Enterprise SSD "small deals"

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

josh

Active Member
Oct 21, 2013
644
212
43
That SATA isn't so bad actually :).

I just grabbed 4 of these INTEL DC P4511 2TB (no more left, sorry :p ), 75 USD negotiated down to 60 USD / Piece for 2TB M.2 22110:


Just to say that it might be worth watching this Seller, some of their Offer are quite good (although I could NOT negotiate THAT much on this one, I tried and got autorejected for 40-45-50-55 then finally on the last one I got accepted at 60 USD).
Any reason to pick SATA over SAS?
 

josh

Active Member
Oct 21, 2013
644
212
43
Lower Power Consumption and works with your onboard SATA Controller, thus no need for an extra PCIe HBA, unless your Server Motherboard has an onboard already SAS HBA Controller (e.g. LSI 3108).
SAS is backward compatible
 

josh

Active Member
Oct 21, 2013
644
212
43
Who keeps buying these at these high prices man.. 265 of these were sold at only $100 just days ago
 

ca3y6

Well-Known Member
Apr 3, 2021
869
863
93
Who keeps buying these at these high prices man.. 265 of these were sold at only $100 just days ago
But that's a great deal, and as you can tell from the timeline of the thread, it was gone in minutes. The vast majority of buyers wait to make a decision to buy, and once they do, take the first price on ebay. But opportunistic buyers like most followers of this thread, who are happy to wait for months to grab a great deal (i.e. have the patience, the funds and available slots) are atypical. And I am actually out of that population (no more available slots). In fact I have a pile of unused 3.84TB SATA SSDs waiting in a closet which breaks my heart! I have been a net seller on ebay YTD.
 

luckylinux

Well-Known Member
Mar 18, 2012
1,675
548
113
Granted these are U2 at 80% health, but if you're looking for that would these be an ok deal?

Lenovo TOSHIBA 3.84TB
According to https://datacentersupport.lenovo.com/bm/de/downloads/ds557652 the Drive in the Listing seems to be at least 2 Firmware Revisions behind (3202T2GB) compared to the latest (3205T2GE and 3205T2GC).

Can we run the Linux Update Tool even though we don't own a Lenovo System ? Is it sufficient that the SSD itself is Lenovo ?

EDIT 1: I gave it a Try ... I could download lnvgy_fw_drives_all-1.39.11-0_linux_x86-64.bin and lnvgy_fw_drives_all-1.50.16-0_linux_x86-64.bin.

They don't really run as is (nobody checks if the File is running as User, just complain that they cannot write to /var/log/Lenovo_Support/ux.log).

Using binwalk one can somewhat extract the File, but then we are left with a .std File that doesn't really give out its Type via file Command.

Seems to just be Binary File, that's all, approx 1.3MB:
Code:
ls -lh ./_lnvgy_fw_drives_all-1.39.11-0_linux_x86-64.bin.extracted/image/files/3205t2ge.std
-rw-rw-r-- 1 USER USER 1,3M Jun 14  2022 ./_lnvgy_fw_drives_all-1.39.11-0_linux_x86-64.bin.extracted/image/files/3205t2ge.std
Opening via nano shows a few Lines of Header, then it's all Binary and/or encrypted:
Code:
CX05DHWF

Wednesday, March 17, 2021 20:16:42 bytsbgle
                                                          C5M:32:05:03:80

^Z        T2GE^@^@^@^@^@^@^H^@^@^@^@^@^@^@^H^@^@^@^@^@^@^B ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^A^@�^S^@�^@^@^@2^E^C�^D^@^@^@^@�^S^@^W[�>^\���~��Ε�h�^W�e"��՚^D^C9�>o5���#����?1�����O�#��^P�J�^C+*>��h�������C>
�Ce�kPv2�^?'��Dӷ7,^S�܌����w^O�i7�f�^Y����v ��^K�"3��4^N"+���e^F^E�FJ��V���M�^U^Ws8����6�ۊ���m^V~^W$ʿa^_#�^Wz^K`|�w^W�`눺�ߚt-fǥV��8J@^U�Ҽ�Th)F^V�"�F�Ϫc1^O�&[dk^O?��v�כݱ�5�aJ'����v^T������1>-`T2^T^^~E2^?⁗T^O�7ނ���i�^BF>>
;d(��!^B2^Y�_<L���      D*G�w^?KHv���^H�>^\��MS��b���
^P[p�Tk����u�7^N�^K�W��aD�;"��G-�I^P�x������^C����lj��w��^Y�(�k^Q^?lڗ�*^P��~^B^\�^[@�j,M�����T���q�T���^\Ô'���O�K^Tl�T^_^N�^Q@�^Hk�D��R��+�ĥ�\�eڭ��m�9�|c A�l^Q�7�Y��5^��.E+��&A�6^U��^A�^K���ܡ]^?�^D�|�1�/�$�R�-$`��^V^Q�>
����O��A^S9�^D�!�8^W'HVk^F���^\&9�w^K���^Cv6y�^?�^N^X�  ���.J^]^V��+y�Kp�y�^T�5�ߟd}��Kv"^@�x&�rMn^X�s�^S%⼒��W�^Ri�T=R^S�{(>ix��?L�C+���^^_^H]��"�]���^G��^D��Yp*zup3�"N�f^F��*�H-M��{��^W�Ƃ:k�^C^E/
^Q^N�U)^DM;��   h-��^Pk�$���a�OZ^Y���7l��^N�^U�0:Nk^V"�>��`��^VH�p5�^K^Y�2
...
...
...
EDIT 2: I asked Seller and he was like "I have no Idea what you are talking about", supposedly about updating the Firmware.

Rejected Offer at 75 USD / Piece for 2 Pieces.

EDIT 3: Offered 85 USD / Piece, not feeling very optimistic. Only 3 Pieces left :oops: .

EDIT 4: Rejected 85 and 100 USD / Piece. Countered at 120 USD the first Time then 140 USD the Second Time :rolleyes: .

EDIT 5: accepted 105 USD / Piece for 2 Pieces :cool:
 
Last edited:

Cruzader

Well-Known Member
Jan 1, 2021
980
953
93
15.36TB ruler format $500 Pardon Our Interruption...

(caution with seller)
I was tempted by their previous listing, but from a that fresh seller it feels a bit risky.
(Reusing the same pictures with same serials when first listing already sold just adds to that sense of risk for me)

It did peak my interest for ruler format prices in general tho.
A fair bit of 600-750$ 15.36tb listings that has been sitting for a long time that would probably accept some okay offers.
 

luckylinux

Well-Known Member
Mar 18, 2012
1,675
548
113
Mixed Workload / Mixed Use Drive :) .

0.5 Drive Writes Per Day (DWPD) for a period of five years -> 0.5 x 3.84 TB x 365 x 5 = 3504 TBW, quite Good I'd say

Datasheet:

No mention about PLP or Power Loss Protection though, I wonder if it has another Name or if it's not there at all ?

EDIT 1: Never Mind, it just has their Proprietary Name described in the DataGuard Technology Section:

1753901544895.png
 

ca3y6

Well-Known Member
Apr 3, 2021
869
863
93
Same pictures for both listing. That's a big red flag, unless they were a big seller using stock photos, which doesn't seem to be the case (or unless the previous sale was cancelled, which is a possibility).
 
Last edited:

kapone

Well-Known Member
May 23, 2015
2,046
1,390
113
Well Cheap, Good, Fast, pick 1 or 2 at most.

So Cheap it is, Good I don't know, Fast probably not :p .
That is a Fusion IO SX350 card. It'll do >3GBps easy. Granted not as fast as the latest NVME etc, but 6.4TB for $150 or less? I'm almost tempted to buy them.

p.s. They work great even in the latest Proxmox.
 
  • Like
Reactions: nexox

Markess

Well-Known Member
May 19, 2018
1,251
891
113
Northern California
Well somebody bought them. Seller rejected my Offer of 90 USD / Piece :rolleyes: .
I was tempted, but trying to do a little research before jumping.

I set up the home NAS with SSDs to keep noise down (a handful of 1.92tb Samsung I got for $40 each when prices dipped). But, it was so popular with the family….its “full”. So have to get something else.

Little cooling fans on a PCIe card shouldn’t be too bad, right?
 

luckylinux

Well-Known Member
Mar 18, 2012
1,675
548
113
yeah, and in this case SAS2 is kind of obsolete, but SAS3 wouldn't be. To be honest, I am not sure age is a primary factor in SSD reliability, as long as you are not pushing them (in term of operating temperature and load level). If the wear levels are healthy. But I have yet to find some stats to back that up, that show the impact of age on SSD reliability independently of TBW.
Microchips are not made to last Indefinitively (thinking mainly about the Controller). They mostly get killed by Over-Voltage Transients / Static Discharge / etc and of course they age much faster the higher in Temperature you go (Arrhenius Law). Which one "kills" first, whether it's the Calendar Aging or the Number of Write Cycles that each Cell experienced over its Lifetime (TBW), it's for sure a Combination of both.

Please Note that old SSDs have MLC Flash with "big" Cells Production Process (in Terms of Nanometers, ~ 20 - 30 nm) so that is better in terms of Longevity and Reliability compared to TLC/QLC NAND. Although in Terms of Calendar Aging, not sure if that makes a Difference. If it does I'd say that the MLC and "big" Production Process is actually better. Of course the same Remark applies to the Controller.

Capacitors mostly age over Time (Capacitance decreases over Time, Electrolyte dries up, ...) so that's for sure a Negative Thing. In these applications aging of Capacitors due to Load Cycling isn't a Factor. Since Manufacturers should design their PLC and Filtering according to the EOL Drive Characteristic (and hopefully they do so), then there should be some reserve Capacity to take that into account so even when old the Drive should perform correctly.
 
  • Like
Reactions: itronin

Markess

Well-Known Member
May 19, 2018
1,251
891
113
Northern California
Those are some incredible Prices :oops: .


If you have the Space for it and can give up the next PCIe Slot, sure. Or centrifugal Fan. Or Side Chassis fan depending on the Chassis Model.
Price was good, but not "incredible" for when I bought them 2 years ago. Back then, $45-50 for single pieces seemed pretty common and I offered for 5.

Yeah, if I went the AIC route for expansion, I'd need to switch to ATX/mATX. Ports are maxed out on the current ITX based NAS, so no way to add capacity there unless I went fresh with larger SATA SSDs. I have a couple suitable chassis and boards I could repurpose from the homelab for a "new" NAS. So, I think keeping the current pool and adding additional storage is my best bet to save $$$. Just need to decide what to do :p
 

luckylinux

Well-Known Member
Mar 18, 2012
1,675
548
113
Better negotiator than I lol. I took 3 at 120 last night. I was getting impatient with deals so these should scratch that itch until something great pops up. Plan to run them in zfs raidz1 for a download cache on my unraid server.

Buying these also means I need to pick up an 826-el1-n4 backplane for my case. Oh, and I don’t have pcie lanes to support so picking up an epyc 7302p with a romed8-2t.

if you have any luck updating them let me know. As long as they work I’m not overly concerned but certainly will if you pave the way.
Well that's already a Premium compared to the 75 USD I was initially willing to spend :rolleyes: .

Maybe I should have indeed gone with 3x, not sure, this way I can only do a Mirror. But at the same Time not being sure if I could update Firmware was a bit of a Risk Factor for me.

Now I'm having quite a different collection of U.2 SSDs though: 1.92TB, 3.2TB, 3.84TB and 4TB. Of course the 4TB (Intel P4500) can be "downgraded" to 3.84TB and the 3.2TB will be in another System so not really an Issue.
 

kapone

Well-Known Member
May 23, 2015
2,046
1,390
113
I'd need to switch to ATX/mATX.
Those are the exact reasons, I'm redoing my SAN nodes. Needed more flash, but was out of slots. So...



10x pci-e 3.0 x8 slots...The chassis (Chenbro NR40700) will need a bit of surgery to fit this monster, but should give me enough I/O for years to come. The Chenbro already has ~1PB of spinning rust, but more flash is needed (who doesn't need more flash!).
 

Wasmachineman_NL

Wittgenstein the Supercomputer FTW!
Aug 7, 2019
2,374
883
113
Those are the exact reasons, I'm redoing my SAN nodes. Needed more flash, but was out of slots. So...



10x pci-e 3.0 x8 slots...The chassis (Chenbro NR40700) will need a bit of surgery to fit this monster, but should give me enough I/O for years to come. The Chenbro already has ~1PB of spinning rust, but more flash is needed (who doesn't need more flash!).
god that would be so rad to run LLMs on. get a shitton of 3090s, 512GB of DDR3, a pair of 2697V2 or 2687W v2's and go crazy
 
  • Like
Reactions: kapone

luckylinux

Well-Known Member
Mar 18, 2012
1,675
548
113
Capacitor health has always made me wonder. Some enterprise drives report a cap health metric but most dont. Has anyone experienced SSDs going bad from caps?
No direct Experience on Capacitors and these Enterprise Drives but IMHO an Enterprise Drive with PLP where the Capacitor aged too much "simply" means that PLP is no longer effective (either not working at all i.e. Lost Function, or for too short period of Time i.e. Degraded Function), effectively it's the same Thing as having a Consumer Drive without PLP, and for most Scenarios with ZFS (even if PLP is highly recommended), then in the worst case you are only going to lose the last few Seconds after an unexpected Power Loss occurred. So the Issue will only manifest itself after an unexpected Power Loss, otherwise none at all. Some PLP Drives do indeed do a "self-Test" of the PLP Capacitor Function and report that Information in the SMART Attributes / Log.

Of course if the PLP Capacitor goes into Short-circuit that's a whole other Story, but that will likely not be reported at all, if anything a fuse Blows or your PSU trips. But how likely is that going to happen ...

Small Filtering Capacitors will not report anything to the Controller obviously, so the only Effect will be a slightly less filtering / decoupling of e.g. 3.3V Supply and some Signals.
 
  • Like
Reactions: crlt

kapone

Well-Known Member
May 23, 2015
2,046
1,390
113
Well Flash is also SAS/SATA. I guess you WANTED (some at least) NVMe
True. Should have been more specific.

ou'll need several HBAs in Parallel anyways
True. Hence the giant monster motherboard.

LSI 9306-24i for 24 Drives but can "only" do 64 gbps
I'm already using two Adaptec 8 series in the chassis for the spinning rust, one per expander/24 drives.

10 SATA Drives or ~ 5 SAS Drives you will bottleneck
Nah. That's fantasy stress testing stuff. Real life...a single pcie 3.0 x8 HBA is more than enough for 24x spinning rust, even the latest ones.

1PB of Spinning Rust
Yup. And that's replicated (well, not all, about half to 2/3) to the other SAN node. Both are identical.

EDIT 1: That Chembro Chassis seems to be super-expensive and difficult to find though
True. Bought three of them years back. They couldn't give them away at the time. I think it was $200/ea shipped and came with a nice motherboard.

Officially it only supports 7xPCI Slots though, so I'm curious how you are going to make that work
I don't need all 11 slots to have "openings". The Fusion IO SX350, the Adaptec HBA etc are all internal only. They just need to be secure enough in the chassis. It's only the NICs etc that need the slot openings.

so I don't think it will work in this one either:


the Xeon E5 v3/v4 Model, the Supermicro X10DRX, has 11 PCIe Slots even
It does. And they're open ended even! But...that motherboard is like $500...this was $27...less than a lunch.

also the Motherboard you mentioned and the Picture shows 11 PCIe Slots. Why did you say 10x Slots, is one of them not working
No, sorry. It does have 11 slots except one of them is pcie-2.0 x4 from the PCH. I don't even look at these slots. blah...
 
  • Like
Reactions: nexox

kapone

Well-Known Member
May 23, 2015
2,046
1,390
113
Probably better the X10DRX at that Point, 300-350 USD instead of the 50 USD for the X9DRX, but potentially at least double the RAM and CPU Power. You'll need the GPUs anyways but I mean budget wise the increase in Motherboard Cost isn't that much when looking at the Total Cost of the System at that Point.

But ... I Googled for half an Hour but couldn't find any reasonable SSI-MEB or HPTX nowadays, there are a few Threads around, but it's about Cases that were built over a decade ago and were quite a Niche Market (i.e. low availability nowadays):
The Chenbro NR400700 motherboard tray is essentially a full width (~17") 3U open chassis, since the power supplies sit "below" it, making the whole system 4U.



At one point, I had a stack of HGST SAS SSDs mounted next to the motherboard... :)
 
  • Like
Reactions: luckylinux