Samsung SSD 860 QVO 1 TB and 4 TB: the first consumer SATA QLC (part 3)

Original author: Billy Tallis
  • Transfer
Part 1 >> Part 2 >> Part 3

Mixed performance test - random read / write


The mixed test uses mixes of operations from pure reading to pure record, with a step of 10%. Each mix is ​​tested up to 1 minute, or up to 32 GB of transferred data. The test is performed with a queue depth of 4 and is limited to a disk capacity of 64 GB. Between each operation package, the drive is given idle time up to one minute, so the total duty cycle is 50%.



The performance of the Samsung 860 QVO in the mixed random I / O test is significantly lower than that of the 860 EVO, but comparable to other mainstream TLC drives. Running a full disk test significantly slows down the 1 TB QVO 860, but it still outperforms the TLC drive without DRAM.





Comparison charts

Samsung 860 QVO 1TB


Samsung 860 QVO 1TB (Full)


Samsung 860 QVO 4TB


Samsung 860 QVO 4TB (Full)


Samsung 850 PRO 1TB


Samsung 860 EVO 1TB


Samsung 860 EVO 4TB


Crucial MX300 2050GB


Crucial MX500 1TB


Crucial P1 1000GB


Crucial P1 1000GB (Full)


Intel SSD 660p 1TB


Intel SSD 660p 1TB (Full drive)


Toshiba TR200 960GB


WD Black 7200RPM 1TB


WD Blue 1TB 3D NAND


Energy efficiency ratings for 860 QVO are not much better than performance ratings. The power consumption is slightly higher than that of the 860 EVO, but does not greatly depend on the capacity or state of the fill, so the energy efficiency indicators roughly reflect the performance indicators.

The 860 QVO starts at a fairly low random read speed, but the performance increases steadily as the workload shifts to the write side, eventually overtaking the 860 EVO. When the test is performed on a full drive, the 1TB 860 QVO exhausts the SLC cache in the final stages of the test and slows down, instead of acceleration.

Mixed performance test - sequential read / write


The mixed sequential read and write test differs from the previous one by performing sequential access of 128 KB, instead of 4 KB in random places, and is performed at queue depth 1. The range of test mixes is the same, and the timing and restrictions on data transfer are the same as in the test above .



The 4TB 860 QVO does a pretty good job with the mixed serial I / O test, but the 1TB model is slightly slower than the DRAMless TLC drive, and significantly behind the main TLC drives.





The power efficiency of SATA drives differs more than the performance, so the 860 EVO and Toshiba TR200 are particularly effective, while the 860 QVO 4 TB is average and the TB model is experiencing difficulty.

Comparison charts

Samsung 860 QVO 1TB


Samsung 860 QVO 1TB (Full)


Samsung 860 QVO 4TB


Samsung 860 QVO 4TB (Full)


Samsung 850 PRO 1TB


Samsung 860 EVO 1TB


Samsung 860 EVO 4TB


Crucial MX300 2050GB


Crucial MX500 1TB


Crucial P1 1000GB


Crucial P1 1000GB (Full)


Intel SSD 660p 1TB


Intel SSD 660p 1TB (Full drive)


Toshiba TR200 960GB


WD Black 7200RPM 1TB


WD Blue 1TB 3D NAND


Both models of the 860 QVO provide decent performance at either end of the test with pure reading or pure writing, and do worse with mixed operations. The 1TB 860 QVO sinks much more in performance during the first two thirds of the test, but in the end it pulls up to the 4TB model.

Idle power consumption


SSD testing is conducted with the SATA channel power control disabled, to measure their power consumption in active standby mode, as well as to measure in deep standby mode and to check for “wake-up” delay. Our test bench, like any ordinary desktop system, cannot cause a deep idle state of DevSleep.

Power management in standby mode for NVMe solid-state drives is much more complicated than for SSD SATA. NVMe drives can support several different idle states, and using the Autonomous Power State Transition function - APST, the operating system can set a drive behavior policy in case you need to switch to low power mode. As a rule, the trade-off is that low-power states take longer to enter and wake up, so the choice of which power states to use may differ between desktops and laptops.

We provide the results of two idle state measurements. Active idle mode is typical of a typical desktop computer, in which none of the advanced PCIe connectivity features or NVMe power savings are enabled, and the drive is immediately ready for processing new commands. Standby power consumption is measured when the PCIe Active State Power L1.2 is turned on and NVMe APST is on, if supported.



It looks like the 1TB Samsung 860 QVO was still busy with background processing a few minutes after the test data was written to disk, so our automatic idle power measurement found that it still consumes 2 watts. 4 TB dropped the SLC cache much faster and showed a good indicator of energy consumption in active standby mode. Both disks show a good result in the deep standby mode, although we saw a consumption slightly higher than the officially stated 30 mW.



The wake-up delay for the 860 QVO is the same as for other SATA SSDs, about 1.2 ms. This is not the best that can be obtained on SATA, but there is nothing to complain about.

findings


The Samsung 860 QVO is not the first QLC consumer drive we have tested, but in many ways it fits better with our expectations for the new QLC technology than the Intel 660p and Crucial P1. These NVMe SSDs are poorly suited to meet the demand for a cheap entry-level drive, or a high-capacity drive — two applications where QLC NAND seems most useful. We sometimes perceived QLC as a replacement for a hard disk, and not as a high-speed drive. It was strange when QLC first appeared on NVMe solid-state drives. In contrast, the 860 QVO is a predictable product that has not brought any surprises in its design. Samsung builds on a tried and true formula by simply adapting the 860 EVO to work with QLC NAND.

QLC NAND is based on sacrificing quality for quantity. The viability of QLC drives is based on the assumption that existing drives are fast enough, which is true for many Samsung SSDs. The Samsung 860 QVO drive is not as fast or energy efficient as the 860 EVO, but it’s not necessary. Samsung, as a rule, stays away from the “entry-level” segment in the SSD market, and in their product line there is room for such a product for much longer than QLC technology itself.



As with the two other tested QLC drives, it is important to note that the use of QLC NAND does not have a revolutionary effect on the final product. Model 860 QVO is still suitable for consumer purposes. It is slower than the 860 EVO, and yet much faster than the slowest SATA SSD we tested. Thanks to the combination of SLC caching and the limitations of the SATA channel, the behavior of the 860 QVO is often indistinguishable from other SATA solid-state drives. Based only on test results, it would be difficult to definitively define QVO as a QLC-based disk, and not just as a relatively slow TLC disk. The true benefits are the stable write performance after filling the SLC cache and the idle time required to recover the disk after using the write cache.

From a technological point of view, QLC NAND seems to have an impact on the consumer storage market. The speed is large enough to leave the hard drives far behind, and the write speed is still adequate. Samsung deserves praise at least for the proposal of 860 QVO in the amount of at least 1 TB. Competitors using QLC in small disks will face shortcomings that are hard to miss. Despite the fact that Samsung introduces a lower level product, they do not let their products go out of control.

Thanks to the Intel and Micron QLC drives using NVMe, there are many ways to compare Samsung's QLC with competing Intel / Micron QLC. Judging by our testing, there is not yet a clear winner. Tests in which the 860 QVO goes beyond the SATA interface are useless. Among other tests, Intel / Micron QLC, as a rule, is slightly faster, but in part it is still connected with the NVMe interface. The energy efficiency between the two QLC projects is generally similar.



The disadvantages of QLC NAND, be they small or serious, are accepted in exchange for the promise of availability. Other things being equal, QLC NAND ideally should be 25% cheaper than TLC NAND. There are several reasons why this goal is currently unattainable, and the few QLC drives that we have are still not able to provide improved availability. Prices for NAND flash memory are falling across the board, so now is not the best time to try using new technologies for price pressure. Most likely, the 860 QVO will suffer the same fate as many entry-level drives - DRAMless SATA SSD: mainstream drives are at the forefront of falling prices, and often reduce the gap with the entry-level drives to almost zero.

Samsung's MSRP for 860 QVOs reflect this. Current street prices for 860 EVO are lower than 860 QVO for two of the three capacities, and this is in comparison with one of the best SATA solid-state drives. There are a lot of basic drives with slightly lower performance. The exception is the 4 TB segment, where Samsung has no analogues. The 4 TB segment is just beginning to wake up, but at $ 400 for a QVO 4 TB, it is still outside the normal consumer price range. It may be worth reviewing the 860 QVO after a few months to check prices.

Samsung plans that the 860 QVO will be available for purchase from December 16. By that time, shaking prices due to holiday sales should pass, and Samsung will have the opportunity to reconsider its prices. Meanwhile, the 860 EVO is clearly the best choice.

Thank you for staying with us. Do you like our articles? Want to see more interesting materials? Support us by placing an order or recommending to friends, 30% discount for Habr users on a unique analogue of the entry-level servers that we invented for you: The whole truth about VPS (KVM) E5-2650 v4 (6 Cores) 10GB DDR4 240GB SSD 1Gbps from $ 20 or how to share the server? (Options are available with RAID1 and RAID10, up to 24 cores and up to 40GB DDR4).

VPS (KVM) E5-2650 v4 (6 Cores) 10GB DDR4 240GB SSD 1Gbps until spring for free if you pay for a period of six months, you can order here .

Dell R730xd 2 times cheaper? Only we have 2 x Intel Dodeca-Core Xeon E5-2650v4 128GB DDR4 6x480GB SSD 1Gbps 100 TV from $ 249in the Netherlands and the USA! Read about How to build an infrastructure building. class c using servers Dell R730xd E5-2650 v4 worth 9000 euros for a penny?

Also popular now: