I think that changed a bit with gen4 NVMe, but yeah, warmer memory and cooler controller used to be the ideal.
1 Like
May I interrest you in a Transcend MTE220S SSD 2TB, M.2 SSD? TLC NAND, and most importantly TBW of 4.4PB with 5 year warranty. for 210-230EUR.
First Party Products
SK Hynix has seemingly received the appropriate approvals from US authorities and then PRC/Chinese authorities for its acquisition of Intel’s NAND business; so it is mainly Micron, SK Hynix, and Samsung that produce NAND, yes?
From what I have heard, the benefit of buying from a manufacturer’s brand is that there is more of a guarantee of consistency that a given product will always be sold with the same NAND and the same controllers; so buying a Crucial (Micron) or SK Hynix brand SSD would be as good as Samsung from that perspective.
In many ways though, I would imagine it is more important to avoid QLC than any particular manufacturer, even third party ones.
SSD End of Life
As far as SSD death is concerned, is not there usually is a counter in SMART that shows some form of wear information?
Either unusable blocks or bytes written could give you a sense of where you are in its life. I remember there was a multi-year SSD write test that ran until the drives permanently died; I think the conclusion was that there was ample warning in most cases before true failure,
In theory my ideal would be to (after some initial testing/wear-in) use a drive until some multiple of its warrantied TBW rating, then switch to using it for low-importance re-downloadable, or lose-able data until it actually dies.
The expensive ideal would of course be to set up something with ZFS and multiple drive error recovery.
SSD Unpowered Retention
One thing I sadly see brought up vary rarely is the possibility of not-wear-induced bit rot. I have no good source at the moment for how long a given drive or particular cell type (2-bit MLC vs 3-bit MLC (TLC) vs 4-bit MLC (QLC)) can last sitting on the shelf unpowered without losing data.
Since the famously long-lasting format of digital tape is, like HDDs, magnetic, I would guess that mature CMR HDD storage would have better retention than SSDs.
There is of course the caveat that old bearings in HDDs can lose lubrication or for some other reason seize up. Well retained magnetic data is of little help if you need to risk platter swapping to be able to spin it up.
2 Likes
One of the very first hard drives I had was a 20mb drive (yes 20mb) and it filled an entire 5&1/4 bay. In order to get it to spin I had to punch the drive to jog the platters enough to get the drive spinning when I started the computer. It was pretty hilarious. Very reliable drive besides needing a good punch every now and again 
(I even found a picture of the first hard drive I had online)
I do not suggest whacking your spinning hard drives though to get them working. Mine was a very specific case and only happened because the drive wouldn’t spin on startup without a bit of a jiggle.
It was pretty hilarious - made me think of Han Solo in Star Wars punching the Millennium Falcon to get it to work.
The nice thing about HDD’s, even if they stop spinning entirely, you can send them to data recovery centers and often they can pull the data off. Louis Rossman offers a repair service specifically for data recovery, and his service is actually affordable. *data recovery centers will charge you thousands of dollars for the same service Rossman offers.
SSD’s do have lifespan tracking, I use a software called CrystalDiskInfo - Crystal Dew World [en] which keeps track of drive health. This particular software is free and has been around for quite a long time (since 2008). It was authored and is kept up to date by the original Japanese developer who wrote it initially. I have found it to be quite reliable for monitoring the health of drives.
1 Like
Just to clarify and add some nuance, NAND flash cells are only temperature sensitive when writing, meaning that they will only wear down faster when writing at 25C than at 50C. Read operations are not affected by this, so I think it is inaccurate to say that NAND flash or NVMEs “like to run hot” when you are specifically referring only to write operations. Furthermore, it is entirely unclear how much this extra wear actually affects drive longevity, so I would still recommend using an appropriate cooling solution for your NVME drives.
I’ve been down this rabbit hole before, and as far as I can tell some internet journalist released a report around 2015 where he tested the write endurance of NAND flash cells on SATA SSDs at both 25C and 50C and found up to a 50% reduction of max TBW at the lower temps. The key distinction here being SATA speeds and not NVME. A few years later Gamers Nexus did a video referencing the original study and then a few years after that LTT did another video referencing both. But then about a year ago LTT snuck in a redaction of their original statements in the previous video into a non-related video and basically stated they couldn’t get anyone in the industry to corroborate or elaborate on any of this.
So either there’s a giant industry-wide conspiracy, or writing to NAND Flash at NVME speeds produces enough heat on it’s own to mitigate any of the unnecessary degradation. Heat sinks also work both ways and the excessive heat from the NVME controllers can also spread to the cooler NAND flash chips, potentially heating them up enough to also negate any of the extra wear. This last paragraph is speculation on my part, but I also think that most people don’t even come remotely close to 50% max TBW in general, so even if it is an issue most users won’t ever be affected either way and never know it was a problem in the first place.
1 Like
I usually use smartmontools (cross-platform)
or hdparm (Linux-only) for drive info. I wonder if CrystalDisk____ uses smartmontools drive database, or if it has its own, or maybe does not use such a thing at all?
@Trooper_ish It appears that Samsung does still sell sizeable 2-bit MLC drives: 860 Pro and 970 Pro. The 860 Pro does say out of stock currently though…
Regarding Optane, @JunTrunko had mentioned in a thread a while ago that it has a three retention guarantee, but apparently that is standard for enterprise NAND as well.
I have never thought to look, I wonder what t he retention guarantees are for consumer NAND SSDs, assuming there are any. In theory, it might make sense to bin consumer drives for retention longevity, but I doubt anyone would actually bother doing that.
In that Optane thread I had laid out my dreams for long-lived data storage; in case others here would be amused by such daydreaming:
1 Like
If you are worried about longevity of your SSDs, get one 8 TB Mech and one 2 TB NVMe, do a single backup write to the mech a day and leave the Mech mostly alone otherwise.
Yep, no RAM cache which will tank the random read/write performance.
True, but I was describing the thinking for limited space/hardware setup, where one drive is all you can have.
Personally, my desire is longevity, even at the cost of speed; I would still prefer to financially support a manufacturer that makes a storage device that lasts, and make that device last as long as possible. I would take reasonably-sized SLC without RAM caching at SATA speeds over PCIe 5-speed QLC any day.
From the vivid (and only tangentially topic-related discussion), I assume I did not pick the worst components on the market.
Topic done.
2 Likes
Sorry man. I apologise for dragging your topic off. I haven’t heard any gotchas about the Kingston parts you listed
You were not the only one, I have a suspiciously sizeable chunk of tangent text ↑ there as well; apologies @MazeFrame.
nah, it was all storage related. and most of it was useful information if he didn’t already know it.
I would avoid Kingston SSDs. When my Kingston A2000 500GB lost 1% endurance after only 3 months of use, Kingston Support argued this was “normal” and nothing to worry about. Now 3 months on or so it has lost 2% endurance and the other NVMe Drives I bought at the same time, (Kingston and Samsung) show no wear at all. This insistance on Kingston’s part that the product was behaving normally for an SSD, when it was obviously defective was infuriating.
1 Like
Well, without knowing much about your situation and use case, that does sound alarming but at the same time does not necessarily need to be a defective drive.
It all depends on how much you use that drive, if you or one of your installed programs write to it a lot, and what the other drives do. You may be correct in your analysis but another 500 GB QLC with the same file system could very well behave similarly if put as your primary drive.
Well, I’ve had a WD 570 Black as a C: Drive on another system for a year or so now and it hasn’t lost any storage cells yet. As it happens, I have an identical A200 500GB and a KC2500 2TB Drive that were purchased at the same time, both at 100% Health (as they should be).
Not saying you didn’t have a defective drive, just that there is a lot more difference on wear and tear between a server drive, a desktop drive, a drive that is installed into a secondary laptop, a drive installed on your daily driver, a drive installed on a crypto miner bot (knowingly or unknowingly)…
It might just have been that that particular install happened to have a cryptomining bot that stole your unused CPU cycles and SSD writes. Or it could’ve been a faulty drive. No way to know without doing a deep dive diagnostic here.
Bad drives happen to everyone but you’d be amazed how many blame the drive when it was an overzealous write program. That said, doesn’t matter if you were mistaken or not or what your case was, just that you are happy now. 
And for the record, this is my smartctl on a 2 year old, Kingston A2000 512 GB drive. Like most consumer companies, Kingston is pretty O.K. Not too worried about the drive life, yet.
$ sudo smartctl -a /dev/nvme0
smartctl 7.2 2020-12-30 r5155 [x86_64-linux-5.13.0-35-generic] (local build)
Copyright (C) 2002-20, Bruce Allen, Christian Franke, www.smartmontools.org
=== START OF INFORMATION SECTION ===
Model Number: KINGSTON SA2000M8500G
Serial Number: 50026B72824E8B9C
Firmware Version: S5Z42102
PCI Vendor/Subsystem ID: 0x2646
IEEE OUI Identifier: 0x000000
Controller ID: 1
NVMe Version: 1.3
Number of Namespaces: 1
Namespace 1 Size/Capacity: 500 107 862 016 [500 GB]
Namespace 1 Utilization: 349 708 521 472 [349 GB]
Namespace 1 Formatted LBA Size: 512
Namespace 1 IEEE EUI-64: 0026b7 2824e8b9c5
Local Time is: Thu Mar 17 17:31:50 2022 CET
Firmware Updates (0x14): 2 Slots, no Reset required
Optional Admin Commands (0x0017): Security Format Frmw_DL Self_Test
Optional NVM Commands (0x005f): Comp Wr_Unc DS_Mngmt Wr_Zero Sav/Sel_Feat Timestmp
Log Page Attributes (0x0f): S/H_per_NS Cmd_Eff_Lg Ext_Get_Lg Telmtry_Lg
Maximum Data Transfer Size: 32 Pages
Warning Comp. Temp. Threshold: 75 Celsius
Critical Comp. Temp. Threshold: 80 Celsius
Supported Power States
St Op Max Active Idle RL RT WL WT Ent_Lat Ex_Lat
0 + 9.00W - - 0 0 0 0 0 0
1 + 4.60W - - 1 1 1 1 0 0
2 + 3.80W - - 2 2 2 2 0 0
3 - 0.0450W - - 3 3 3 3 2000 2000
4 - 0.0040W - - 4 4 4 4 15000 15000
Supported LBA Sizes (NSID 0x1)
Id Fmt Data Metadt Rel_Perf
0 + 512 0 0
=== START OF SMART DATA SECTION ===
SMART overall-health self-assessment test result: PASSED
SMART/Health Information (NVMe Log 0x02)
Critical Warning: 0x00
Temperature: 34 Celsius
Available Spare: 100%
Available Spare Threshold: 10%
Percentage Used: 3%
Data Units Read: 26 494 436 [13,5 TB]
Data Units Written: 32 151 636 [16,4 TB]
Host Read Commands: 374 571 834
Host Write Commands: 450 758 281
Controller Busy Time: 7 267
Power Cycles: 928
Power On Hours: 12 123
Unsafe Shutdowns: 30
Media and Data Integrity Errors: 0
Error Information Log Entries: 0
Warning Comp. Temperature Time: 0
Critical Comp. Temperature Time: 0
Error Information (NVMe Log 0x01, 16 of 256 entries)
No Errors Logged
1 Like
I don’t think you quite understand. My issue with Kingston, is not about their Hardware per se. The products are well concieved, and we accept there will be faults, and natural variations as with any complex device.
My caution to any prospective Buyer is that, in my experience, the reponse of their Support Dept was exceedingly poor.
1 Like
Useful to hear; I have a bunch of Kingston drives, so when I need to make a ticket, I can brace myself.
Haha - Well hopefully you won’t need to. But really, I have a bunch too and theyr’e mostly great, but I just won’t go there again.
1 Like
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.