Original Link: https://www.anandtech.com/show/10274/the-crucial-mx300-750gb-ssd-review-microns-3d-nand-arrives
The Crucial MX300 750GB SSD Review: Micron's 3D NAND Arrives
by Billy Tallis on June 14, 2016 9:00 AM EST
Micron is the second company to the 3D NAND flash milestone, with today's release of the Crucial MX300 750GB Limited Edition. This follows on from two years ago when Samsung became the first manufacturer to ship SSDs with 3D NAND flash, the 850 Pro - a drive launched as a flagship high end drive and remains the fastest consumer SATA drive today. By contrast, the Crucial MX300 is intended to be a mainstream affordable SSD based on 3D TLC NAND, making the MX300 more of a direct competitor to the 850 Evo.
Micron's strategy for transitioning to 3D NAND is quite different from the path that Samsung took. Similar to other projects, Micron is once more partnered with Intel to produce this generation of flash memory, and their 3D NAND is different compared to Samsung in several ways. While the rest of the industry chose to combine the 3D NAND transition with a switch from a floating gate memory cell to a charge trap design, Intel and Micron's 3D NAND is still a floating gate design, making the end-point a 'simpler' two-step process. As a result, the capacity is also quite different: Samsung's 3D NAND debuted with a 32-layer but small 86Gb MLC die and 128Gb TLC, later followed up with a 128Gb MLC, whereas the Intel/Micron first effort on 32-layer 3D NAND is an MLC part with a capacity of 256Gb and a TLC part with 384Gb, making for a much larger die. While Samsung has continued to develop planar NAND alongside their 3D NAND and recently reintroduced planar NAND to their retail SSD family with the 750 EVO, Micron is committed to going all-3D with new products. To be clear however, Micron's existing planar NAND products won't be going away quickly (especially in the enterprise market).
The Crucial MX300 750GB
The MX300 was first made official at a Micron enterprise storage event with a release planned for April/early Q2. While the launch of the full range of drives was pushed back to allow for a new firmware revision with improved performance to go through a QA cycle, Micron is sticking with the original plan of first launching a limited edition single 750GB capacity, with the full range of capacities and M.2 versions to fill in over the rest of the year. With seven extra weeks of firmware work behind them, supplies of the 750GB MX300 may not be as limited as they would have been for an April launch, but availability could be temporarily constrained as the drives with the updated firmware make their way through the supply chain. Despite this, we also expect that a tight supply of the 3D TLC was a contributing factor in a staggered release, and as a result we start with the Limited Edition first.
750GB is not a typical capacity for a SSD, but the 384Gb (48GB) die capacity of Micron's 3D TLC makes it hard to hit the usual power of two drive capacity points without having a much larger overprovisioning ratio than is typical of a consumer drive. Instead, the MX300 line will probably be the first time we see TLC clearly resulting in an increase of capacity relative to MLC drives across the full range of products, rather than just offering a lower price per GB (and one extra capacity at the high end). Micron hasn't nailed down the final specifications for the other capacities of MX300, but we do know that the smallest drive will use six dies. This will give it a raw capacity of 288GB and the usable capacity may be around 275GB rather than something in the 240-256GB range. The largest (and last to launch) capacity will be in the neighborhood of 2TB. The 750GB model uses sixteen dies arranged in eight packages of two dies each for a raw capacity of 768GB.
In the past, when NAND transitions to a smaller lithography node and a larger die capacity, the reduction in die count has caused a decrease in available parallelism and thus lower performance for drives of the same capacity. Micron is offsetting this by partitioning their 3D NAND into a four plane design, where 128Gb planar flash chips are at most two plane designs. This allows for almost as many commands to be processed simultaneously by the 3D TLC, but will also lead to somewhat higher contention on the link between the controller and the flash.
| Crucial MX Series Specifications | ||||
| Capacity | MX300 750GB | MX200 500GB | MX200 1TB | |
| Form Factors | 2.5" | 2.5", mSATA, M.2 | 2.5" | |
| Controller | Marvell 88SS1074 | Marvell 88SS9189 | ||
| NAND | Micron 384Gbit 32-layer 3D TLC |
Micron 16nm 128Gbit MLC |
||
| DRAM | 512MB | 512MB | 1GB | |
| Sequential Read | 530MB/s | 555MB/s | 555MB/s | |
| Sequential Write | 510MB/s | 500MB/s | 500MB/s | |
| 4KB Random Read | 92K IOPS | 100K IOPS | 100K IOPS | |
| 4KB Random Write | 83K IOPS | 87K IOPS | 87K IOPS | |
| Dynamic Write Acceleration | Yes | Yes (mSATA and M.2 models only) | No | |
| DevSleep Power | 4mW | 2mW | 2mW | |
| Slumber Power | 75mW | 100mW | 100mW | |
| Max Power | 5.2W | 4.7W | 5.2W | |
| Encryption | TCG Opal 2.0 & IEEE-1667 (eDrive) | |||
| Endurance | 220TB | 160TB | 320TB | |
| Warranty | Three years | |||
| Price | $199.99 (MSRP) | $139.00 | $269.94 | |
Aside from adopting 3D TLC and upgrading to the latest in Marvell's long line of SATA SSD controllers, the MX300 offers all the usual features of the MX series: encryption support, SLC caching, partial power loss protection and a three year warranty. Performance and endurance specifications are down slightly from the MX200 due to the new hardware, but are still fine for a mainstream SATA SSD. The final decisions about overprovisioning for the other capacity models could mean they will have better endurance ratings than the MX200 of comparable capacity.
From the exterior the MX300 appears virtually identical to a MX200. Opening it up to look at the PCB, we see a full size circuit board with plenty of room for more NAND, but Micron won't actually need more than the 8 packages in this drive for even the largest 2TB class model. This 750GB drive stacks just two dies per package; a 2TB class drive would need at most six dies per package, but stacking 8 or 16 dies per package is common when dealing with smaller NAND chips and especially smaller form factors. The pads for the power loss protection capacitors are almost all populated.
The 750GB Crucial MX300 comes with a MSRP of $199.99. On a price per GB basis, this puts it on par with the current actual retail prices for the MX200 and we expect competition can drive the MX300 price down further. It will probably take up position near the top of the budget planar TLC segment of the market and be cheaper than all but the best sale prices on MLC drives. So while at first glance the MX300 might appear to be the natural match against the 850 EVO, Micron is clearly aiming lower in price.
In addition to the Crucial MX200 and 3D TLC based 850 EVO, the MX300 will compete against the top performing planar TLC drives like the SanDisk X400, which is based on the same Marvell 88SS1074 controller used in the MX300.
| AnandTech 2015 SSD Test System | |
| CPU | Intel Core i7-4770K running at 3.5GHz (Turbo & EIST enabled, C-states disabled) |
| Motherboard | ASUS Z97 Pro (BIOS 2701) |
| Chipset | Intel Z97 |
| Memory | Corsair Vengeance DDR3-1866 2x8GB (9-10-9-27 2T) |
| Graphics | Intel HD Graphics 4600 |
| Desktop Resolution | 1920 x 1200 |
| OS | Windows 8.1 x64 |
- Thanks to Intel for the Core i7-4770K CPU
- Thanks to ASUS for the Z97 Deluxe motherboard
- Thanks to Corsair for the Vengeance 16GB DDR3-1866 DRAM kit, RM750 power supply, Carbide 200R case, and Hydro H60 CPU cooler
Performance Consistency
Our performance consistency test explores the extent to which a drive can reliably sustain performance during a long-duration random write test. Specifications for consumer drives typically list peak performance numbers only attainable in ideal conditions. The performance in a worst-case scenario can be drastically different as over the course of a long test drives can run out of spare area, have to start performing garbage collection, and sometimes even reach power or thermal limits.
In addition to an overall decline in performance, a long test can show patterns in how performance varies on shorter timescales. Some drives will exhibit very little variance in performance from second to second, while others will show massive drops in performance during each garbage collection cycle but otherwise maintain good performance, and others show constantly wide variance. If a drive periodically slows to hard drive levels of performance, it may feel slow to use even if its overall average performance is very high.
To maximally stress the drive's controller and force it to perform garbage collection and wear leveling, this test conducts 4kB random writes with a queue depth of 32. The drive is filled before the start of the test, and the test duration is one hour. Any spare area will be exhausted early in the test and by the end of the hour even the largest drives with the most overprovisioning will have reached a steady state. We use the last 400 seconds of the test to score the drive both on steady-state average writes per second and on its performance divided by the standard deviation.
The MX300's steady state random write performance is below the MX200 but stays slightly ahead of the 850 EVO and well ahead of any planar TLC drive.
The MX300 earns a consistency score that is below the MX200 but about average, and certainly high enough for ordinary consumer workloads.
 |
|||||||||
| Default | |||||||||
| 25% Over-Provisioning | |||||||||
The MX300 lasts for over eight minutes before the SLC cache and spare area are completely exhausted, but during that phase it alternates between operating at about 22k IOPS and 65-70k IOPS. During the steady state phase there is gradual improvement in performance along with an increase in variability.
 |
|||||||||
| Default | |||||||||
| 25% Over-Provisioning | |||||||||
The steady state random write performance of the MX300 varies periodically between about 6k IOPS to 10k IOPS, with no outliers of slow performance. With extra overprovisioning performance is substantially higher and consistency is a bit better.
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 and unlike our Iometer tests, 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.
We quantify performance on this test by reporting the drive's average data throughput, a few data points about its latency, and the total energy used by the drive over the course of the test.
The average data rate for the MX300 is between the scores of the 1TB and 500GB MX200 but a little closer to the latter. This is acceptable performance but not really an upgrade over the MX200.
The average service time of the MX300 is slow enough to put it among the budget planar TLC drives and well behind the SanDisk X400.
The MX300 has a higher frequency of both moderate and severe latency outliers than the MX200, but it isn't a huge discrepancy. The performance relative to the SanDisk X400 shows that there is room for improvement.
The MX300's power consumption comes in a distant second place behind the BX100, but this is still remarkable efficiency for a TLC drive.
AnandTech Storage Bench - Heavy
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.
The average data rate of the MX300 is virtually identical to that of its predecessors, except that when full the MX300's identity as a TLC drive is revealed by a precipitous drop.
The average service time of the MX300 when running the test on an empty drive places the MX300 between the slow MLC drives and the fast planar TLC. When the test is run on a full drive the average service time grows to be close to that of some of the cheapest drives on the market.
When run on an empty drive, the write-oriented Heavy test doesn't produce a significant number of latency outliers, but on a full drive delays are quite common.
The MX300's power efficiency is once again very close to the top of the charts, even for the full-drive test.
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.
The MX300 establishes a small but clear lead in average data rate over the MX200 when the test is run on an empty drive, but for a full drive the performance again drops down to the level of budget planar TLC drives.
As with the Heavy test, the MX300 has an average service time on the Light test that puts it below the MLC drives but ahead of most TLC drives except when full.
The frequency of latency outliers shows the same split personality: keep plenty of spare area around and the MX300 performs like a (low-end) MLC drive, but fill it up and it will begin to struggle.
For power efficiency the MX300 is once again second only to the Crucial BX100, and substantially better than any other TLC.
Random Read Performance
The random read test requests 4kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test spans the entire drive, which is filled before the test starts. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
The random read performance of the MX300 is a major regression relative to the MX200, which is a disappointment given that both the SanDisk X400 with planar TLC and the Samsung 850 EVO with 3D TLC manage to perform close to the MX200's level. Since these scores are based on performance at low queue depths of reads that cannot be cached, they reflect an inherently high read latency of Micron's 3D TLC.
Despite the MX300 being close to the top of the power usage chart on this test, the poor performance means that the efficiency is not that great considering.
 |
|||||||||
The MX300's random read performance does increase with queue depth, but not fast enough for it to catch up to the competition or reach a plateau.
Random Write Performance
The random write test writes 4kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test is limited to a 16GB portion of the drive, and the drive is empty save for the 16GB test file. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
The MX300's random write speed is slightly worse than the MX200, but it is still quite clearly in MLC territory well above the performance level of planar TLC drives.
The MX 300 breaks the power usage record by a shockingly wide margin. Since this average is based on the first 9 minutes of the test, the drive is mostly writing to its SLC cache, which it can apparently do with very little power.
 |
|||||||||
Performance scales very well from QD1 to QD4, but the limit reached by the MX300 is slightly lower than the fastest competitors and there's a slight drop at the end suggesting that the SLC cache began to run out.
Sequential Read Performance
The sequential read test requests 128kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test spans the entire drive, and the drive is filled before the test begins. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
The earlier MX series drives had somewhat poor sequential read speeds, though the spread from best to worst is quite small. The MX300 brings things up to average.
The MX300 is still quite efficient, but the SX930 and BX100 500GB both beat it by a wide margin.
 |
|||||||||
With a higher starting point, the MX300's performance scaling is not as pronounced as the MX200's. The MX300 is a little bit slower at higher queue depths.
Sequential Write Performance
The sequential write test writes 128kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test spans the entire drive, and the drive is filled before the test begins. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
Sequential write speeds on the MX300 are much slower than the MX200 or any other MLC drive, but it does manage to come out ahead of all the planar TLC drives.
Power efficiency is still great by planar TLC standards, but the Samsung 850 EVO managed to saturate the SATA connection while using the same amount of power.
 |
|||||||||
The MX300 shows essentially no scaling with queue depth. For this test we usually don't see much scaling as caching large writes allows for full performance even at low queue depths, but full performance for the MX300 is still disappointing.
Mixed Random Read/Write Performance
The mixed random I/O benchmark starts with a pure read test and gradually increases the proportion of writes, finishing with pure writes. The queue depth is 3 for the entire test and each subtest lasts for 3 minutes, for a total test duration of 18 minutes. As with the pure random write test, this test is restricted to a 16GB span of the drive, which is empty save for the 16GB test file.
Despite poor random read speeds, the MX300 is only slightly slower than the MX200 on mixed random workloads, and is faster than most MLC drives.
Once more setting a power usage record, the MX300 is more efficient than even the BX100.
 |
|||||||||
The MX300 manages to never decrease in performance when the proportion or writes increases, showing that its SLC write caching is working very effectively. Power consumption doesn't begin to increase until the test is almost completely writes.
Mixed Sequential Read/Write Performance
The mixed sequential access test covers the entire span of the drive and uses a queue depth of one. It starts with a pure read test and gradually increases the proportion of writes, finishing with pure writes. Each subtest lasts for 3 minutes, for a total test duration of 18 minutes. The drive is filled before the test starts.
Since SATA drives all perform about the same on sequential reads, rankings on this test are determined mainly by sequential write performance and whether the controller can process the mixed workload effectively. The MX300 is slower than the SanDisk X400 but faster than the other planar TLC drives.
With another third place ranking for power usage, the MX300 beats all the planar TLC drives on efficiency but is unremarkable by MLC standards.
 |
|||||||||
The MX300 bottoms out around 126MB/s which would be respectable for a planar TLC drive, but the MX200 never drops below 200MB/s.
ATTO
ATTO's Disk Benchmark is a quick and easy freeware tool to measure drive performance across various transfer sizes.
 |
|||||||||
The ATTO performance plot shows that small reads are slow but performance is reasonably stable for large transfers, albeit a little slower than the MX200.
AS-SSD
AS-SSD is another quick and free benchmark tool. It uses incompressible data for all of its tests, making it an easy way to keep an eye on which drives are relying on transparent data compression. The short duration of the test makes it a decent indicator of peak drive performance.
AS-SSD shows the MX300 read speed as suffering slightly, but the (SLC cached) write speed as perfectly normal.
Idle Power Consumption
Since the ATSB tests based on real-world usage cut idle times short to 25ms, their power consumption scores paint an inaccurate picture of the relative suitability of drives for mobile use. During real-world client use, a solid state drive will spend far more time idle than actively processing commands. Our testbed doesn't support the deepest DevSlp power saving mode that SATA drives can implement, but we can measure the power usage in the intermediate slumber state where both the host and device ends of the SATA link enter a low-power state and the drive is free to engage its internal power savings measures.
We also report the drive's idle power consumption while the SATA link is active and not in any power saving state. Drives are required to be able to wake from the slumber state in under 10 milliseconds, but that still leaves plenty of room for them to add latency to a burst of I/O. Because of this, many desktops default to either not using SATA Aggressive Link Power Management (ALPM) at all or to only enable it partially without making use of the device-initiated power management (DIPM) capability. Additionally, SATA Hot-Swap is incompatible with the use of DIPM, so our SSD testbed usually has DIPM turned off during performance testing.
Marvell's latest controller allows for great slumber power savings, but Micron needs to work on improving active idle power draw.
Conclusions
It appears that the Crucial MX300 will be priced as a mid-range SATA drive or slightly below that. On a 'Price per GB' metric alone the 3D TLC NAND isn't starting any revolutions, which means that again the association between TLC NAND and lower performance still rings true. Despite this, the performance is clearly higher and above the current glut of planar TLC drives that are competing in a race to the bottom.
One of the issues that Crucial will face is that despite being plus-one generation above the MX200, The MX300 is slightly slower and only by a small amount. It frequently straddles the dividing line between MLC performance and planar TLC performance. One issue on performance will be that it is also surpassed on several benchmarks by SanDisk's X400, one of the fastest planar TLC drives and a drive that will likely beat the MX300 on price. The 850 EVO level of performance is simply out of reach; Micron's 3D TLC drive is slower than Samsung's 3D TLC drive, and so will have to compete on price.
One thing to point out is that through our testing, we see that the MX300 has an acute weakness in its random read latency. At all but the highest queue depths it is half as fast as the top MLC drives that are only moderately faster than the MX200. Since this pattern holds at the lowest queue depths where parallelism and caching don't apply, there's a danger that this means Micron's 3D TLC is inherently quite slow to read from. This is most likely a carry on from when Micron implemented SLC write caching for the Crucial MX200:
With the MX200, the short-term performance boost of the SLC wasn't always worth the eventual cost of moving data from SLC to MLC. The SLC caching on the MX300 seems to greatly lower the power requirements of handling a small volume of writes, which may be a reason to use it even with the 3D MLC, especially if performance is sufficient to handle flushing a full write cache under load without a drastic slowdown. However, when the MX300's SLC write caching and spare area are exhausted, it slows down to the level of budget planar TLC drives. This is a drive that should not be filled to the brim and should not be subjected to enterprise workloads with heavy sustained writes.
Crucial SSDs: MX, BX and The Future
The future of Micron's Crucial SSDs is uncertain. When the MX100 launched, it was a hit by offering mainstream performance at great prices for the time. The BX100 showed up at even lower prices and with performance that was pretty close to the MX100. The MX200 added just enough performance to somewhat justify keeping two models around. Later the BX200 adopted TLC and sacrificed a lot of performance to cut costs, but failed to compete against the wave of budget drives based on Toshiba and Hynix TLC. Now that the MX line has also adopted TLC, it seems likely that the BX line will be retired along with planar NAND.
The interesting question is whether Crucial will introduce a higher end 3D MLC drive. We learned at Computex that a 3D MLC NVMe SSD will be released under Micron's Ballistix brand, a now separate sub-brand of Micron and different to Crucial. Thus the only potential for a new MLC drive from Crucial would be a high-end SATA drive. Many companies have been wondering whether it is worth trying to compete directly against the 850 Pro that has reigned for two years as the fastest SATA SSD and is very nearly the fastest possible SATA SSD (barring the use of pure SLC or 3D XPoint, neither of which will happen). Crucial might have the opportunity with Micron's 3D MLC to introduce a drive that is just as fast as the 850 Pro while being more power efficient, but it would still be tough to dethrone the 850 Pro unless Micron could also clearly undercut Samsung on price. Alternatively, we may see MLC become something that is mostly used on PCIe SSDs while the SATA SSD market is overrun by TLC.
| SSD Price Comparison (Sorted by Price/GB of Highest Capacity Drive) |
|||
| Drive | 960GB 1TB |
750GB | 480GB 512GB |
| OCZ Trion 150 | $199.99 (20.8¢/GB) | $109.99 (22.9¢/GB) | |
| SanDisk X400 | $229.49 (22.4¢/GB) | $124.49 (24.3¢/GB) | |
| SanDisk Ultra II | $219.56 (22.9¢/GB) | $127.31 (26.5¢/GB) | |
| Mushkin Reactor | $249.99 (24.4¢/GB) | $149.99 (29.3¢/GB) | |
| Crucial MX300 | $199.99 (26.7¢/GB) | ||
| Crucial MX200 | $269.94 (27.0¢/GB) | $139.00 (27.8¢/GB) | |
| PNY CS2211 | $289.99 (30.2¢/GB) | $129.99 (27.1¢/GB) | |
| Samsung 850 EVO | $306.76 (30.7¢/GB) | $153.95 (26.7¢/GB) | |
| SanDisk Extreme Pro | $338.08 (35.2¢/GB) | $189.99 (39.6¢/GB) | |
Final Words
To put this into perspective, under ordinary consumer and end-user/home workloads, the MX300 performs at its peak near the top of the TLC charts. On most tests we found the MX300 to be remarkably power efficient. Other things being equal, TLC is typically slower and more power hungry than MLC, but the MX300 is more power efficient on most benchmarks than most MLC drives. Having this level of efficiency is extremely promising for Micron's 3D MLC and an accomplishment worth some kudos.
Facebook
Twitter
RSS
