SSD Write Endurance 25nm Vs 34nm

[Christopher]

Samsung 830 256GB Day 134

(GiB) 3,308,292
(TiB) 3,230
(PiB) 3.17


(Avg) 297.92 MB/s

(B1) Wear Leveling Count: 14,328

(B6) Erase Fail Count: 9
(05) Reallocated Sectors: 36864
(POH) 3275


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

[Anvil]

Intel 330 120GB

698.40TB Host writes
2.46TiB Host reads
Reallocated sectors : 05 118 +18
Available Reserved Space : E8 100
MWI 18
[B5] Program fail count 58
[B6] Erase fail count 60
[F1] Total LBAs Written 22885243
[F2] Total LBAs Read 80556
[F9] Total NAND Writes 504156GB // ~492TiB
POH 1677
MD5 OK

125.42MiB/s on avg (~48 hours)

--

Kingston SSDNow 40GB (X25-V)

1167.76TB Host writes (38265203*32)
Reallocated sectors : 05 69 +1
Available Reserved Space : E8 98
POH 10414
MD5 OK

33.69MiB/s on avg (~48 hours)

--

Endurance_cr_20120802.png

--

Not sure what to think about the 330, it's accumulating a lot of reallocated sectors, the Force 3 120GB had just 2 when it died but it did suffer from bad FW, most of the time that is.
The 830 looks to handle lots of reallocation's so there is still hope for the 330.

[Christopher]

The 830 looks to handle lots of reallocation's so there is still hope for the 330.

I don't know whether it does or not -- it's only got 9, totalling 18MB. It might not handle reallocations as gracefully as the X25-V.

[canthearu]

Some drives, like the vertex 4/octane and samsung, record whole erase blocks bad when they are found.

Other drives, like the sandforce drives, only record individual pages when they go bad ... for example, on the Intel 330, each page is 8kb, so when a write/erase error occurs on a block, and the page is marked bad, the bad block count increase by 2 (as each page is 2x4kb blocks)

[Christopher]

Some drives, like the vertex 4/octane and samsung, record whole erase blocks bad when they are found.

Other drives, like the sandforce drives, only record individual pages when they go bad ... for example, on the Intel 330, each page is 8kb, so when a write/erase error occurs on a block, and the page is marked bad, the bad block count increase by 2 (as each page is 2x4kb blocks)

If there is an erase failure on a SandForce it doesn't mark the whole erase block as bad? Just one page? I assume if there is one bad page in an erase block, it would just dump the block, not just a page. That's kinda wierd, but then the 830 hasn't had anything but erase failures.

[Anvil]

Intel 330 120GB

703.51TB Host writes
2.47TiB Host reads
Reallocated sectors : 05 133 +15
Available Reserved Space : E8 99 -1
MWI 17
[B5] Program fail count 65
[B6] Erase fail count 68
[F1] Total LBAs Written 23052662
[F2] Total LBAs Read 81020
[F9] Total NAND Writes 507849GB // ~496TiB
POH 1689
MD5 OK

125.34MiB/s on avg (~60 hours)

--

Kingston SSDNow 40GB (X25-V)

1169.12TB Host writes (38309652*32)
Reallocated sectors : 05 71 +2
Available Reserved Space : E8 98
POH 10426
MD5 OK

33.59MiB/s on avg (~60 hours)

--

@christopher
Isn't [05] Reallocated sectors (in your case 36864) listed in SMART or is that calculated?

In any case, it's on a roll, it's steadily increasing and triggered a movement on [E9] Available Reserved Space

[Christopher]

Anvil -- reallocated sectors are tracked in SMART data. 1 reallocation event = one 2MB erase block, or 4096 512b sectors. So 9 erase block failures equals 18MB reallocated, or 36864 sectors. The M3P went through almost 2GB of reallocations.

[Computurd]

Today JEDEC released some Standard SSD Endurance Workloads that they feel are relevant for endurance testing.

The enterprise endurance workload shall be comprised of random data with the following payload size distribution:
512 bytes (0.5K) 4%
1024 bytes (1K) 1%
1536 bytes (1.5K) 1%
2048 bytes (2K) 1%
2560 bytes (2.5K) 1%
3072 bytes (3K) 1%
3584 bytes (3.5K) 1%
4096 bytes (4K) 67%
8192 bytes (8K) 10%
16,384 bytes (16K) 7%
32,768 bytes (32K) 3%
65,536 bytes (64K) 3%

This is just a bit of it, very detailed. reading now.

[Christopher]

Samsung 830 256GB Day 136

(GiB) 3,375,642
(TiB) 3,296
(PiB) 3.24


(Avg) 298.04 MB/s

(B1) Wear Leveling Count: 14,619

(B6) Erase Fail Count: 9
(05) Reallocated Sectors: 36864
(POH) 3339


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

[Anvil]

It's all downhill for the 330, I'd be surprised if it last another week.

Intel 330 120GB

724.63TB Host writes
2.52TiB Host reads
Reallocated sectors : 05 199 +66
Available Reserved Space : E8 69 -30
MWI 14
[B5] Program fail count 96
[B6] Erase fail count 103
[F1] Total LBAs Written 23744862
[F2] Total LBAs Read 82722
[F9] Total NAND Writes 523106GB // ~511TiB
POH 1737
MD5 OK

125.39MiB/s on avg (~109 hours)

--

Kingston SSDNow 40GB (X25-V)

1174.70TB Host writes (38492805*32)
Reallocated sectors : 05 71
Available Reserved Space : E8 98
POH 10474
MD5 OK

33.39MiB/s on avg (~109 hours)

[Anvil]

Today JEDEC released some Standard SSD Endurance Workloads that they feel are relevant for endurance testing.

This is just a bit of it, very detailed. reading now.

Thanks for the link, will read asap

[kensiko]

Kensiko...this thread is dedicated to endurance testing, maybe you should post in the hardware-storage section.

Sorry, I had to ask somewhere

I'll look in the hardware section.

[Anvil]

Intel 330 120GB

734.08TB Host writes
2.55TiB Host reads
Reallocated sectors : 05 228 +29
Available Reserved Space : E8 58 -11
MWI 13
[B5] Program fail count 110
[B6] Erase fail count 118
[F1] Total LBAs Written 24054404
[F2] Total LBAs Read 83516
[F9] Total NAND Writes 529922GB // ~517TiB
POH 1759
MD5 OK

125.42MiB/s on avg (~131 hours)

--

Kingston SSDNow 40GB (X25-V)

1177.20TB Host writes (38574684*32)
Reallocated sectors : 05 72 +1
Available Reserved Space : E8 98
POH 10496
MD5 OK

33.35MiB/s on avg (~131 hours)

[B.A.T]

Here are todays update:

m4
1352.2965 TiB
5616 hours
Avg speed 74.59 MiB/s.
AD 112 to 94
P/E 23236.
C3 4457
CE 58
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
846.46 TiB
24902 hours
Reallocated sectors : 115 to 255 to 8
Available Reserved space: 31 to 22
MWI= 104
MD5 =OK
27.05 MiB/s on avg


Intel X25-E 64GB
665.36 TiB
2083-30=2053 hours
Reallocated sectors : 2 to 7
Available Reserved space: 99
MWI= 95
MD5 =OK
91.73 MiB/s on avg


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

The G1 continues its path to death and the E had 5 more reallocated sectors. The m4 just steams on

[Christopher]

It would appear that TRIM and 'modern' firmware are worth their weight in gold. Imagine the X25-E and G1 with TRIM -- they'd be forever in the killing. And the G1 has already made it quite a ways.

[Zaxx]

Very good point Christopher. Although we're talking a totally different animal, seeing these trimless drives w/o GC doing this well sustains my confidence that my 60GB Ocz Apex with 51nm nand may live an exceptionally long life, all things considered. Biggest gripe I have is that smart only has one attribute, the temp which is locked at 44C. No lifetime reads/writes, no POH, no MWI...no nothing, doesn't even support SE! Guess that's the price paid for an internally raided gen1 SSD. Gotta give it props tho...after consolidating free space and a run of AS_Cleaner 1.0 (aka FreeSpaceCleaner) w/FF, dubbed 'Tony Trim' about once a month it still benches just as fast as the day I bought it.

[Anvil]

The system had halted 30 minutes within the latest report, found out 6 hours later.

Intel 330 120GB

741.81TB Host writes
2.57TiB Host reads
Reallocated sectors : 05 260 +32
Available Reserved Space : E8 46 -12
MWI 12
[B5] Program fail count 125
[B6] Erase fail count 135
[F1] Total LBAs Written 24307515
[F2] Total LBAs Read 84087
[F9] Total NAND Writes 535502GB // ~523TiB
POH 1776
MD5 OK

124.82MiB/s on avg (~18 hours)

--

Kingston SSDNow 40GB (X25-V)

1179.37TB Host writes (38645807*32)
Reallocated sectors : 05 72
Available Reserved Space : E8 98
POH 10514
MD5 OK

35.04MiB/s on avg (~18 hours)

[canthearu]

Sandisk Extreme G25 - 120gb - Day 63

Drive hours: 1446
ASU GiB written: 625,108.60 GiB (610.46 TiB)
Avg MB/s: 124.05 MB/s (23.08 hours)
MD5: OK

Host GB written (F1): 629,613 GiB (614.86 TiB)
NAND writes (E9): 469,999 GiB (458.98 TiB)

Retired Block Count (05): 4 raw
Failure count (AB, AC): 1 program, 1 erase
Raw Error Rate (01): 108 normalized
Media Wearout Indicator (E7): 10 normalized
Wear Range Delta (B1): 3 raw
Reported Uncorrectable Errors (BB): 0 raw

I did a retention test for 3.5 days. Drive didn't complain! Drive might outlive the Intel 330, which I really wasn't expecting :P

I might do another drive (maybe a cheap 60gig sandforce) after 5.0.3 firmware is released with the TRIM fix. Could be interesting to see if the drive reverts to the old wear range delta behaviour

[Christopher]

I might do another drive (maybe a cheap 60gig sandforce) after 5.0.3 firmware is released with the TRIM fix. Could be interesting to see if the drive reverts to the old wear range delta behaviour

My 60GB Chronos Deluxe 60GB 2281/32nm Toggle on 3.3.0 and then 3.3.2 never went above 11 I think. But due to the 5K NAND, I think the WRD acted differently. Like it allowed a larger percentage of PE cycle deltas than 3K flash or something. Either that, or it just wear leveled better.

[canthearu]

My 60GB Chronos Deluxe 60GB 2281/32nm Toggle on 3.3.0 and then 3.3.2 never went above 11 I think. But due to the 5K NAND, I think the WRD acted differently. Like it allowed a larger percentage of PE cycle deltas than 3K flash or something. Either that, or it just wear leveled better.

No, I'm absolutely certain your 60gig Chronos Deluxe drive was programmed with 3K cycles, not 5K cycles, from where it landed at MWI=10.

In any case, the Sandisk extreme hasn't done worse then 4 WRD in it's entire life. That is a huge improvement over the old Corsair Force 3, and even your 60gig Chronos Deluxe

[Christopher]

Samsung 830 256GB Day 137

(GiB) 3,428,092
(TiB) 3,347
(PiB) 3.29


(Avg) 297.86 MB/s

(B1) Wear Leveling Count: 14,845

(B6) Erase Fail Count: 9
(05) Reallocated Sectors: 36864
(POH) 3390


-----------------------------------
canthearu -- I don't know what the FW was programmed with for certain, but the NAND was 5K rated. Not that it matters a whole lot, but it could just be that Mushkin didn't bother changing the FW for 5K and 3K models (Chronos and Chronos Deluxe). OCZ did the same thing with the 1.6FW for Indilinx.

[Anvil]

[B]
...
I did a retention test for 3.5 days. Drive didn't complain! Drive might outlive the Intel 330, which I really wasn't expecting :P

I might do another drive (maybe a cheap 60gig sandforce) after 5.0.3 firmware is released with the TRIM fix. Could be interesting to see if the drive reverts to the old wear range delta behaviour

I expected the 330 to outlast my Force 3, it doesn't look like that is going to happen.

No signs of slowdowns during benchmarks though.

INTEL SSDSC2CT120A3_120GB_1GB-20120807-1908.png

...the rig had rebooted, frequent reboots and or frozen systems are signs of a drive going bad.

[Anvil]

Here's the 330 as of 12 hours ago.
I330_MWI_vs_RSC_2012_08_07_2343.png

and finally an update.
Endurance_cr_20120807.png

[Anvil]

Intel 330 120GB

749.94TB Host writes
2.60TiB Host reads
Reallocated sectors : 05 305 +45
Available Reserved Space : E8 27 -19
MWI 11
[B5] Program fail count 147
[B6] Erase fail count 158
[F1] Total LBAs Written 24574169
[F2] Total LBAs Read 85096
[F9] Total NAND Writes 541377GB // ~529TiB
POH 1800
MD5 OK

125.84MiB/s on avg (~14 hours)

--

Kingston SSDNow 40GB (X25-V)

1181.71TB Host writes (38722319*32)
Reallocated sectors : 05 72
Available Reserved Space : E8 98
POH 10534
MD5 OK

36.93MiB/s on avg (~14 hours)

[Anvil]

It's getting close...

HDD_PROBLEM_CDI.PNG

HDD_PROBLEM.PNG