SSD Write Endurance 25nm Vs 34nm

[Anvil]

I've been traveling lately, just got home and business is as usual for the little drive, tomorrow will be 11 months of testing.
(closing in on a year of testing)

Kingston SSDNow 40GB (X25-V)

873.60TB Host writes
Reallocated sectors : 05 41
Available Reserved Space : E8 99
POH 7846
MD5 OK

33.55MiB/s on avg (~117 hours)

[Ao1]

^ Incredible. My 830 is coming up for the 3 month data retention test next week I believe.

[Christopher]

The X25-V is a rockstar. 11 months and 1.1 million thread views.

Anvil, have you put any more thought into the next drive?

[Christopher]

More likely a firmware bug than the quasi-9174 controller being at fault. Although if OCZ is running the controller at a higher clock rate, I suppose there could be some strange interaction between the controller and the NAND flash.

A01 -- I've not rebooted or cycled the system in weeks. I've been out of town for this week and won't be back for another week or so. I feared reallocations might reset with a power cycle, but the more I think about it, the less likely I think it is to happen.

I believe one plane/die is sub par. If that one plane/die/etc goes, that's the end of the road. If these reallocations are occuring over the entirety of the flash, then the outlook is slightly more hopeful. I don't want this damn thing to die before I get home to yank it out of the rig.

But there have to be implications for substantially increasing clockspeed of the controller. It could bite them in the ass long term.


-------------------------------------------
170 reallocations and 89 MWI

Jesus, I just looked 5 seconds later and reallocations are now 172.
-------------------------------------------

The 830 is at 2810PE and the Vertex Turbo is 6032PE without reallocations. I estimate the Octane has used somewhere in the neighborhood of ~2600.

[B.A.T]

Todays update:

Intel X25-M G1 80GB
511.13 TiB
22349 hours
Reallocated sectors : 88 to 89
Available Reserved space: 94
MWI=220 to 216
MD5 =OK
46.46 MiB/s on avg


m4
702.9650 TiB
3015 hours
Avg speed 73.00 MiB/s.
AD gone from 210 to 206.
P/E 12183.
MD5 OK.
Reallocated sectors : 00

[Christopher]

[canthearu]

@Christopher: I guess the octane hit that brick wall of reallocated sectors it was creating for itself. Very unusual for a marvell controller to do that. I must get one of these drives one day and see if it is repeatable.

@BAT: Your new M4 is looking even better then the old M4, and is getting close to the same wear. It is showing less CRC corrections, and fewer read errors. Maybe 25nm fabrication quality has improved over the last 6 months.

[Christopher]

226 reallocations at the last log. MWI decreased by 115 from the 226 popped blocks.

[canthearu]

226 reallocations at the last log. MWI decreased by 115 from the 226 popped blocks.

Maybe a secure erase would help .... but probably not.

[kensiko]

Too bad for the octane. Let's hope the Vertex 4 doesn't suffer the same misbehavior.

True the x-25V is something else. Got 2 of them, very pleased to see it breaking records.

[canthearu]

Patriot Torqx-2 64GB - Day 59

Drive hours: 1,438 (926 Hours, 512 missing)
GiB written: 94,736 GiB (92.52 TiB)
Avg MB/s: 20.80 MB/s
MD5: OK

Bad blocks: 2 grown, 83 factory
Wear cycle counter: 5921 average erase cycles, 9084 maximum erase cycles.

Intel 520 60GB - Day 51

Drive hours: 1,236
Avg MB/s: 93.22 MB/s
MD5: OK

Host GB written (F1): 388,069.63 GiB (378.97 TiB, 12418228 raw)
NAND writes (F9): 274,787 GiB (268.35 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 118 normalised
Avaliable Reserved Space (AA): 100 normalised
Media Wearout Indicator (E9): 11

Notes:

I have a Vertex 4-120gig sitting here ... I'm having difficulty deciding whether to replace one of my older drives or just put the vertex 4 to the sword!

I can confirm that the vertex 4 has a working total write sector count and hour/poweroff counters! Don't know about lifetime counter.

[B.A.T]

@BAT: Your new M4 is looking even better then the old M4, and is getting close to the same wear. It is showing less CRC corrections, and fewer read errors. Maybe 25nm fabrication quality has improved over the last 6 months.

Biggest difference betwheen them is firmware. The first had fw 0002 and this one uses fw 0009.
The first had higher writespeed (88 MiB/s) but I don't know if that count for anything.

This time I'm going to break 800 TiB and maybe it'll reach 1 PiB.....

[Christopher]

I'm not there in person, but the octane is un detectable. I was thinking a reflash of 1.13fw might help.

But the OCZ toolbox can't detect it either.

[Nizzen]

canthearu:
Put th Vertex 4 to the sword!

[Christopher]

I'm never testing a drive without decent smart data again. Not happening. Hell, OCZ should thank me for buying (it's not even a good value) and putting their -- (insert adjective here) -- drive in my precious endurance rig. I was thinking they were gonna bring the rain... And I was wrong.

[canthearu]

Less then 300TiB written, probably can still get RMA replacement. *evil grin*

[Christopher]

Less then 300TiB written, probably can still get RMA replacement. *evil grin*

That presupposes I want that drive anywhere near me. And, if I'm honest, it's no bug. I think the issue here is that much of the flash being used is not that great. It only takes one not so good die to ruin the whole drive. 7 of 8 NAND devices may be straight golden, but that 8th one can just be terrible sometimes. The first turbo I did was like that. But my displeasure with the Octane really centers more on its Godawful SMART data and poor relative value. All drives will have some sort of redundancy in higher and mid range capacities soon, and for this very reason. A drive is only as good as the worst piece of flash on it. Make sure the worst flash on the drive is still pretty good, and you've got a deal IMHO.

[canthearu]

I'm never testing a drive without decent smart data again. Not happening.

Well, the vertex 4 has working smart information. Looks like my drive got a 600 GiB wear in test as well before being packaged (my drive was sealed and the sticker was intact for the anti-static back, so definitely new.

[canthearu]

Patriot Torqx-2 64GB - Day 60

Drive hours: 1,462
ASU GiB written: 96,352.33 GiB (94.1 TiB)
Avg MB/s: 25.48 MB/s
MD5: OK

Bad blocks: 2 grown, 83 factory
Wear cycle counter: 6050 average erase cycles, 9283 maximum erase cycles.
Total ECC Error Count: 1

Intel 520 60GB - Day 52

Drive hours: 1,260
ASU GiB written: 392,671.79 GiB (383.47 TiB)
Avg MB/s: 93.91 MB/s
MD5: OK

Host GB written (F1): 395,413.34 GiB (386.15 TiB, 12653227 raw)
NAND writes (F9): 279,989 GiB (273.43 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 108 normalised
Avaliable Reserved Space (AA): 100 normalised
Media Wearout Indicator (E9): 9

Notes:

Changed to Intel rapid storage drivers 10.8. Getting somewhat of an improvement with the torqx 2 drive, and a little improvement for the Intel 520 drive.

Updated windows. Rebooted and the Torqx 2 fixed it's power on hours counter ... weird.

Also worked out how to correct the ASU counter, so using that for the Torqx 2, and including it in the intel stats now.

[Christopher]

Well, the vertex 4 has working smart information. Looks like my drive got a 600 GiB wear in test as well before being packaged (my drive was sealed and the sticker was intact for the anti-static back, so definitely new.

Really? The Plextors supposedly go through "20 hours of burn in testing" which probably means they just write a couple hundred GiB. That's a good thing, but I wonder how they swing that at the factory. Adding 20 hours to the build process can't be cheap.

[Christopher]

690,895 GiB
289.96MB/s

The 830 is now over 3000 PE cycles, but it's still at 17 MWI. I've been told it's "3000" PE cycle flash, so one of those numbers is wrong probably.



------------------------------------------
351,527GiB
84.53MB/s

Just... awesome.

[Anvil]

The X25-V is a rockstar. 11 months and 1.1 million thread views.

Anvil, have you put any more thought into the next drive?

I haven't decided yet, not sure that I've got one that really needs testing, some are already getting a bit old...
If I can't make up my mind in due time for the 1 year "anniversary" for the thread I'll find some way to randomly select one.

...
-------------------------------------------
170 reallocations and 89 MWI

Jesus, I just looked 5 seconds later and reallocations are now 172.
-------------------------------------------

The 830 is at 2810PE and the Vertex Turbo is 6032PE without reallocations. I estimate the Octane has used somewhere in the neighborhood of ~2600.

A good thing that one can spot when things really start to fail...
If this was a real drive with important data you'd have ample time for a "backup".

I'm never testing a drive without decent smart data again. Not happening.
...

That's why I haven't considered using the Plextor, it's just not worth it.

Endurance_cr_20120417.png

I'll leave the Octane in the chart. (until it can be physically checked)

[Anvil]

Kingston SSDNow 40GB (X25-V)

879.25TB Host writes
Reallocated sectors : 05 41
Available Reserved Space : E8 99
POH 7895
MD5 OK

33.37MiB/s on avg (~166 hours)

[B.A.T]

Todays update:

Intel X25-M G1 80GB
516.76 TiB
22387 hours
Reallocated sectors : 89 to 93
Available Reserved space: 94
MWI=216 to 205
MD5 =OK
42.18 MiB/s on avg


m4
712.7236 TiB
3053 hours
Avg speed 73.09 MiB/s.
AD gone from 206 to 201.
P/E 12346.
MD5 OK.
Reallocated sectors : 00

[canthearu]

That's a good thing, but I wonder how they swing that at the factory. Adding 20 hours to the build process can't be cheap.

The vertex 4 had 4 hours on the clock when I first put it in. So at least in the case of the vertex 4, it isn't so bad.