SSD Write Endurance 25nm Vs 34nm

[sergiu]

I can't seem to keep the vertex 4 out of storage mode. Secure wipe only helped for half an hour or so, then performance dropped again. I'll try a firmware reflash followed by another secure wipe. See if that helps.

Strange thing is that it ran fine for 3 days .....

If I can't get this working right, I may need to abandon breaking this V4 ... which would mean I would need to go find another victim in its place.

Try increasing the OP by 5% steps and see if it can survive longer. SLC based drives also need alot of overprovisioning to behave very fast. 128GB with 50% overprovisioning is like a 64GB SLC with little to nothing overprovisioning.

[canthearu]

It always perfoms fine in benchmarks ... benchmarks simply don't push enough writes to the drive.

You see my graphs above showing performance degredataion while running ASU ... the first 10gig or so of writes was always ultra-quick, even when the run was being slowed significantly by the storage mode later in each cycle

[canthearu]

My ASU runs always keep 5 gig free, and I am running a 61000meg parition.

But I'll try smaller partitions if I have problems again (I've just reflashed and secure erased, then ran WEI after creating partition to make sure windows knows it is an SSD and to use TRIM.

[Anvil]

My rig had frozen 2 hours ago, can't find what caused it, nothing in the EventLog.
I suspect it was caused by the X25-V as there were 0 files in the test folder, as in, all files were deleted and just about to start a new loop.

Both drives look to be OK, I'll keep an extra eye on the rig this evening.

[Anvil]

Intel 330 120GB

604.10TB Host writes
2.22TiB Host reads
Reallocated sectors : 05 26 +2
Available Reserved Space : E8 100
MWI 29
[B5] 13
[B6] 13
[F1] Total LBAs Written 19795323
[F2] Total LBAs Read 72844
[F9] Total NAND Writes 436080GB // ~426TiB
POH 1460
MD5 OK

125.29MiB/s on avg (~3 hours)

Once restarted, the 330 quickly threw a few reallocated sectors.

I330_MWI_vs_RSC_2012_07_24.png


--

Kingston SSDNow 40GB (X25-V)

1142.45TB Host writes (37435802*32)
Reallocated sectors : 05 65
Available Reserved Space : E8 98
POH 10195
MD5 OK

38.44MiB/s on avg (~3 hours)

--

Endurance_cr_20120724.png

[Christopher]

Only 40 more hours to 3PiB. That feels like a monumental number, but that would be the equivalent of a 64GB drive hitting 775TiB -- ie 3/4ths of a PiB.

[Christopher]

Samsung 830 256GB Day 126

(GiB) 3,098,472
(TiB) 3,025
(PiB) 2.97


(Avg) 299.06 MB/s

(B1) Wear Leveling Count: 13,422

(B6) Erase Fail Count: 9
(05) Reallocated Sectors: 36864
(POH) 3071


-------------------
The 830 actually has 3040TiB on it, so a little over 24 hours from now should be 3PiB

[Anvil]

About 14 hours later...

Intel 330 120GB

609.96TB Host writes
2.24TiB Host reads
Reallocated sectors : 05 31 +5
Available Reserved Space : E8 100
MWI 29
[B5] 15
[B6] 16
[F1] Total LBAs Written 19987316
[F2] Total LBAs Read 73298
[F9] Total NAND Writes 440309GB // ~430TiB
POH 1474
MD5 OK

125.21MiB/s on avg (~17 hours)

This time, B5 is not equal to B6.

--

Kingston SSDNow 40GB (X25-V)

1144.07TB Host writes (37488837*32)
Reallocated sectors : 05 65
Available Reserved Space : E8 98
POH 10209
MD5 OK

35.39MiB/s on avg (~17 hours)

--

[B.A.T]

Here are todays update:

m4
1276.6606 TiB
5320 hours
Avg speed 74.35 MiB/s.
AD 144 to 137
P/E 21954.
C3 4457
CE 58
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
818.04 TiB
24607 hours
Reallocated sectors : 153 to 201
Available Reserved space: 44 to 41
MWI= 152 to 135
MD5 =OK
31.07 MiB/s on avg


Intel X25-E 64GB
572.00 TiB
1788-30=1758 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 96
MD5 =OK
91.22 MiB/s on avg


Mtron Pro 7025 32GB
Paused until new firmware is updated.

[Christopher]

Samsung 830 256GB Day 127

(GiB) 3,122,132
(TiB) 3,048
(PiB) 3.00


(Avg) 298.61 MB/s

(B1) Wear Leveling Count: 13,524

(B6) Erase Fail Count: 9
(05) Reallocated Sectors: 36864
(POH) 3094


-------------------

[canthearu]

Reallocated sectors : 05 31 +5

Never saw a sandforce drive degrade gracefully like this. Maybe the sandforce controller works better with crappy NAND :-)

[Anvil]

It's behaving nicely

I'll make sure to check what chips are used on both drives. (eventually)

Intel 330 120GB

620.21TB Host writes
2.26TiB Host reads
Reallocated sectors : 05 33 +2
Available Reserved Space : E8 100
MWI 27
[B5] 16
[B6] 17
[F1] Total LBAs Written 20323087
[F2] Total LBAs Read 74149
[F9] Total NAND Writes 447708GB // ~437TiB
POH 1497
MD5 OK

125.20MiB/s on avg (~41 hours)

--

Kingston SSDNow 40GB (X25-V)

1146.79TB Host writes (37578229*32)
Reallocated sectors : 05 65
Available Reserved Space : E8 98
POH 10233
MD5 OK

34.12MiB/s on avg (~41 hours)

--

[canthearu]

Sandisk Extreme G25 - 120gb - Day 51

Drive hours: 1221
ASU GiB written: 527,275.97 GiB (514.9179.83 TiB)
Avg MB/s: 125.81 MB/s (7.19 hours)
MD5: OK

Host GB written (F1): 531,263 GiB (518.81 TiB)
NAND writes (E9): 396,545 GiB (387.25 TiB)

Retired Block Count (05): 4 raw
Failure count (AB, AC): 1 program, 1 erase
Raw Error Rate (01): 100 normalized
Media Wearout Indicator (E7): 10 normalized
Wear Range Delta (B1): 4 raw
Reported Uncorrectable Errors (BB): 0 raw

Vertex 4 - 50% OP - 128gb - PAUSED

Drive hours: 101
ASU GiB written: 62,800.31 GiB (61.33 TiB)
Avg MB/s: 0 MB/s (0 hours)
MD5: OK

Host GB written (E8): 63,967.47 GiB (62.47 TiB, 134149509504 raw)

Raw Read Error Rate (01): 5 raw
Reallocated Block Count (05): 0 raw
Remaining Life (E9): 84 normalised

My last attempt to get the vertex 4 working in performance mode only failed after 20 minutes or so. I also tried increasing the time between runs to 30 seconds to give the drive more recovery time, but that didn't help.

I am not testing with a 50,000 Megabyte partition, which is about what an enterprise focused SSD would get on a 64gig SLC drive.

[Christopher]

It's too bad about the V4. I honestly think the 128 is one of the most interesting drives on the market, but not giving end users a choice to stick with pre 1.4fw if they so choose is a bit off-putting. 1.4+ is not for everyone IMHO.

[Anvil]

Both drives produced a few reallocations.

Intel 330 120GB

623.95TB Host writes
2.27TiB Host reads
Reallocated sectors : 05 35 +2
Available Reserved Space : E8 100
MWI 27
[B5] 17
[B6] 18
[F1] Total LBAs Written 20445743
[F2] Total LBAs Read 74490
[F9] Total NAND Writes 450411GB // ~440TiB
POH 1506
MD5 OK

125.15MiB/s on avg (~49 hours)

I330_MWI_vs_RSC_2012_07_26.png

--

Kingston SSDNow 40GB (X25-V)

1147.79TB Host writes (37610723*32)
Reallocated sectors : 05 67 +2
Available Reserved Space : E8 98
POH 10241
MD5 OK

33.95MiB/s on avg (~49 hours)

--

Endurance_cr_20120726.png

Endurance_PEsummary_20120726.png

[B.A.T]

Here are todays update:

m4
1282.7254 TiB
5344 hours
Avg speed 74.38 MiB/s.
AD 137 to 133
P/E 22056.
C3 4457
CE 58
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
820.52 TiB
24631 hours
Reallocated sectors : 201 226
Available Reserved space: 41 to 40
MWI= 135 128
MD5 =OK
31.00 MiB/s on avg


Intel X25-E 64GB
579.46 TiB
1812-30=1782 hours
Reallocated sectors : 0
Available Reserved space: 100 to 99
MWI= 96
MD5 =OK
91.26 MiB/s on avg


Mtron Pro 7025 32GB
Paused until new firmware is updated.

[Hopalong X]

Never saw a sandforce drive degrade gracefully like this. Maybe the sandforce controller works better with crappy NAND :-)

Maybe Sandforce controllers work better with properly written FW.

[Kain665]

Maybe Sandforce controllers work better with properly written FW.

You should see what happened to the 520.

[Demo]

1000Tb of host writes seems a lot!

I wonder if the Corsair Force GT 240Gb can handle that kind of a write endurance?

[Anvil]

Intel 330 120GB

634.53TB Host writes
2.30TiB Host reads
Reallocated sectors : 05 43 +8
Available Reserved Space : E8 100
MWI 26
[B5] 21
[B6] 22
[F1] Total LBAs Written 20792174
[F2] Total LBAs Read 75341
[F9] Total NAND Writes 458046GB // ~447TiB
POH 1530
MD5 OK

125.21MiB/s on avg (~74 hours)

--

Kingston SSDNow 40GB (X25-V)

1150.59TB Host writes (37702684*32)
Reallocated sectors : 05 67
Available Reserved Space : E8 98
POH 10266
MD5 OK

33.70MiB/s on avg (~74 hours)

--

[Christopher]

Samsung 830 256GB Day 128

(GiB) 3,159,768
(TiB) 3,085
(PiB) 3.03


(Avg) 298.40 MB/s

(B1) Wear Leveling Count: 13,686

(B6) Erase Fail Count: 9
(05) Reallocated Sectors: 36864
(POH) 3130


-------------------

[canthearu]

Bah, my vertex 4 is still not performing right.

Will decide whether to just kill it as is, or just keep it for something else.

[canthearu]

It's too bad about the V4. I honestly think the 128 is one of the most interesting drives on the market, but not giving end users a choice to stick with pre 1.4fw if they so choose is a bit off-putting. 1.4+ is not for everyone IMHO.

Yeah, a 1.3.bugfixed firmware would have been good ... but I think the vertex 4 is solely focused on write benchmarks at all costs. It kinda has tunnel vision to other ssd requirements.

The plextor/liteon drives are a much better buy at the moment for performance.

[canthearu]

1000Tb of host writes seems a lot!

I wonder if the Corsair Force GT 240Gb can handle that kind of a write endurance?

Your drive is rated for somewhere between 600 and 800 TiB writes. (depending on what data you use)

But if you were to keep using it past that point, I imagine you would get to 1000TiB without any real problem.

[Christopher]

I still have an unopened Vertex Turbo 64GB laying around somewhere, maybe the third time is the charm?

...Probably not.