Shouldn’t 67TB written result in just 2.2% (67TB/3PB) of the total TBW used so far, instead of 5%?
What exactly do those “Data Units Read/Written” and “Host Read/Write Commands” counts tell me?
Interestingly, if I sum up the totals of both “Read” and “Write” respectively I end up with 1.53PB for the Transcend and 0.145PB for the Lexar drive. 1,53PB/2,2PB = 69.5% and for the Lexar 0.145PB/3PB = 4.8%. Those two numbers come fairly close to what is reported as wear looking at the reported percentages of 30% and 95% as overall health.
Do both, reads and writes count against the proclaimed TBW? If so, why is it called “Total Bytes Written”?
From your experience, is the Lexar drive deteriorating too fast?
I have Seagate Firecuda’s (510 type).
used percentage by sudo nvme smart-log /dev/nvme (in linux) is (much) closer to the calculated lifespan from data unit written vs TBW specs.
Seagate Firecuda 510 (500GB)
TBW (spec)
data-units written
bytes written
TBW
% used calculated
% used reported (nvme smart-log)
650
690.968.674
353.775.961.088.000
354
54%
61%
Seagate Firecuda 510 (2000 GB)
TBW (spec)
data-units written
bytes written
TBW
% used calculated
% used reported (nvme smart-log)
2600
3.982.751.174
2.039.168.601.088.000
2.039
78%
80%
I think your idea about adding read and written bytes is a coincidence, if I do that my firecudas are far over 100% in total.
Looks like the marketing people of Transcend and Lexar are exagerrating a bit. As marketing people tend to do… Seagate marketing people also but a bit less
The Lexar link shows .44DWPD so
TBW = 1x.44x365x5 = 803TB
Difference between host read/write and data units read/write relates to how many blocks are actually need to be written. The drive manages the writes in order to wear level the drive. This means it may organise writes on the drives DRAM in order to reduce the number of writes to the NAND. Remember that each time you write to a NAND cell you degrade it a little bit and the drive does it’s best to mitigate this.
Also the drive may have some data that the host is requesting in its read buffer already so it doesn’t need to go to the media and services the read from cache.
I think (well, hope) those Lexar marketing people are just lazy and dumb, copy pasting their spec sheets and doing an occasional whoepsie.
The 0.44 for their ‘Professional’ series is exactly the standard/consumer DWPD, so they must have copied from consumer to professional
If not, the claimed 3000TBW is not an exaggeration but a downright lie.
Also possible of course, Lexar of today is not Lexar of 10 years ago…
Ouch. I filtered for the TBW and DWPD on a price comparison site and it seems that the site calculates the DWPD. I blindly relied on it without paying much attention to DWPD.
I’ll contact Lexar and ask about this decrepancy, because this doesn’t check out at all.
Sometimes “Professional” means a better controller and extra DRAM for better performance. To get more endurance you need to add more NAND to the overprovisioning pool which increases price. The NAND itself generally has the same endurance unless you change the number of bits being stored in each cell i.e SLC/MLC/TLC etc. the more bits per cell the lower the endurance…
