SSD Write Endurance 25nm Vs 34nm

[canthearu]

Intel 520 60GB - Day 117

Drive hours: 2,753
ASU GiB written: 878,292.56 GiB (857.70 TiB)
Avg MB/s: 93.37 MB/s (3 hours)
MD5: OK (verified off disk now)

Host GB written (F1): 884,428.22 GiB (863.70 TiB, 28301703 raw)
NAND writes (F9): 626,106 GiB (611.43 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 116 normalized
Available Reserved Space (AA): 100 normalized
Media Wearout Indicator (E9): 1

Sandisk Extreme G25 - 120gb - Day 17

Drive hours: 408
ASU GiB written: 173,413.05 GiB (169.35 TiB)
Avg MB/s: 123.36 MB/s (3 hours)
MD5: OK (verified off disk now)

Host GB written (F1): 174,517 GiB (170.43 TiB)
NAND writes (E9): 130,157 GiB (127.11 TiB)

Retired Block Count (05): 4 raw
Failure count (AB, AC): 1 program, 1 erase
Raw Error Rate (01): 116 normalized
Media Wearout Indicator (E7): 67 normalized
Wear Range Delta (B1): 4 raw
Reported Uncorrectable Errors (BB): 0 raw

Edit: The Sandisk extreme MWI value seems to be tracking very close to the MWI on the Corsair Force 3. The sandisk seems to be controlling wear range delta much better though ... which could be an advantage at endgame!

http://www.xtremesystems.org/forums/...=1#post4967340

[Anvil]

Ah, moving the source file to another drive has fixed the md5 testing

I will post an update in a few hours ... let speeds somewhat stabalize again.!

Thanks for letting me know!

It should work even if the file is on the same drive, it will just not copy the file.
Give it another shot with the new build. (should be available this weekend, most likely Sunday)

--

Intel 330 120GB

274.10TB Host writes
1.02TiB Host reads
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 69
[F1] Total LBAs Written 8981645
[F2] Total LBAs Read 32467
[F9] Total NAND Writes 197849GB
POH 697
MD5 OK

123.55MiB/s on avg (~70 hours)

--

Kingston SSDNow 40GB (X25-V)

1053.85TB Host writes (34532729*32)
Reallocated sectors : 05 48
Available Reserved Space : E8 99
POH 9427
MD5 OK

33.18MiB/s on avg (~65 hours)

--

Endurance_cr_20120622.png

And again, the Samsung goes through the roof.

[Christopher]

Both the 830 and M3P are, inexplicably, slightly faster. Not on average (those are higher too) but instantaneous speeds are faster. Crazy.

[canthearu]

The Intel 520 is dead at day 118

Drive just panic locked without warning at 9:43am on the 23/6/2012. Attempts to resurrect it by plugging it into other computers have failed.

Final Stats:

Drive hours: 2,765
Host GB written (F1): 888,488.78 GiB (867.66 TiB, 28431641 raw)
NAND writes (F9): 628,987 GiB (614.25 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 116 normalized
Available Reserved Space (AA): 100 normalized
Media Wearout Indicator (E9): 1

While I did expect a bit more from the Intel 520, I guess this isn't a terrible result.

[Zaxx]

Damn...I had hoped it'd join the 1PB club.

I'd be nice if Intel could analyze the corpse and see why the controller choked. It sure wasn't the nand...lol

[Christopher]

Damn, canthearu, that's unfortuanate. And disappointing. You'd think this being the number one cause of death, more manufacturer resources would be devoted to making sure it doesn't happen.

----------------------------------------------------
Samsung 830 256GB Day 96

(GiB) 2,349,189.01
(TiB) 2 294.13

(Avg) 295.06 MB/s

(B1) Wear Leveling Count: 10,192

(B6) Erase Fail Count: 7
(05) Reallocated Sectors: 28672
(POH) 2316

Plextor M3P 128GB Day 25

(GiB) 649,620.10
(TiB) 634.39

(Avg) 276.74 MB/s
(POH) 666

-------------------------------------------------------
Averages are still gradually creeping higher.

[AMD_Freak]

Both the 830 and M3P are, inexplicably, slightly faster. Not on average (those are higher too) but instantaneous speeds are faster. Crazy.

I have seen other M3P get faster as they get some wear on them, might be just now getting broke in

[Christopher]

The speed increase from "break in" isn't probably what you're seeing with the M3P. You have to put a lot of wear on a drive before it's NAND's characteristics start to change, several thousand PE cycles. However, some drives also tune for workload. While it's uncommon with consumer drives, enterprise drives do this frequently.

[Anvil]

Intel 330 120GB

292.06TB Host writes
1.07TiB Host reads
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 67
[F1] Total LBAs Written 9570306
[F2] Total LBAs Read 34942
[F9] Total NAND Writes 210815GB
POH 739
MD5 OK

123.57MiB/s on avg (~112 hours)

--

Kingston SSDNow 40GB (X25-V)

1058.69TB Host writes (34691177*32)
Reallocated sectors : 05 48
Available Reserved Space : E8 99
POH 9469
MD5 OK

33.16MiB/s on avg (~107 hours)

--

B.A.T updated on the Norwegian site so the figures are more or less up to date.

Endurance_cr_20120624.png

Endurance_PEsummary_20120622.png


The 520 did fine, I did expect some kind of warning though, a few disconnects or some other sign.
Well, it sets the bar for my 330, 2x of the 520 is acceptable.
The Force 3 had async NAND and a poor firmware, I believe the Force 3 would have done better if there was a solid fw from the start.

[Christopher]

Samsung 830 256GB Day 97

(GiB) 2,366,892.23
(TiB) 2 311.42

(Avg) 294.86 MB/s

(B1) Wear Leveling Count: 10,269

(B6) Erase Fail Count: 7
(05) Reallocated Sectors: 28672
(POH) 2334

Plextor M3P 128GB Day 26

(GiB) 666,775.26
(TiB) 651.15

(Avg) 276.90MB/s
(POH) 684

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

[B.A.T]

R.I.P Intel 520. A little early to die at only 600TiB of nand write.

Since XS was down last time I updated I'll just skip to the latest numbers.

Todays update:
m4
1095.5206 TiB
4592 hours
Avg speed 73.65 MiB/s.
AD 251 to 239
P/E 18871.
C3 3323
CE 58
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
735.32 TiB
23883 hours
Reallocated sectors : 200 to 231
Available Reserved space: 73 to 70
MWI= 42 to 20
MD5 =OK
39.62 MiB/s on avg


Intel X25-E 64GB
344.34 TiB
1064-30=1034 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 97
MD5 =OK
90.98 MiB/s on avg


ASU write 64.6592 TiB
Host write 121,0753 TiB
Total host write 168,6109 TiB
Day 21
MWI= 95
MD5 =OK
46.63 MiB/s on avg


The Mtron has dropped out one time this morning but is OK now. The G1 is still losing reserve space but still looks strong. I also noted that CDI latest ver has added some information on the m4.
C3 has gone from unknown to Cumulative Corrected ECC and will help us to know the status of the nand. Looking back this value started increasing at 500 TiB but had the largest increase after the last retention test.

[Christopher]

Both drives are just a little faster

[Anvil]

Intel 330 120GB

301.50TB Host writes
1.28TiB Host reads
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 66
[F1] Total LBAs Written 9879456
[F2] Total LBAs Read 41807
[F9] Total NAND Writes 217629GB
POH 761
MD5 OK

123.75MiB/s on avg (~12 hours)

--

Kingston SSDNow 40GB (X25-V)

1061.34TB Host writes (34778194*32)
Reallocated sectors : 05 48
Available Reserved Space : E8 99
POH 9492
MD5 OK

36.49MiB/s on avg (~12 hours)

[canthearu]

Sandisk Extreme G25 - 120gb - Day 20

Drive hours: 478
ASU GiB written: 203,794.93 GiB (199.02 TiB)
Avg MB/s: 122.44 MB/s (56.11 hours)
MD5: OK (verified off disk now)

Host GB written (F1): 205,050 GiB (200.24 TiB)
NAND writes (E9): 152,940 GiB (149.36 TiB)

Retired Block Count (05): 4 raw
Failure count (AB, AC): 1 program, 1 erase
Raw Error Rate (01): 112 normalized
Media Wearout Indicator (E7): 61 normalized
Wear Range Delta (B1): 4 raw
Reported Uncorrectable Errors (BB): 0 raw

[B.A.T]

Todays update:

m4
1101.3860 TiB
4615 hours
Avg speed 74.21 MiB/s.
AD 239 to 236
P/E 18970.
C3 3323
CE 58
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
738.09 TiB
23906hours
Reallocated sectors : 231 to 255 to 86
Available Reserved space: 70 to 63
MWI= 20
MD5 =OK
34.81 MiB/s on avg


Intel X25-E 64GB
351.83TiB
1087-30=1057 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 97
MD5 =OK
91.80 MiB/s on avg


Mtron Pro 7025 32GB
ASU write 67.3465 TiB
Host write 126,2116 TiB
Total host write 173,7471 TiB
Day 22
MWI= 95
MD5 =OK
48.03 MiB/s on avg


The Mtron dropped out this afternoon but after a restart is OK again. What's more interesting is the X25-M G1. A huge round of reallocated sectors and a 7% drop decrease in spare area makes me believe that it'll not reach 1 PiB at this rate.

[Christopher]

Samsung 830 256GB Day 98

(GiB) 2,401,525.13
(TiB) 2 345.24

(Avg) 295.01 MB/s

(B1) Wear Leveling Count: 10,418

(B6) Erase Fail Count: 8
(05) Reallocated Sectors: 32768
(POH) 2367

Plextor M3P 128GB Day 27

(GiB) 699,695.62
(TiB) 683.30

(Avg) 277.22MB/s
(POH) 715

------------------------------------------------------
Finally, the 830 loses another block. New count is 8, up from 7.-

[kaktus1907]

AFAICR first reallocted sectors did occur at ~1.180GiB on Samsung 830 256GB, one would expect The Plextor behave similar at roughly 600GiB for a 128GB drive, but fortunately there is no sign of it.. I think it's more important living long time with bad blocks, handiling them etc than bad block amount.. But it still a good thing for The Plextor for now, considering it writes like a SLC drive, right

[Conquistador SW]

Stupid me does not really understand what I am seeing here and I have some questions if that is ok.

What is a sector? It seems a block on the 830 contains 4096 sectors. Are these bytes? I assume things can differ between NAND chips, so can you compare "reallocated sector" on the 830 with the intel G1 (why did the count suddenly go down??) ?

[Anvil]

Intel 330 120GB

311.54TB Host writes
1.30TiB Host reads
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 65
[F1] Total LBAs Written 102086727
[F2] Total LBAs Read 42658
[F9] Total NAND Writes 224878GB
POH 785
MD5 OK

123.86MiB/s on avg (~35 hours)

--

Kingston SSDNow 40GB (X25-V)

1064.05TB Host writes (34866899*32)
Reallocated sectors : 05 48
Available Reserved Space : E8 99
POH 9515
MD5 OK

34.32MiB/s on avg (~35 hours)

--

Endurance_cr_20120626.png

I'll update the other stats tonight.

[canthearu]

Stupid me does not really understand what I am seeing here and I have some questions if that is ok.

What is a sector?

A sector is 512 bytes. An erase block is 2 meg in size, so each block contains 4096 of the 512 byte sectors.

Whenever the samsung controller sees an erase failure, it marks the entire erase block as bad, so the bad block count increases 4096.

It seems a block on the 830 contains 4096 sectors. Are these bytes? I assume things can differ between NAND chips, so can you compare "reallocated sector" on the 830 with the intel G1 (why did the count suddenly go down??) ?

Not sure how comparable it is to the G1.

The reason why the G1 reallocated sector count went down was because the counter is only able to store a number between 0-255. Intel made a mistake putting the reallocated sector count inside the normalized value rather then the raw smart value. So when the normalized value reaches 255 and is incremented again, it loops back to 0.

[Conquistador SW]

Thanks! So these 8 blocks that failed (16MB) is not that much if I understand correctly.

[Christopher]

Thanks! So these 8 blocks that failed (16MB) is not that much if I understand correctly.

No, they're not really all that much. The 830 has over 17GB of spare area, so it's got a lot of blocks to spare.

[B.A.T]

Todays update:

m4
1107.2110 TiB
4639 hours
Avg speed 74.55 MiB/s.
AD 236 to 233
P/E 19068.
C3 3323
CE 58
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
741.09 TiB
23929 hours
Reallocated sectors : 86 to 91
Available Reserved space: 63
MWI= 20 to 12
MD5 =OK
38.53 MiB/s on avg


Intel X25-E 64GB
359.00 TiB
1110-30=1080 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 97
MD5 =OK
92.01 MiB/s on avg


Mtron Pro 7025 32GB
ASU write 69.0235 TiB
Host write 129,4164 TiB
Total host write 176,9520 TiB
Day 23
MWI= 95
MD5 =OK
48.20 MiB/s on avg


The Mtron did its usual thing and dropped out early this morning. Runs normal now.

[Christopher]

The M3P had it's first reallocation event today.

[kaktus1907]

amusing, it seems to me that the plextor was waiting for me to pop some reallocation..
at least i learnt what BB, B2 SMART ID's stands for for my CPP.. for example it lacks C4 and there was no way to see other than 05..