Newer versions are not always better. Recently, SSD manufacturers have begun to accept speed and reliability in order to put more space into their drives. Protocols like NVMe and PCIe are getting faster and faster, but some SSDs are running backwards.
QLC flash is the problem
Here's the problem. Manufacturing SSDs is expensive, and few people want to pay $ 200 for a 51
SSD manufacturers increase storage capacity while keeping costs low – but this is bad for performance and endurance. Big SSDs are getting cheaper, but there's a trade-off for every step in SSD technology. Currently, Quad Level Cell (QLC) SSDs are on the rise, capable of storing 4 bits of information per memory cell. QLC has not completely replaced standard SSDs, but some drives that use it have arrived in the market and they have problems.
In particular, SSD manufacturers need to find a way to fit more space in the same size NAND flash chips (the actual data-storing part of the SSD). Traditionally, this has been done with a process node downsizing, which has made the transistors in the flash smaller. But if Moore's law slows down, you have to be more creative.
The ingenious solution is the multistage NAND flash. NAND-Flash is able to store a certain voltage level in a cell over a longer period of time. Traditional NAND Flash stores two levels: On and Off. This is called SLC flash and is very fast. However, because NAND essentially stores an analog voltage, you can represent several bits with slightly different voltage levels:
The problem shown here is that it expands exponentially . SLC flash only needs voltage or its absence. MLC flash requires four voltage levels. TLC needs eight. And last year, the QLC flash has entered the market with 16 different voltage levels.
This leads to a variety of problems. The more voltage levels added, the harder it is to distinguish the bits from each other. This makes QLC Flash 25% more dense than TLC, but much slower. The reading speed is not so much affected, but the writing speed takes a jump in purchasing. Most SSDs (using the newer NVMe protocol) are about 1500 MB / s for long-term reading and writing (that is, loading or copying large files). However, QLC Flash only manages between 80 and 160 MB / s for persistent writes, which is worse than a decent hard drive.
QLC SSDs break much faster
All SSDs generally have an unfavorable write time compared to hard drives. Each time you write to an SSD cell, it wears off slowly. Deleting a cell is said to free it of electrons, but some are always stuck, causing a "0" cell to be closer to "1" over time. This is compensated by the controller by applying a more positive voltage over time. This is fine if you have a lot of tension room available. QLC not.
SLC has an average write time of 100,000 program / erase cycles (write operations). MLC has between 35,000 and 10,000. DC has around 5,000. But QLC only has 1000. This makes QLC for drives with frequent access, such as: For example, your boot drive, which is described very frequently, is inappropriate.
Conclusion: Do not buy a QLC drive that can be used with the system drive of your operating system. They are far too unreliable to be sure that it will not deteriorate in a few years. We recommend that you use a large QLC drive to replace a rotating hard drive and use a fast SLC, MLC, or TLC drive as the primary operating system drive. This can be a problem with laptops that do not have the option, but QLC is still very new and has not yet entered laptops.
Efficient caching hides these problems.
At this point, you may be wondering why QLC is a thing at all if it's objectively slower and breaks much faster than the other types of lightning. Of course, you can not market a downgrade, but SDD vendors have found a way to hide the problem: caching.
QLC SSDs allocate part of the drive to a cache. This cache ignores the fact that it should be QLC, and works like SLC Flash instead. The cache is 75% smaller than the actual used space but much faster.
Data from the cache can be written at the same speed as other high-end SSDs and slowly emptied by the controller and sorted into the QLC cells. However, if this cache is full, the controller must write directly to the slow QLC cells, resulting in a significant performance degradation on long writes.
Take a look at this benchmark from Tom's hardware review of the Crucial P1 500GB. A consumer QLC SSD that shows this problem quite clearly:
The red line for the Crucial P1 works with solid NVMe speeds, although somewhat compared to some slowly the high-end deals. However, after about 75 GB of write access, the cache is full and you can see the actual speed of the QLC flash. The transfer speed drops to about 80 MB / s, which is slower than most hard drives for permanent writes.
The ADATA XPG SX8200, a TLC drive, has the same characteristics, except that the raw TLC flash is even faster after it has been dropped. Most other drives also use this caching method because they speed up fast, small writes to the drive (which is most common). However, long-term writing is what you'll notice most – you will not notice if a small file copy takes .15 seconds instead of .22 seconds, but you'll notice if it takes another 10 minutes.
You could easily write this out as an edge case scenario, but this cache will not last forever 75 GB. When you fill the hard disk, the cache gets smaller. According to Anandtech's tests, for the Intel SSD 660p series, the cache for the 512 GB model is reduced to only 6 GB when the drive is mostly full, even though there is still 128 GB of space remaining Drive Fills "width =" 650 " height = "606" data-credittext = "Anandtech" data-crediturl = "https://www.anandtech.com/show/13078/the-intel-ssd-660p-ssd-review-qlc-nand-arrives" src = "/pagespeed_static / 1.JiBnMqyl6S.gif" onload = "pagespeed.lazyLoadImages.loadIfVisibleAndMaybeBeacon (this);" onerror = "this.onerror = null; pagespeed.lazyLoadImages.loadIfVisibleAndMaybeBeacon (this);" />
Granted, you'll likely get lost in this example by downloading Gesc speed, but in the case of updates (which must effectively download the existing files and then replace them) the problem would be much more obvious. They would stop downloading and then have to wait forever for it to be installed.
Should you avoid QLC?
You should definitely avoid 512GB QLC drives (and less as soon as production becomes cheaper). because they do not make much sense. They fill them up much faster and the cache gets smaller when it's full, which slows it down considerably. Besides, they are currently not much cheaper than the alternatives.
Despite its shortcomings, QLC Flash is not a big problem for higher capacity drives. The 2 TB model of the 660p has at least 24 GB of cache when it is full. It's still QLC flash, but an acceptable compromise for a cheap 2 TB SSD that works very fast most of the time.
Because of their gigantic capacity, QLC-based SSDs can serve as a reasonable replacement for a spinning disk, provided you create regular backups in the event that they fail. It is ideal for things that you rarely access but that you want to do very quickly. With a large enough SLC cache, most persistent writes are relatively fast until you get the drive full.
Reliability issues should not be used as a boot drive or for things that are written very often.
There is still much to be done in other aspects of manufacturing – better controllers that can address more flash chips, cheaper flash chips than process nodes, and perhaps other technologies as a whole. QLC Flash will not become standard in the foreseeable future. Currently, this is just another option. Just make sure you check the specifications when purchasing an SSD and pay attention to the type of flash used to make it.