Drives
Sabrent Rocket Q 8TB Samsung 870 QVO 8TB

Plotted against the full set of results from all the SATA SSDs we’ve tested, the performance and power consumption of the 8TB 870 QVO on the sequential write test appears to be good but not pushing any limits. The Rocket Q’s performance is higher than most entry-level NVMe drives, but its power consumption creeps up to unusually high levels (over 6W).

Our test of mixed random reads and writes covers mixes varying from pure reads to pure writes at 10% increments. Each mix is tested for up to 1 minute or 32GB of data transferred. The test is conducted with a queue depth of 4, and is limited to a 64GB span of the drive. In between each mix, the drive is given idle time of up to one minute so that the overall duty cycle is 50%.

Mixed 4kB Random Read/Write

The 8TB Sabrent Rocket Q’s performance on the mixed random IO test is much better than any of the other low-end NVMe drives; the DRAMless TLC drives are the slowest in this bunch, and the Intel 660p with its four-channel controller cannot keep up with the Rocket Q’s 8-channel Phison E12. The 8TB Samsung 870 QVO is slower than most of the other SATA drives in this bunch, but still has a clear advantage over the 1TB model.

Sustained 4kB Mixed Random Read/Write (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

The high-end consumer NVMe drives and the Samsung 860 EVO TLC SATA drive top the power efficiency chart for the mixed random IO test. The Sabrent Rocket Q’s efficiency is a significant step down from there, but still a bit better than any of the other low-end drives. The 8TB 870 QVO’s efficiency score is worse than the 4TB model’s, but clearly better than the 1TB model or either of the DRAMless TLC NVMe drives.

Both of the 8TB QLC drives show fairly typical performance curves for the mixed random IO test: little or no performance drop when writes are first added to the mix, and then increasing performance that accelerates toward the end of the test as write caching becomes more effective. The 8TB 870 QVO doesn’t show the signs of a filled SLC cache that we see from the 1TB model, and neither 8TB QLC drive shows the nearly-flat performance exhibited by the two DRAMless TLC drives.

Our test of mixed sequential reads and writes differs from the mixed random I/O test by performing 128kB sequential accesses rather than 4kB accesses at random locations, and the sequential test is conducted at queue depth 1. The range of mixes tested is the same, and the timing and limits on data transfers are also the same as above.

Mixed 128kB Sequential Read/Write

The Sabrent Rocket Q’s performance on the mixed sequential IO test is competitive with the high-end consumer TLC drives, and far better than the other low-end NVMe options. The 8TB Samsung 870 QVO has distinctly lower performance than the smaller capacities, but isn’t quite the worst overall performer.

Sustained 128kB Mixed Sequential Read/Write (Power Efficiency)
Power Efficiency in MB/s/W Average Power in W

The good performance of the Rocket Q on the mixed sequential IO test comes at the cost of worse power efficiency than the DRAMless TLC competition, but its efficiency scores are still decent. The 8TB 870 QVO’s efficiency scores are worse than any of the other consumer SSDs in this bunch.

As with several other synthetic tests in our suite, the mixed sequential IO test has the Sabrent Rocket Q showing rather variable performance, though fortunately without any severe drops. It performs a bit better during the more write-heavy half of the test.

The Samsung 870 QVO shows relatively flat and consistent performance throughout this test, but as is common for Samsung drives there’s a bit of a decreasing performance trend during the read-heavy half of the test.

Real-world client storage workloads leave SSDs idle most of the time, so the active power measurements presented earlier in this review only account for a small part of what determines a drive’s suitability for battery-powered use. Especially under light use, the power efficiency of a SSD is determined mostly be how well it can save power when idle.

For many NVMe SSDs, the closely related matter of thermal management can also be important. M.2 SSDs can concentrate a lot of power in a very small space. They may also be used in locations with high ambient temperatures and poor cooling, such as tucked under a GPU on a desktop motherboard, or in a poorly-ventilated notebook.

Sabrent Rocket Q 8TB
NVMe Power and Thermal Management Features
Controller Phison E12S
Firmware RKT30Q.2 (ECFM52.2)
NVMe
Version
Feature Status
1.0 Number of operational (active) power states 3
1.1 Number of non-operational (idle) power states 2
Autonomous Power State Transition (APST) Supported
1.2 Warning Temperature 75°C
Critical Temperature 80°C
1.3 Host Controlled Thermal Management Supported
 Non-Operational Power State Permissive Mode Supported

The Sabrent Rocket Q claims support for the full range of NVMe power and thermal management features. However, the table of power states includes frighteningly high maximum power draw numbers for the active power states—over 17 W is really pushing it for a M.2 drive. Fortunately, we never measured consumption getting that high. The idle power states look typical, including the promise of quick transitions in and out of idle.

Sabrent Rocket Q 8TB
NVMe Power States
Controller Phison E12S
Firmware RKT30Q.2 (ECFM52.2)
Power
State
Maximum
Power
Active/Idle Entry
Latency
Exit
Latency
PS 0 17.18 W Active
PS 1 10.58 W Active
PS 2 7.28 W Active
PS 3 49 mW Idle 2 ms 2 ms
PS 4 1.8 mW Idle 25 ms 25 ms

Note that the above tables reflect only the information provided by the drive to the OS. The power and latency numbers are often very conservative estimates, but they are what the OS uses to determine which idle states to use and how long to wait before dropping to a deeper idle state.

SATA SSDs are tested with SATA link power management disabled to measure their active idle power draw, and with it enabled for the deeper idle power consumption score and the idle wake-up latency test. Our testbed, like any ordinary desktop system, cannot trigger the deepest DevSleep idle state.

Idle power management for NVMe SSDs is far more complicated than for SATA SSDs. NVMe SSDs can support several different idle power states, and through the Autonomous Power State Transition (APST) feature the operating system can set a drive’s policy for when to drop down to a lower power state. There is typically a tradeoff in that lower-power states take longer to enter and wake up from, so the choice about what power states to use may differ for desktop and notebooks, and depending on which NVMe driver is in use. Additionally, there are multiple degrees of PCIe link power savings possible through Active State Power Management (APSM).

We report three idle power measurements. Active idle is representative of a typical desktop, where none of the advanced PCIe link power saving features are enabled and the drive is immediately ready to process new commands. Our Desktop Idle number represents what can usually be expected from a desktop system that is configured to enable SATA link power management, PCIe ASPM and NVMe APST, but where the lowest PCIe L1.2 link power states are not available. The Laptop Idle number represents the maximum power savings possible with all the NVMe and PCIe power management features in use—usually the default for a battery-powered system but not always achievable on a desktop even after changing BIOS and OS settings. Since we don’t have a way to enable SATA DevSleep on any of our testbeds, SATA drives are omitted from the Laptop Idle charts.

Note: Last year we upgraded our power measurement equipment and switched to measuring idle power on our Coffee Lake desktop, our first SSD testbed to have fully-functional PCIe power management. The below measurements are not a perfect match for the older measurements in our reviews from before that switch.

Idle Power Consumption - No PMIdle Power Consumption - DesktopIdle Power Consumption - Laptop

The Samsung 870 QVO SSDs have lower active idle power consumption than the NVMe competition, though our measurements of the 4TB model did catch it while it was still doing some background work. With SATA link power management enabled the 8TB 870 QVO draws more power than the smaller models, but is still very reasonable.

The Sabrent Rocket Q’s idle power numbers are all decent but not surprising. The desktop idle power draw is significantly higher than the 49mW the drive claims for power state 3, but it’s still only at 87mW which is not a problem.

Idle Wake-Up Latency

The Samsung 870 QVO takes 1ms to wake up from sleep. The Sabrent Rocket Q has almost no measurable wake-up latency from the intermediate desktop idle state, but takes a remarkably long 108ms to wake up from the deepest sleep state. This is one of the slowest wake-up times we’ve measured from a NVMe drive and considerably worse than the 25ms latency the drive itself promises to the OS.

The sheer capacity alone is enough to make the 8TB Sabrent Rocket Q and 8TB Samsung 870 QVO impressive and groundbreaking products. But reaching this new capacity point for consumer SSDs has required significant tradeoffs. These two drives rely on QLC NAND flash memory with worse performance and write endurance than the TLC NAND used by mainstream consumer SSDs. Thankfully, the sheer high capacity of these drives offsets some of the downsides of QLC NAND, but it but does not eliminate all of them.

The result is a pair of drives that blur the lines between low-end and premium products. The price tags are unquestionably premium territory, and even on a per-GB basis these drives aren’t the cheapest. Rather than offering economies of scale, the niche status of such high-capacity SSDs carries a bit of a price premium. This is especially true of the 8TB Sabrent Rocket Q: it is currently at its cheapest-ever price, but is still 45% more expensive than the 8TB Samsung 870 QVO. The Rocket Q’s use of an NVMe controller (rather than a SATA controller) only accounts for a few dollars of this vast difference. Sabrent is probably paying more to buy Micron’s QLC on the open market than it costs Samsung to use their own QLC, but a large portion of this price disparity can simply be blamed on lack of competition. The Sabrent Rocket Q was the first 8TB consumer NVMe SSD, and only one competitor has showed up since: the Corsair MP400, based on the same basic formula as the Rocket Q.

While its price tag certainly appears exorbitant next to the cheaper Samsung 870 QVO, there’s no question that the 8TB Rocket Q deserves more premium pricing. The Samsung 870 QVO is slow even by SATA SSD standards, and is best used as a secondary drive for bulk data with low performance requirements. Ignoring the price, it looks great in comparison to an 8TB hard drive: silent, faster (usually), more compact. But compared against other SSDs it is lackluster. The fact that it’s no faster than the 2TB and 4TB models is another disappointment, and a clear sign that 8TB is far beyond the sweet spot of the SSD market.

The Rocket Q on the other hand is fast enough to provide a good experience as a primary drive, even if it gets loaded down with several TB of data. It won’t always match the performance of a smaller high-end drive, but it doesn’t suffer as much from the worst-case performance problems that plague most QLC SSDs (and likely the smaller capacities of the Rocket Q as well). At its worst, the Rocket Q only degrades down to a bit slower than mainstream SATA drives. Rocket Q doesn’t quite manage to provide that magical combination of maximum capacity and maximum performance, but comes surprisingly close.

High-Capacity Consumer SSD Price Comparison
December 4, 2020
  1TB 2TB 4TB 8TB
ADATA XPG SX8100
TLC
$119.99 (12¢/GB) $229.99 (11¢/GB) $499.99 (12¢/GB)  
Addlink S92
QLC
$145.88 (15¢/GB) $277.88 (14¢/GB) $649.99 (16¢/GB)  
Corsair MP400
QLC
$137.00 (14¢/GB) $288.00 (14¢/GB) $662.00 (17¢/GB) $1498.00 (19¢/GB)
Corsair MP510
TLC
$142.99 (15¢/GB) $289.99 (15¢/GB) $744.99 (19¢/GB)  
Inland Platinum
QLC
$94.99 (9¢/GB) $191.99 (10¢/GB) $499.99 (12¢/GB)  
Sabrent Rocket Q
QLC
$109.98 (11¢/GB) $219.98 (11¢/GB) $599.98 (15¢/GB) $1299.99 (16¢/GB)
Sabrent Rocket Q 4.0
QLC, PCIe Gen4
$149.98 (15¢/GB) $279.98 (14¢/GB) $689.98 (17¢/GB)  
Sabrent Rocket
TLC
$129.98 (13¢/GB) $249.98 (12¢/GB) $699.99 (17¢/GB)  
WD Black AN1500
TLC, PCIe Gen3 x8
$299.99 (30¢/GB) $549.99 (27¢/GB) $999.99 (25¢/GB)  
SATA SSDs:
Samsung 870 QVO
QLC
$89.99 (9¢/GB) $199.99 (10¢/GB) $419.99 (10¢/GB) $899.99 (11¢/GB)
Samsung 860 EVO
TLC
$99.99 (10¢/GB) $199.99 (10¢/GB) $540.99 (14¢/GB)  
WD Blue 3D
TLC
$104.99 (10¢/GB) $179.00 (9¢/GB) $499.99 (12¢/GB)  

Looking at the overall state of pricing in the SSD market, among NVMe drives, the current 8TB options are the Sabrent Rocket Q and the Corsair MP400, which use almost identical hardware. The Sabrent Rocket Q currently has better pricing than the more recently-released MP400. Dropping down to less extreme capacities, neither product is the best option. Microcenter’s Inland Platinum is their version of the Phison E12 with QLC, and it’s cheaper than the Rocket Q at 1TB, 2TB and 4TB. There’s also the ADATA XPG SX8100, by far the cheapest multi-TB NVMe SSD with TLC NAND. It uses Realtek’s RTS5762 controller so it’s really not a high-end drive even by PCIe 3 standards, but it’s definitely a step up from the QLC drives, especially for heavier workloads. The 4TB SX8100 is currently $499 and was recently on sale for $399.

 

In the consumer SATA SSD market, there are far fewer options for very large drives. The 870 QVO is unopposed at the 8TB capacity, and the only 4TB alternatives are TLC drives. However, the 4TB WD Blue at 20% more than the 4TB 870 QVO seems like a pretty good upgrade. At 1TB and 2TB the 870 QVO is uncompetitive: the 860 EVO is currently only $10 more at 1TB, and the same price at 2TB.

 

Looking Forward

For most consumers, 8TB SSDs will not become a realistic proposition for several more generations of 3D NAND technology. These drives are an early preview of that future, and highlight what else needs to improve aside from just the price. Even though QLC NAND has a reputation for poor performance, both of these 8TB drives are often bottlenecked instead by the controller: partly a result of putting 64 NAND flash dies behind 8 channel controllers. The consumer SSD market is unlikely to reverse direction and start moving towards wider controllers, so in order for 8TB drives to go mainstream without the limitations of today’s models, we’ll need to see higher per-die capacities and much higher IO speeds per channel.

Higher die capacities will go hand in hand with cost reductions in future generations of 3D NAND flash memory, and by the time 8TB drives are mainstream we’ll probably see 1TB drives as the same kind of baseline that 256GB drives are today. Movement toward higher interface speeds between the NAND and controller is already underway, spurred on by the arrival of PCIe 4.0. There’s now demand for 4-channel NVMe SSD controllers capable of several GB/s, which requires NAND interface speeds far in excess of what the Sabrent Rocket Q’s Phison E12 is capable of.

We will soon be continuing our exploration of newer QLC SSDs with a look at the 1TB Corsair MP400, which should be very similar to the 1TB Rocket Q. At lower capacities, the limitations of QLC NAND are a bigger challenge, and there’s more competition from entry-level TLC drives. We’re also testing the Sabrent Rocket Q4, the PCIe 4.0 successor to the Rocket Q—another hybrid of high-end and low-end features. However, this one currently only goes up to 4TB.