The Crucial P1 1TB SSD Review: The Other Consumer QLC SSDby Billy Tallis on November 8, 2018 9:00 AM EST
The Crucial P1 SSD isn't quite the first at anything, but it is still a very novel product. It is the second consumer SSD on the market to use four bit per cell (QLC) NAND flash memory, after the Intel SSD 660p. It's the second QLC SSD from Micron, after their 5210 ION enterprise SATA SSD that started shipping to select partners in May (and is now starting to be more widely available).
More importantly however, it is the first consumer NVMe SSD that Micron has actually shipped. For all of their storied history in the SSD industry, Micron is pretty much the last SSD maker to enter the consumer NVMe market; and not for a lack of trying. The company's first attempt at an NVMe drive, the 2016 Ballistix TX3, was ready to hit the market but was canceled when it became clear that it would not have been competitive. So this drive is a very important one for the company, despite the fact that its use of an NVMe/PCIe interface is hardly the most interesting aspect of the Crucial P1.
Under the hood, the Crucial P1 starts from the same basic ingredients as the Intel 660p: Intel/Micron 64-layer 3D QLC NAND and the Silicon Motion SM2263 controller. Micron has added their own firmware customization atop Silicon Motion's work, and the design of the Crucial P1 differs from the Intel 660p in several aspects — so this is not a case of two brands selling the exact same reference design SSD.
Where the Intel 660p includes just 256MB of DRAM regardless of drive capacity, the Crucial P1 includes the same 1GB DRAM per 1TB NAND ratio that is used by most mainstream SSDs. This extra DRAM on the Crucial P1 should enable marginal improvements on benchmarks and workloads that touch large amounts of data, and probably allows a slight simplification to the drive's firmware. The Crucial P1 also has slightly lower usable capacities, eg. 500GB instead of 512GB, so there's a bit more spare area for the controller to work with. And whereas Intel's 2TB 660p is still a single-sided M.2 module, the upcoming 2TB Crucial P1 will have NAND and DRAM on both sides.
The rest of the architecture of the Crucial P1 follows the same general strategies as the Intel 660p. The SM2263 controller is the smaller four-channel design from Silicon Motion's current generation, though it is a step above the DRAMless SM2263XT variant. The host interface is a PCIe 3.0 x4 link, but the Crucial P1 barely needs more than the PCIe 3.0 x2 link used by some competing entry-level NVMe controllers: peak sequential transfers for the P1 are only about 2GB/s.
|Crucial P1 SSD Specifications|
|Capacity||500 GB||1 TB||2 TB|
|Form Factor||Single-sided M.2 2280||Double-sided M.2 2280|
|Interface||NVMe 1.3 PCIe 3.0 x4|
|Controller||Silicon Motion SM2263|
|NAND Flash||Micron 64L 3D QLC NAND|
|DRAM||512MB DDR3||1GB DDR3||2GB DDR4|
|Sequential Read||1900 MB/s||2000 MB/s||2000 MB/s|
|Sequential Write||950 MB/s||1700 MB/s||1750 MB/s|
|Random Read||90k IOPS||170k IOPS||250k IOPS|
|Random Write||220k IOPS||240k IOPS||250k IOPS|
|SLC Write Cache (approximate)||5GB min
|Idle||2mW (PS4), 80mW (PS3)|
|Write Endurance||100 TB
|MSRP||$109.99 (22¢/GB)||$219.99 (22¢/GB)||TBA|
The use of QLC NAND means that the Crucial P1 is highly reliant on its SLC write cache to enable performance that can exceed what SATA SSDs provide. This is because the drive's QLC NAND isn't all that high performing on its own; it's dense, but it takes longer to program a block than MLC or TLC NAND. TLC for that matter is cache-sensitive for similar reasons, but QLC in turn has cranked up the importance of cache sizes and caching algorithms another notch, as the performance delta between the cache and the actual storage has increased.
For Crucial's P1 the SLC cache is variable in size, and on a nearly-empty drive the cache will be substantially larger than what is usually found on TLC SSDs. Consumer drives that use TLC NAND often try to limit the maximum size of their SLC caches in order to reduce the amount of background work necessary in the event that the SLC write cache should overflow. The Crucial P1 is designed to do as much as possible to avoid falling off that performance cliff, rather than attempt to mitigate the effects when it does happen.
When the SLC cache fills up, writes to the P1 get very slow. The P1 doesn't bypass the cache when it is full, so everything written to the drive is written to SLC first before being folded into QLC blocks. (This helps the P1 offer similar partial power loss protection to the Crucial MX series of SATA SSDs.) The P1 also tends to keep data in SLC so it can serve as a read cache instead of aggressively folding data into QLC blocks during idle time.
All told, the caching strategy of the Crucial P1 maximizes performance and endurance for typical lightweight consumer/client storage workloads, but at the cost of performance on storage-intensive workloads. The P1 is definitely not the SSD to use in a workstation that regularly reads and writes datasets of many gigabytes, but it should be fine for more common desktop usage that is fairly read-heavy and only does multiple GB of writes on rare occasions such as when installing large software packages. This is in some sense just an amplification of the trends we saw as the SSD market moved from MLC to TLC NAND, but we do now have high-end TLC drives that can maintain high write speeds even after their SLC caches have filled. This is not true of the current two consumer QLC SSDs, and will probably always be a significant weakness of QLC SSDs.
The other major tradeoff to the Crucial P1's use of QLC NAND is the lower write endurance compared to TLC SSDs. The P1 is rated for about 0.1 drive writes per day under a five-year warranty, while most consumer SSDs are rated for 0.3 up to about 1.0 DWPD for either three or five years. The P1 somewhat mitigates this by only offering large capacities of 500GB and up, so the total write endurance starts at a minimum of 100TB. This is adequate provided that most of the drive's capacity is used for static data. If the P1 has to hold hundreds of GB of data that changes as frequently as a web browser's cache and history, then 0.1 DWPD won't be enough. But in a more normal scenario where most of the data is media like movies and video games, then there's no problem.
The early projections for QLC NAND write endurance were in the ballpark of a few hundred program/erase cycles at most, which would have required the QLC SSDs to be treated very carefully. The QLC NAND that Micron is now producing in volume can last for a similar number of P/E cycles as early TLC NAND, which is how the Crucial P1 can be usable for general-purpose consumer storage duties. Even after accounting for the write amplification caused by SLC caching and a realistic proportion of writes being random, the P1 is still rated for the equivalent of 200 full drive writes on the host side. (If those drive writes were entirely large-block sequential writes such as from re-imaging an entire drive, then the P1 should last much longer, but Micron doesn't want to complicate the endurance specs for their consumer drives that much.)
In spite of the tradeoffs of lower performance and endurance, QLC drives like the Crucial P1 are worth a look because of their potential to also have significantly lower prices. At around 22¢/GB currently, the P1 isn't setting any records yet. Several industry reports have indicated that yields of Intel/Micron QLC are still poor, so the production costs of SSD-quality QLC aren't meaningfully lower than TLC yet. The other major NAND manufacturers are being less aggressive about bringing QLC to market, but once they have introduced their competitors we will probably see QLC products offering a more significant discount over TLC. NAND prices in general are also in decline, with some higher-volume TLC products leading the way ahead of even the QLC drives.
The primary competition for the Crucial P1 is its close relative Intel 660p, as well as other entry-level NVMe SSDs. There is a low-end NVMe market segment with numerous options, but they are all struggling under the pressure from more competitively priced high-end NVMe SSDs. Products like the DRAMless Toshiba RC100 and HP EX900 haven't been able to get any traction when the HP EX920 is hitting the same prices. Thus, it is also fair to compare the Crucial P1 against such faster NVMe drives. The P1 is a modest step up in price over mainstream SATA SSDs, so this review includes benchmark results from the Crucial MX500.
|AnandTech 2018 Consumer SSD Testbed|
|CPU||Intel Xeon E3 1240 v5|
|Motherboard||ASRock Fatal1ty E3V5 Performance Gaming/OC|
|Memory||4x 8GB G.SKILL Ripjaws DDR4-2400 CL15|
|Graphics||AMD Radeon HD 5450, 1920x1200@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|
- Thanks to Intel for the Xeon E3 1240 v5 CPU
- Thanks to ASRock for the E3V5 Performance Gaming/OC
- Thanks to G.SKILL for the Ripjaws DDR4-2400 RAM
- Thanks to Corsair for the RM750 power supply, Carbide 200R case, and Hydro H60 CPU cooler
- Thanks to Quarch for the XLC Programmable Power Module and accessories
- Thanks to StarTech for providing a RK2236BKF 22U rack cabinet.
Post Your CommentPlease log in or sign up to comment.
View All Comments
DanNeely - Thursday, November 8, 2018 - linkWhen DDR2 went mainstream they stopped making DDR1 dimms. The dimms you could still find for sale a few years later were old ones where you were paying not just the original cost of making them, but the cost of keeping them in a warehouse for several years before you bought them. Individual ram chips continued to be made for a while longer on legacy processes for embedded use but because the same old mature processes were still being used there was no scope for newer tech allowing cost cutting, and lower volumes meant loss of scale savings meaning that the embedded world also had to pay more until they upgraded to new standards.
Oxford Guy - Thursday, November 8, 2018 - linkThe point was:
"QLC may lead to higher TLC prices, if TLC volume goes down and/or gets positioned as a more premium product as manufacturers try to sell us QLC."
Stopping production leads to a volume drop, eh?
romrunning - Thursday, November 8, 2018 - link"There is a low-end NVMe market segment with numerous options, but they are all struggling under the pressure from more competitively priced high-end NVMe SSDs."
I really wish all NVMe drives kept a higher base performance level. QLC should have died on the vine. I get the technical advances, but I prefer tech advances increase performance, not ones that are worse than their predecessor. The price savings, when it's actually there, isn't worth the trade-offs.
Flunk - Thursday, November 8, 2018 - linkIn a year or two there are going to be QLC drives faster than today's TLC drives. it just takes time to develop a new technology.
Oxford Guy - Thursday, November 8, 2018 - linkFaster to decay, certainly.
As I understand it, it's impossible, due to physics, to make QLC faster than TLC, just as it's impossible to make TLC faster than MLC. Just as it's impossible to make MLC faster than SLC.
Workarounds to mask the deficiencies aren't the same thing. The only benefit to going beyond SLC is density, as I understand it.
Billy Tallis - Thursday, November 8, 2018 - linkOther things being equal, MLC is faster than TLC and so on. But NAND flash memory has been evolving in ways other than changing the number of bits stored per cell. Micron's 64L TLC is faster than their 32L MLC, not just denser and cheaper. I don't think their 96L or 128L QLC will end up being faster than 64L TLC, but I do think it will be faster than their 32L or 16nm planar TLC. (There are some ways in which increased layer count can hurt performance, but in general those effects have been offset by other performance increases.)
Oxford Guy - Thursday, November 8, 2018 - link"Other things being equal, MLC is faster than TLC and so on"
So, other than density, there is no benefit to going beyond SLC, correct?
Billy Tallis - Thursday, November 8, 2018 - linkPretty much. If you can afford to pay for SLC and a controller with enough channels and chip enable lines, then you could have a very nice SSD for a very unreasonable price. When you're constrained to a SATA interface there's no reason not to store at least three bits per cell, and even for enterprise NVMe SSDs there are only a few workloads where the higher performance of SLC is cost-effective.
Great_Scott - Monday, November 12, 2018 - linkThey should drop the SLC emulation and just sell the drive as an SLC drive. Sure, there may be some performance left on the table due to the limits of the NVME interface, but the longevity would be hugely attractive to some users.
They'd make more money too, since they could better justify higher costs that way. In fact, with modern Flash they might be able to get much the same benefit from MLC organization and have roughly half the drive space instead of 25%.
Lolimaster - Friday, November 9, 2018 - linkDo not mix better algorithms of the simulated SLC cache and dram with actual "performance", start crushing their simulated cache and the TLC goes to trash.