Crucial P1 1TB
Crucial's new SSD is a quad-level conundrum
INCREASED STORAGE CAPACITY comes at a price. That’s not exactly a revolutionary idea, but it is more complicated than it sounds. That’s because the costs involved aren’t just dollars and cents; they can also be measured in performance and endurance terms.
Those latter concerns are front and center with Crucial’s P1 drive, its latest M.2 NVMe drive, reviewed here in 1TB configuration. It’s one of the first on the market to use Intel and Micron’s new QLC or quad-level cell NAND flash memory. That’s one bit up on the previous industry best that is TLC or triple-level cell memory. In simple terms, the appeal is obvious. The extra bit makes for 33 percent more storage capacity per cell. However, the downsides can be brutal.
In layman’s terms, upping the per-cell capacity of NAND flash makes it more fragile. A single-level cell, for instance, can handle something like 100,000 program-erase cycles before conking out. That endurance progressively degrades as you add cell capacity. By the time you get to QLC memory, you’re looking at just 1,000 cycles before a cell is toast. In practice, wear-leveling algorithms ensure that won’t actually be a major concern for most home users, but it does illustrate the cost of increasing capacity via cell density.
As for performance, we detailed in last issue’s feature on the state of SSD tech that reading and writing to multi-level NAND cells is increasingly laborious as per-cell density increases. It requires distinguishing between and applying 16 possible voltage levels in each cell. That limitation is more apparent in practice than the impact on endurance. PERFORMANCE ART A quick thumb-in-the-air evaluation of the new P1 reveals that sequential data transfer performance eventually drops off to about 60MB/s if you throw enough data at the drive. What’s happening is obvious: The P1 relies on an allocation of its memory operating in SLC or singlelevel cell mode to deliver its claimed peak performance of 2,000MB/s reads and 1,700MB/s writes. Once that is exhausted, the true underlying performance of those QLC cells is exposed. And it ain’t pretty.
You won’t always see that limitation in synthetic tests. Although the P1 doesn’t get on that well with CrystalDiskMark’s sequential throughput test, it still cranks out pretty quick numbers of 1.15GB/s reads and 1.7GB/s writes, for instance. That said, the P1’s performance in Anvil’s QD1 random access benchmarks is disappointing; 32MB/s for reads and 126MB/s for writes are not disastrous, but remain a little off the pace for a modern M.2 NVMe drive.
Our more real-world 30GB internal file copy test is also a mixed bag. The P1 completes it in 60 seconds, which means most of the data transfer is managed before the SLC cache runs out. But a high-quality drive that doesn’t rely on SLC caching, such as Intel’s Optane 905p, can burn through the data transfer in about half that time.
In the end, the P1 is posed with two critical questions. First, how often are you likely to bump into the P1’s built-in limitations? Unlike the ADATA SU630 opposite, which uses the same QLC NAND flash memory, we’d say not that often. However, the problem the P1 shares with the ADATA drive is its answer to the second question. Does it provide an advantage in terms of cost per GB?
At this price point, you have a number of TLC options, and this QLC drive doesn’t offer a clear value advantage on capacity alone. That may be because, reportedly, Intel and Micron are having problems manufacturing QLC NAND at yields healthy enough to make drives much cheaper than the TLC alternatives. Whatever the truth about that, the fact is that QLC technology isn’t cheap enough now to offset the impact on performance and endurance.