Micron's consumer-oriented Crucial brand is finally entering the world of NVMe SSDs with the new Crucial P1 M.2 SSD. The P1 is an entry-level NVMe drive using four bit per cell (QLC) NAND flash memory and the Silicon Motion SM2263 controller. This is the same basic formula as used in the Intel 660p, the only other consumer QLC drive on the market so far. Micron has brought their own firmware customizations, so while the performance characteristics are similar to the Intel 660p they are definitely not the same drive. The Crucial P1 has slightly lower usable capacities than the Intel 660p, which translates into slightly more spare area available for garbage collection and SLC caching. Unlike the Intel 660p, the Crucial P1 uses the same 1GB DRAM per 1TB NAND ratio as most MLC and TLC SSDs.

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
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
50GB max
12GB min
100GB max
24GB min
200GB max
Power Max 8W
Idle 2mW (PS4), 80mW (PS3)
Warranty 5 years
Write Endurance 100 TB
0.1 DWPD
200 TB
0.1 DWPD
400 TB
0.1 DWPD
MSRP $109.99 (22¢/GB) $219.99 (22¢/GB) TBA

With top sequential speeds of only 2GB/s, the Crucial P1 doesn't really need all four PCIe lanes, but Silicon Motion's entry-level SM2263 controller still has four instead of the two that some other low-end NVMe controllers use. Given the use of QLC NAND, the P1's SLC cache is far more important than it is on drives with TLC NAND. Micron has taken a similar approach to what Intel did with the 660p by making the SLC cache not just a write buffer but a full-time dynamically sized read and write cache. All data written to the Crucial P1 hits the SLC cache first, and is compacted into QLC blocks only when the drive's free space starts running low. This means that a mostly-empty drive will be using tens or hundreds of GB of SLC, but as it fills up the cache will shrink down to just 5-24GB depending on the model. All of Crucial's official performance specifications are for the SLC cache.

As with the Crucial MX series of SATA SSDs, the Crucial P1 features a greater degree of power loss protection than typical consumer SSDs, though not the fully capacitor-backed protection that most enterprise SSDs feature. With the MX500, Crucial had already substantially reduced the number of capacitors required for their partial power loss protection thanks in part to a reduction in write power requirements for their 64-layer 3D NAND. The P1 gains additional data security from its SLC-first write policy, which eliminates the partially-programmed page risk. However, there is still a tiny bit of used data buffered in volatile RAM, on the order of a few MB at the most.

The rated write endurance of 0.1 drive writes per day is low even for an entry-level consumer SSD, but given the large drive capacities it is adequate. The P1 is definitely not intended to be the workhorse of an enthusiast system with a write-heavy workload, but for more typical read-oriented workloads it offers better performance than SATA SSDs.

The initial MSRPs for the Crucial P1 are unimpressive: it's substantially more expensive than the Intel 660p, and about matches some of the most affordable high-end NVMe drives like the HP EX920 and ADATA SX8200. If Micron can catch up to Intel's pricing and compete for the lowest $/GB among all NVMe SSDs then the Crucial P1 has a shot at success.

The 2TB model will be launching slightly later due to using DDR4 DRAM instead of the DDR3 used by the 500GB and 1TB models. We are testing the 1TB Crucial P1, with some results already in our Bench database. Look for our full review next week.



View All Comments

  • chrcoluk - Wednesday, September 25, 2019 - link

    Bang on, I have tried to explain this to others also, I can see what they doing from a mile off. the aim of QLC is to increase profit margins, I expect the manufacturing cost savings are only about 20% at best. There will be no miracle mega cheap drives. They will slightly undercut TLC to get the marketshare, then when QLC is accepted as the norm, TLC and MLC will go up in price as defacto premium products. e.g. I expect the samsung pro drive manufacturing costs are nowhere near double of EVO drives, yet they double the price, they already successfully pulled of the same trick with TLC. Reply
  • chrcoluk - Wednesday, September 25, 2019 - link

    Also to add my 850 pro (A pricey 3D MLC NAND drive) died a month or so back with just 20TB written, the rated endurance I would take with a grain of salt. The manufacturers will know those ratings wont be tested for a while. Reply
  • chrcoluk - Wednesday, September 25, 2019 - link

    QLC is only 33% cheaper to produce than TLC tho and thats assuming it has as good yields. Any bigger price differentials will be for other reasons (such as profit margins). So e.g. TLC is 50% cheaper than MLC to produce yet samsung TLC drives are almost half price of their MLC drives, so that doubling price has some artificial inflation for the premium'ness of the PRO models.

    The 860 QVO actually was same price as the EVO on launch LOL.
  • nathanddrews - Tuesday, October 16, 2018 - link

    If the price is right, this would be perfect for a Steam/GOG drive. Reply
  • Kamgusta - Friday, October 19, 2018 - link

    For STEAM/GOG I am using a 50$ 1TB 2,5" SATA mechanical drive. I would not replace it with this P.1 by any means: 5x price increase, 3x power consumption increase, 0% performance increase. Reply
  • PaoDeTech - Tuesday, October 16, 2018 - link

    If the price is right... Today on Amazon: MX500 1TB SATA M.2 is $160 (basically the SATA M.2 consumer value leader), this one (P1 1TB 3D NAND NVMe PCIe M.2) is $220 (the PCIe M.2 consumer value leader?). Samsung 970 EVO 1TB is $277. Obviously I would't use it for OLTP. Reply
  • hojnikb - Tuesday, October 16, 2018 - link

    Thats just not true. There was never a consumer NAND rated for million p/e. The most we got was a 100k SLC around the 50nm node. Obviously you could find some enterprise solutions for more, but that was pretty niche. When SSD really took off, we were already at MLC with around 3-5k p/e which was perfectly usable. Even todays NAND with ~1k p/e or so is more than enough for typical client workloads. Reply
  • PeachNCream - Tuesday, October 16, 2018 - link

    You are correct that there wasn't a consumer product with that sort of endurance. They were the domain of flash found on devices like thin clients (which were fun to modify into desktop PCs). Regardless of the consumer availability, the fact that NAND hasn't got legs to go further and the writing has been on the wall for as long as we've had MLC makes it a confusing fact that we currently lack a sufficiently durable, cost-effective replacement. We've had years to solve the problem so even though it is a non-trivial task, wringing NAND with QLC as state-of-the-art is a poor example of how far we've fallen. Reply
  • npz - Tuesday, October 16, 2018 - link

    Here's a typical LIGHT client workload from just ONE commonly used application:
    "In my case, SSDLife notified me that 12GB was written to the SSD in one day"
    Every single update of any file, write then delete, Windows updates, application updates, caching of ANY kind for any program, constitutes p/e cycles. For these QLC drives, that pitiful 0.1 DWPD rating = 50GB for the 500GB model, can be exceeded on a client.

    And you have to keep in mind that this doesn't use dedicated SLC. It uses the same NAND cells of the QLC except dynamically in SLC mode (hence why it shrinks as usage goes up) so it's SLC cache usage also constitutes p/e wear on the same drive cells.
  • Destoya - Tuesday, October 16, 2018 - link

    I have had my OS install and most programs on my 120 GB MLC drive since 2012. The drive has over 3.5 years of power-on time so I'm not a light computer user by any means. I also use it as a cache for various games/programs but not any super heavy database-intensive workloads.

    Right now it's reporting 96% life with ~275 P/E cycles (33 TB written), so while I'm at ~0.125 DWPD over the lifetime of the SSD that's really just because of the low initial capacity. My total writes wouldn't change that significantly if it were a 240 GB model for example. My point is that a moderately sized QLC drive is completely adequate for the vast majority of computer users and will needless to say far outperform and outlast a HDD.

Log in

Don't have an account? Sign up now