I imagine that Windows and browser log files being incessantly written to an SSD will cause noticeable wear over time.
SSD Life
- Thread starter ZeeOSix
- Start date
I have that same concern. There's not much you can do except refuse all but essential cookies and set history to delete every thing on browser close down.
Less than you’d think. There should be wear leveling to distribute program/erase cycles, and newer SSDs employ memory and SLC caches that would mitigate the writing to the main flash. There’s been consternation that newer quad level NAND flash used has maybe 500-1000 endurance cycles, but that may be adequate with proper design.
I didn’t really think about it previously, but they’re likely some way to configure the same NAND flash chips to operate under different modes and program different portions to operate differently. There isn’t physically anything different about a single level NAND flash setup than tri or quad level.
Temp plays a role IF YOU KEEP THE POWER OFF.
When it is running the algorithm takes temperature into accounts on how often you need to refresh the content. You don't get these kind of refresh in the background if you keep it in a hot car powered off for a year. Most aim for 40C operating temperature. I know there are data centers who want to buy 65C operating temperature stuff with life as a trade off, and they do lose a lot. However most human would not like to work in 65C and most often AC is turned on before reaching 32C, so a correctly designed desktop / laptop should have operating temperature of the SSD at around 40C.
Usually, the more you write / erase the less accurate they will eventually become. You can be way off in SLC mode and the controller can still correct the tiny amount of errors before failing, but you can only be 1/2 as far off (in voltage) in MLC mode, 1/4 as far off in TLC mode, and 1/8 as far off in QLC mode. That determines how much life is left in a cell and how long would it last.
It is an arm race between the controller and NAND.
The NAND got worse as the cell got smaller and more bits are stored per cell (i.e. a capacitor, but your voltage has to be 2x as accurate in MLC vs SLC, 4x as accurate in TLC vs SLC, and 8x as accurate in QLC vs SLC). The controller now pack a much stronger error correction engine than before and therefore able to tolerate these worse and worse NAND.
The manufacturers always sell you something based on how much they can get away with. They won't leave additional life on the table for free if they can. Instead of 15% reserve capacity for wear leveling they may give you 7% if you don't need as much life, for example.
Last edited:
I was thinking more in the line that a cell could be used for anything. The first time I’d heard of SLC caching was in 2018 when I got my SanDisk Ultra 3D SATA SSD. I was thinking it had a separate SLC chip, but I figure they probably have some way to configure individual NAND arrays to operate as SLC/MLC/TLC.
Yeah the chips are all the same, they can take SLC, MLC, TLC, QLC commands and they "become" SLC, MLC, TLC, QLC etc. How much they allocate to each kind will be based on how fast and how often each operation is done, and benchmark, not chip, but each block on the chip.I was thinking more in the line that a cell could be used for anything. The first time I’d heard of SLC caching was in 2018 when I got my SanDisk Ultra 3D SATA SSD. I was thinking it had a separate SLC chip, but I figure they probably have some way to configure individual NAND arrays to operate as SLC/MLC/TLC.
Typically you tell the chip what voltages you use to separate between 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, then let the chip decide what it is when you "read" and "write". Defining them as each type of cell gives you different advantage and disadvantage, like data retention life vs speed and density.
I look at the major brands (many being the actual NAND flash makers like Micron, SK Hynix, WD/SanDisk, Samsung)and they seem to have coalesced around similar price points. Then I see lesser known brands like Silicon Power or ADATA and I’m wondering why they’re so much cheaper. And finally the no name brands, often where they’re emulating the label of a major brand but where a 1 TB SATA SSD is selling for $30. What corners are they cutting?
They often cut the marketing cost, and OEM may not buy them because they cannot afford warranty claim of the entire machine / send out repair tech vs just mailing you a new SSD.
Some of them use very similar components and are good, while others may use cheap stuff that may cause problems once in a while, it really depends on the brand. A lot of them are made by the same OEM (LiteOn, Kingston, etc) using the same components (Kioxia, Micron, SK Hynix nand, Marvell or Phison controllers, etc).
I would buy from a company that I know can handle the warranty claim (i.e. 3 years, 5 years, etc) and benchmark on review site is good. I think the biggest performance jump from HDD to SSD is big, but most office task won't tell much of a difference between the best vs good enough unless you are gaming or running a server.
I have good experience with Marvell / Phison controller SSD. I typically avoid the cheapest stuff with horrible benchmark, that usually tells me they are using some really extreme compromise and may not last long.
I have limited applications for these, and I rather like Apple where you’re currently locked in. I do have the last of its kind MacBook Pro where I last upgraded to a WD Blue 1 TB in 2019. I was using a 512 TB SSD but that started running out due to using it for backups. But those basically max out the SATA-III bus performance.
I did buy an SSD for work purposes last year but was in a hurry and got SanDisk’s lower priced SSD Plus version. I couldn’t find any official specs on its write performance. It regardless of all that, bulk transfer isn’t really what helps the most. It’s the lack of a HD seek time. There were something’s I’d do where I’d hear my hard drive making noise heading everywhere looking for tiny bits of data. That’s a lot faster now.
If you want to see how much and what is writing to the disk in the background, look at the "Resource Monitor" in Windows. You'll be surprised what's going on all the time. Seems "System" (the Page File and few other functions under System) is the main culprit, followed by your browser if you have a lot of tabs opened up.
Windows loves to use the pagefile for all sorts of crap.
Linux has the ability to use the swap file more or less aggressively. I wonder if there is some way to do the same in Windows. Most of us have more than enough RAM these days that the swap file never *needs* to be used; but some applications and system libraries tend to get a little funky in its absence.
I've noticed there is a lot of Swap File action going on in Windows, even though Task Manager/Resource Manager shows that only ~50% of the RAM is being used. So, who knows how the Swap File logic is setup - ?. So apparently, even if not all the RAM is utilized, the Swap File is still pretty active, for whatever reasons.
So I noticed with the latest Firefox update (112.0.2), they fixed a "high memory usage" issue. I can already see less CPU and SSD read/write activity when the computer is just sitting there and I'm doing nothing but looking at it. I've seen where the CPU will go nearly 100% for minutes at a time, and if I close Firefox it drops way down instantly, like something going on in the Firefox background is really taxing the CPU.
I have a SSD remaining life vs time graph, so will see if the curve changes life % use with time. I use Firefox 99% of the time, and usually have anywhere from 25-50 tabs open at once, so could be Firefox was also burning up my SSD life. If so, can't believe it took this long for Firefox to find a memory leak. Seems that would be near the top of the list for any browser performance requirement.
I have a SSD remaining life vs time graph, so will see if the curve changes life % use with time. I use Firefox 99% of the time, and usually have anywhere from 25-50 tabs open at once, so could be Firefox was also burning up my SSD life. If so, can't believe it took this long for Firefox to find a memory leak. Seems that would be near the top of the list for any browser performance requirement.
It really depends though. If it's being used like a cache, then most newer SSDs have an SLC layer that should reduce the number of writes/erases for the main array.
I'll see if there is a change in the Crystal Disk remaining life info I've been tracking.
But then there could be the matter of what "remaining life" represents. It's just a model and possibly not the same as the manufacturer's model. And then there's the partition into SLC cache and the main (likely TLC or QLC) flash. So if it says 75% remaining, what does it really mean? Those are two parts of the system that have different endurance. Maybe around 100,000 cycles for SLC and 5000 for TLC.
As an SSD ages, I would worry more about the possibility of catastrophic failure since they can fail randomly. And those chances go up as they're used.
All I'm saying is that I will determine if the Firefox update changes the rate of life reduction on this SSD. I've tracked the Crystal Disk info once a week for over a year. When the SSD % life goes down 1% every 2 months (6% in a year), it just seems like the SSD life % reduction rate is too high from just mainly using this laptop for surfing the web with Firefox and saving some files. Maybe it's all due to Windows 10 writing a bunch of stuff in the background. I can see writing to the SSD going on in Task Manager when the computer is just sitting with me doing absolutely nothing with it. What's writing to the SSD in the background all the time? Like said, when I close Firefox I see some of that idle writing stop.
Last edited:
