No, to keep it short, in accordance with this comment: “With the additional speed and efficiency come lower write-endurance ratings, mostly as a result of the switch from 2-bit MLC NAND to 3-bit MLC NAND.” … and additional info to the linked article would be that 970 PRO 512GB model have warrantied TBW of 600 (5-years or TBW, whichever comes first). But, up to my understanding, 980 PRO read endurance and performance isn’t affected, so if needed as reliable drive (on second M.2 slot), frequently used, it might easily be the one to choose.
1 Like
DW/D? What are you guys doing with your disks? 
My desktop/laptop usage would be measured in DW/Y.
Unless you are actually doing e.g. OLTP or DVR then SSD lifetime is unlikely to be an issue.
The one thing that I would say is … periodically check disk health e.g. with smartctl
2 Likes
All I tried to say, to @Julie21, is that 970 PRO (512GB), as offered/recommended from Purism support side, must be very fine SSD, if something reliable (worthwhile) needed. It might be slower in numbers, but not in endurance (DW/D), if 2-bit MLC type of memory cell preferred (instead of something with more TBs inside M.2 SSD size). Just by quick comparison of two 1TB drives I’d buy older generation drive as better one (not even being interested to know what is current price difference of two) for home usage.
Another comparison of two (memory types) would be that 2TB 3-bit MLC (Samsung T7) you can already get packed into external USB-C 3.4×2.2×0.3 inches box, yet for 2TB 2-bit MLC there will be need for quite bigger and probably warmer box, I guess.
2 Likes
the 970-pro is not a slow drive by any stretch of the imagination.
the issue at hand is that if you’re a consumer/home-user doing only light to medium write operations daily (W/D - 0.1 - 0.5 max) then a consumer grade drive will do just fine.
both 970/980-pro are consumer grade drives but they are also oriented more towards the pro/enterprise/data-center conscious users.
for instance if you’re planning on doing daily whole drive wipes (format > completely fill with zeros the whole drive) + other ops’ then you’ll want one that’s enterprise-grade/data-center grade drive ( i.e more than 1 W/D certified)
more explanations - yanked from the WesternDigital tech-brief paper :
Converting Between DW/D and TBW
As long as the capacity and warranty are available, the conversion
between the two measurement units is straightforward.Convert from DW/D to TBW: TBW =
Capacity(TB) * DW/D * 365 * Warranty(years)
Convert from TBW to DW/D: DW/D =
TBW / (365 * Warranty(years) * Capacity(TB) )
1 DW/D Does Not Equal 1 DW/D
A common misconception when examining SSD specifications is
that “1 DW/D” on one drive means the same endurance as “1 DW/D”
on another drive. That’s not true, even for drives of the same model
but different capacities. To understand why, examine the case of a
3-DW/D, 800GB1 Ultrastar DC SS300 SSD and a 3-DW/D, 1.6TB Ultrastar
DC SS300, both of which share a 5-year limited warranty:800GB Ultrastar DC SS300 @ 3 DW/D:
0.8TB * 3 writes/day * 365 days/year * 5 years =
~4.4PB lifetime data written1.6TB Ultrastar DC SS300 @ 3 DW/D:
1.6TB * 3 writes/day * 365 days/year * 5 years =
~8.8PB lifetime data writtenThis effect becomes even more pronounced as drives increase in
capacity. When comparing the highest capacity model of the Ultrastar
DC SS200 at 1 DW/D with the highest capacity Ultrastar DC SS200
available in a 3-DW/D specification, we see that even though the
DW/D value of one is three times the value of the other, the actual
amount of lifetime data that can be written to the 3.2TB model is only
about 25% more:Ultrastar DC SS200 7.68TB @ 1 DW/D:
7.68TB/day * 365 days/year * 5 years =
~14PB lifetime data writtenUltrastar DC SS200 3.2TB @ 3 DW/D:
3.2TB * 3 writes/day * 356 days/year * 5 years =
~17.5PB lifetime data writtenGiven the dangers of under-specifying the endurance, carefully
selecting an SSD of the proper endurance level maximizes reliability
and minimizes costs.
For workloads in which the amount of written data is well known,
either by design or by measurements of existing systems, the choice
of endurance levels is straightforward: Determine the average
amount of data written per day, multiply it by the number of days
a server is in service (i.e., three or five years), and then use that
number as a lower bound for your endurance selection. This average
number is a lower bound because when using measurements of
today’s workloads and extrapolating, it’s a prudent practice to add
headroom for expected growth.
it’s probably overkill to go this deep into these things. for home users a SATA-3 based SSD is going to be just fine as a boot drive + some occasional R+W.
WRITES are the most taxing for an SSD so keep that in mind though …
Do you have a figure for DW/D for your own computer?
I have a 250GB disk, doing about 4DW/Y i.e. 1TB/Y. Warranty is 150 TBW. So lifetime is 150 years. Barring major advances in medicine, that is going to outlast me! Of course the warranty is also limited to only 5 years. So it will be well out of warranty and could die after 5.1 years. Realistically: I expect the disk to last as long as the computer itself, hopefully a minimum of 10 years.
(Part of the reason for the low DW/Y figure is that this is just an access device. Files are on the file server. There is not much stored on the computer itself. An endless stream of Linux updates. Some downloads there temporarily.)
i could try to guesstimate an average but probably bellow 0.1 W/D more accurate would be an yearly figure since in my case w/d wary wildly. some days it could be 1 w/d and the rest it could be months at a time with 0.01 w/d
it’s very tricky to have a figure. speed is more important than writes for me
i’ve tested a 512 GB m2-NVME-SSD (980-pro) some days ago with my Librem-Mini but only as an external drive trough my usb-c-3.0 connector.
it is severely handicapped by the 5 Gbps usb-3.0 interface but the access times are 0.07 (raising eyebrows) 
you can probably guess that sequential R+W are severly limited by the max usb-3.0 connection.
i’ll test it on my workstation that’s got PCIE 4.0 and 10Gbps usb port and report back when i can 
2 Likes
Up to this German article, size of Pseudo-SLC-Cache within 512GB Samsung 980 PRO is 4+90 GB, meaning you’ll get max. of 5.000 MB/s with TurboWrite and afterwards (>94GB) max. write speed of at most 1.000 MB/s. Anyway, as Samsung 970 PRO doesn’t possess/rely on any SLC-Cache and as such preserves write speeds (which aren’t related to size of data) all the time up to 2.300 MB/s (512GB) or 2.700 MB/s (1024GB), throughout the whole size of particular 970 PRO M.2 drive. Therefore sorry @reC, I wish I can understand your point but I cannot 
, perhaps just another matter of taste 
or my old-fashioned point of view why I find 970 PRO as quite better drive (but don’t own one).
1 Like
the problem with these m2 NVME drives is that they can only turbo for a short while if they aren’t properly cooled/managed. i.e they are best known for their awesome speed bursts but not exactly great when it comes to throttling down because of heat. mine throttled down after a couple minutes to about 170 MB/s (a little LESS than my SATA3 m2 WD red-pro that CAN sustain above 200MB/s for WAY longer periods of time WITHOUT any proper cooling mechanism attached or otherwise.
it’s simply a matter of how each person uses the tools available for the right job …
Unsure if this should be a new thread, but considering many aspects of the hardware are being discussed, I’ll post this here.
On the page for the Librem 14 it says " USB-C Power Delivery Port". My question is there a minimum Wh needed to properly charge through this port? My current laptop says 65+Wh for charging over type C. Was just wondering if there is a similar requirement for the new Librem 14.
USB-PD spec allows for different charging rates based on the capabilities of the charger and host. It’s likely that if you use anything less than 65W it won’t actually charge while in use, but I’ve not tested with lower power chargers. Your standard 15-20W USB-C phone charger won’t do anything useful, that’s for sure.
Even when the system is suspended? It wont ‘trickle’ charge?
That’s more likely to be a problem with a laptop or an ultracompact (like the Mini).
I have a desktop, with nothing more than a standard case fan, and while backing up the NVMe to external media (and limited by the speed of the external media / external connection) it sustains way more than that (> 500MB/s) for long periods.
I have a fanless ultracompact, more limited in hardware than the desktop, both by the NVMe drive and by the external media / connection, and while backing up it still manages to sustain over 300 MB/s for long periods.
So you probably need to look more closely. Unclean vents? Firmware update for the drive?
nope, just external drives connected to the LMini-v1 through usb-3.0. i wasn’t talking about the benchmark from GDisk. simply transfer/back-up of larger than 100MiB files (can go up to 30-40GiB sometimes)
i don’t believe ShangSung has gotten the chance to release firmware updates for the 980-pro yet as it’s still so fresh on the market.
had my 980-pro in this > https://rog.asus.com/storage/rog-strix-arion-model < connected to the LMini-v1 (only reaches 5Gbps sadly and gets HOT)
So is the internal drive NVMe or the external drive is NVMe or both?
Any time you are backing up like that to external drive, you would suspect the external drive / connection before suspecting the internal drive. (I’m backing up the entire drive, so heaps more than 40 GiB.) I’ve seen some pretty disappointing external SSDs.
Maybe start a new topic rather than continue to hijack Librem 14: Almost There!
it’s ok. i’ve quite finished here
i meant INDIVIDUAL files going up to 40 GiB not everything … see above
Ok. My chargers range from 65-100W. So I think I should be good. Just don’t want to overload the thing.
that’s not how USB-PD works, voltage and power are negotiated; it’s pretty much impossible to overload a device or charger if both ends are spec-compliant
3 Likes
So are we now just weeks away from shipping the 14’s?!?!
Similar question was asked on Matrix, and the answer was, no specific time can be given at the moment, but progress is being made.
1 Like
what problems still need to be solved before sending the Librem ?