hello, is the sm883 (HXM7904Q) been tested and works? is it different firmware for different drive capacity?
whats best way to flash this? from dos bootdisk or from windows?
Firmware package for Samsung SM883/MZ7KH3T8HALS
- Thread starter oh2ftu
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
whats best way to flash this? from dos bootdisk or from windows?
Can confirm, run it with 8 SM883 1.92. capacity drives for vSAN. I flashed them on a Windows system.
No. Neither better nor worse behaviour. I think this update adress only edge things which may happen under rare circumstances.
you could infer the changelogs from the HPE version SM883 revision history.
These are the fixes in the latest HPE firmware HPG3 (I assume this is HXM7904Q equivalent)
And HPG2 had these fixes
HPG1 fixes
support.hpe.com
These are the fixes in the latest HPE firmware HPG3 (I assume this is HXM7904Q equivalent)
- PrefetchClear logic error on Write CMD.
- ICRC Data Integrity improvement.
And HPG2 had these fixes
- When SSD statistics are queried, erroneous drive temperatures may be reported. When this occurs, the system may encounter any of the two following conditions, leaving the server in an unstable state.
- Server shutdown
- In VMware vSAN environment, storage volume taken offline.
HPG1 fixes
- Fixed an issue with performance drops during mixed workload operations.
HPE Support Center
SM883 is 64 layer MLC like 970 Pro or 860 Pro. Those 980 Pro uses 128 layer TLC NANDs so your drive should be OK
I am just stating my observations:
- all the models that had problems lately had 128 or higher layers (870EVO, 980 PRO, PM9A1, 970 EVO Plus Elpis controller version, 990 Pro)
- No widespread issues reported on any of the 64 layer models
It's not that easy, but they had some issues with they 128-layer gen6 "V-NAND" like they have on the 176-Layer -oh, where is the pm1743
.Some issues are rare and occur only under certain circumstances, the sm/pm883 had 10 (!) firmware revisions over the time. Older models probably have serious flaws too, which were (and will..) never be corrected.
The 24-layer MLC gen1 had 3.6 DWPD (over 5 years), on the sm863 with 32-layer it was about ~3,5 DWPD, then it sank again to 3 DWPD on the 48-layer sm863a and stayed there for the sm883 (gen4 64-layer MLC) with the same endurance. They stopped the (2-bit) MLC NAND and done the same 3 DWPD on the 128-layer gen6 pm897 with rather heavy overprovisioning... ~6TB (?) raw for 3,84TB usable.
But remember, the more used the cells get, the faster they loose information. Servers are most of the time powered on, but if you take a heavily used/written server SSD off the power and wait for 6 months or even less, then it can be quite possible that the data is gone.
well, data retention is one thing (then the data retention is spec'ed for 3 months @ warranty exhausted TBW, not at zero TBW state)It's not that easy, but they had some issues with they 128-layer gen6 "V-NAND" like they have on the 176-Layer -oh, where is the pm1743
Some issues are rare and occur only under certain circumstances, the sm/pm883 had 10 (!) firmware revisions over the time. Older models probably have serious flaws too, which were (and will..) never be corrected.
The 24-layer MLC gen1 had 3.6 DWPD (over 5 years), on the sm863 with 32-layer it was about ~3,5 DWPD, then it sank again to 3 DWPD on the 48-layer sm863a and stayed there for the sm883 (gen4 64-layer MLC) with the same endurance. They stopped the (2-bit) MLC NAND and done the same 3 DWPD on the 128-layer gen6 pm897 with rather heavy overprovisioning... ~6TB (?) raw for 3,84TB usable.
But remember, the more used the cells get, the faster they loose information. Servers are most of the time powered on, but if you take a heavily used/written server SSD off the power and wait for 6 months or even less, then it can be quite possible that the data is gone.
and abnormal early failure is another. which 128 or higher layer Samsung drives seem prone to.
Seems like more layers is inversely proportional to durability. Maybe I should hoard SV843 1Tb (24 layers) drives....
Then I wonder whether this is Samsung specific trend or something universal
This is a complex question... one part is the NAND flash. Several techniques like Charge-Trap Flash, Floating-Gate, Replacement-Gate, the structure size (lithography, under 50nm is degrades nonlinear fast) , how many bits are stored per cell -SLC, MLC, TLC, QLC... and of course the binning of the NAND-dies; all this and more impacts durability and retention time. With 3D-NAND they observed that the quality of different Layers - bottom, middle, top - differ extremely. This and other effects like write-disturbance, where even read cycles degrade the information integrity of neighboring cells play a big role, the temperature at the data was written too... many many factors.
There comes the controller and especially the firmware "magic" in action. How good and fast is it at error-detection/correction, does it know the strong and weak cells of the often over 100 dies (512dies in 32x HDP packages on a 30TB pm1643). When is the time to refresh/rewrite the cells, wear-leveling, garbage-collection, overprovisioning, spare-cells. Every manufacturer has their own tricks ..and snake-oil.
IMHO one should have the option to choose if he wants to set the cells into pSLC-mode.
You can put a thermal pad on the controller and probably the PMIC to enhance the longevity, and give 20% to 25% of the space for the drive to extend the life due to lower WAF and get better/constant write speeds.
I know not too much about the details, but if you take the firmware and controllers from today and pair them with the older NAND, some unexpected "wonders" can possibly happen, but the price per GB will magical let disappear your $$$.
Just try it or leave it, there is a risk in everything, but the chances are rather high that you gain something. I updated 5 sm883 3,84TB and an sm863a 1,92TB, no troubles so far.