Anyone know how reliable they are?
I didn't know till now that they have wireless SD cards and was just thinking how easy and low power it would be to set a few of them up as file storage. They would also let me setup a place to store files slightly away from my internet connection and hackers.
Any thoughts would be appreciated!
Shall have to join the others here in saying that they aren't too reliable and should preferably not be relied on as backup.
Now as some of you here may know I have tested a number of SSDs, seeing if read speeds drop and how much and/or if they suffer retention failures.
But I have also tested a few usb drives/memory cards though that has been a bit more ad hoc meaning that basically they were just some I was using and not tested in similar ways or after a set amount of time.
Results for them have been mixed, some have barely slowed down at all even when they've been powered down for a long time while others have slowed down or even suffered retention failures though none of them have just been blank after connecting them to the computer.
As for flash technology, it's literally the worst medium of all for long-term unpowered data retention / cold storage. The only reason some people still had readable photo's on very old 32-128MB (yes, Megabyte) SD cards pulled from a digital camera after a year or two unpowered was that the early cards were SLC / MLC and large node (+40nm) based.
Have had flash drives with readable files after being unpowered for probably like 5+ years that were around 4-8 GB so some larger capacities also used decent NAND.
They did suffer from a bit of voltage drift however which isn't all that surprising.
Yes, type of NAND (SLC/MLC/TLC) is very important in terms of retention even if it doesn't tell the whole truth.
Think size of the node is not equally important however - have seen drives with NAND built on a small node impress and drives with NAND built on a large node that are not that impressive.
The old
"1 year guaranteed JEDEC SSD chart (showing unpowered data retention in weeks vs temperature) that everyone loves to quote was based on 2013-2015 era Intel MLC SSD drives that use larger nodes and are now woefully outdated in 2018. Modern consumer flash is now all TLC and smaller node.
JEDEC updated their paper on their requirements for SSDs (JESD218B to JESD218B.01) in 2016 and they still have similar chart included so if it was so outdated they certainly had opportunity to update the chart with more accurate information.
They do note that one of the differences between JESD218A and JESD218B is: "Refine the low-temperature data retention flow to accommodate the realities of the SILC retention mechanism in modern NAND" but I don't know quite when they updated JESD218A to JESD218B.
Think the chart is accurate - it's just that you hit the point where wear is going to cause retention to go below one year significantly faster with MLC NAND than with SLC NAND and so on and so on.
And, well...there's also the question if manufacturers actually make sure that those standards are actually met.
Had a test result recently with one drive that really makes me doubt that they have made sure it is up to snuff.
The "future" is QLC, and PR statements of new QLC stuff sent to tech sites is "missing" its node-size and durability ratings for a reason (QLC PE cycles are down to just 100-300 cycles vs 50k-100k of SLC / 5k-10k of MLC / 1k-3k of TLC). For a small amount of data you write-once, throw in the back of a cupboard for several years and forget about, even burning 25GB HTL / MDISC BD-R's is far better than using anything based on increasingly leaky flash.
While neither of them have mentioned the size of the node both Intel/Micron and Toshiba/WD have said that their 3D QLC NAND is specced for 1000 P/E.
Though if that's what enterprise drives using 3D QLC NAND are specced for then consumer drives may be specced for less than that.
Samsung however has been pretty silent on both counts though as far as I can remember.
Now even if endurance may not be the best retention could be decent (have read that 3D QLC NAND is supposed to be similar to 2D TLC NAND at 15nm), there's one drive I'm testing with fairly low endurance that has been fairly resistant to voltage drift that could possibly endure twice what it's been specced for and still stay within JEDEC specs.
Will want to see how it does in the current test to say anything more definitive than that though.
Well Samsung's first gen 850 3D NAND was 40nm. Since then though there have been rumors that Samsung has been shrinking it down to nearer 20nm again for the 860/960's. That would certainly explain the shrinking warranty sizes and lack of node-size info for newer drives (eg,
5-10 year 850 series vs only 3-5 year 860/960 series), not to mention how 860 ends up much cheaper than 850 (and close in price to the MX500 which is said to be using the 16-20nm 3D Intel/Micron process).
Where have you heard anything about the litography being shrunk for the 860/970?
I've heard that it is in their plans to do that at some point but not that they have already done so.
Regarding one way that can at least partly explain why the 860 EVO is close in price to the MX500:
http://thememoryguy.com/how-samsung-will-improve-3d-nand-costs/
And according to the same person 3D NAND of a smaller lithography can match 2D NAND at larger lithographies:
http://thememoryguy.com/how-3d-nand-shrinks-ecc-requirements/
I thought that was due to the increase from 32 layers in the first gen to 64 layers in the fourth gen NAND. Not shrinking geometries. The new 64 layer NAND also has tremendous write endurance for a consumer drive (1200TB for the 1TB 860/970pro). That's 1200 P/E cycles guaranteed.
That's just the TBW though which can have a pretty loose connection to the endurance.
Samsung says that their 64-layer 3D TLC NAND can take 7000-20000 P/E (think their 48-layer 3D TLC NAND had similar endurance) which is significantly higher than 1200 P/E.
Of course they don't mention how they get those numbers or if that NAND is binned well and if so is the NAND in 860/970 EVO of similar quality.
It's said to be both increasing layers and shrinking nodes. Hence why they go into great detail when layers get increased yet have quietly removed all detail about the node size latter for post 850 series marketing or explained why warranties have halved for premium PRO's. 1.2PB endurance is impressive for the PRO's, but that's MLC which is expected (even 16nm planar MLC of old MX100/200's or 19nm Samsung 830's also hit +1PB in endurance tests).
Think that the reason they have have halved the warranty period for their Pro drives is that they didn't consider it necessary seeing as noone else does either.
If I remember correctly (not entirely sure about that) then originally they intended to have a five year warranty for the 850 Pro but then they decided to up that since the SanDisk Extreme Pro had a 10 year warranty and they couldn't lose to SanDisk now could they?
I kind of miss the fact that they would mention the lithography more than the 10 year warranty to be honest.
Regarding endurance test results, those conducting endurance tests don't really tend to test retention which is a pretty important part.
If anyone here is using SSD's in external USB 3 enclosures as a backup drive, make sure you plug it in at least once every 2-3 months.
Probably more often than strictly necessary but it is good to check that nothing has been corrupted now and then.
Hey, I'm going to test an 850 EVO (Gen 1) after it's been powered down for a year.
Still a few months until then but would you be interested in seeing the results?
Facebook
Twitter