SSD Write Endurance 25nm Vs 34nm

[devsk]

Ha Ha... it's time to get the 320 switched out under warranty.

And tell them it was only sparingly used....

[Hopalong X]

"All I did was plug it in."

[One_Hertz]

I have no idea what happened! I bought this shiny new SSD and it is dead in less than 3 months! I thought SSDs were supposed to be reliable?!

[devsk]

I have no idea what happened! I bought this shiny new SSD and it is dead in less than 3 months! I thought SSDs were supposed to be reliable?!

And then they look at the SMART attributes...

[Hopalong X]

"Well it was that way when I got it."

"All I did was plug it in and it wouldn't work."

[Christopher]

I have a special tool that makes reading smart info difficult.

It's called a hammer!

Take that, Intel!


Dear Intel,

I regret to inform you that your 320 series of Solid State Drives are terrible.

It couldn't even last four months in my Chamber of Doom^tm.

What the hell are you guys doing over there?? This isn't rocket science, people.


Much love,
One_Hertz

[johnw]

I doubt he has it yet. I didn't overnight it. It is going by USPS priority mail, which I don't think gets particularly expedited through Customs. I shipped it on Oct 6, but as of Oct 9, it had only made it to LA.

"Your item is being processed by customs in CANADA at 6:49 pm on October 12, 2011."

I love how specific the USPS tracking information is. It is in "CANADA". That really pinpoints its location! At least it is in the right country...

[Ao1]

Here is a graph that shows the relationship between WRD (B1) & MWI (E7) for each of the SF drives being tested (excluding Vapor’s as there is not enough info). Anvil’s drive is holding up significantly better than SynbiosVyse & Christopher’s drives.

On SynbiosVyse’s drive the B1 value is more or less a perfect inverse of E7. I don’t know what NAND is being used in each drive, but it looks like SynbiosVyse & Christopher have NAND with 3K P/E specs and Anvil has 5K. It looks like the “disconnects” that Christopher experienced allowed better wear levelling, which has decreased the MWI burn out rate in comparison to SynbiosVyse's drive.

I can’t help think that SynbiosVyse’s drive failed due to the difference between least and most worn blocks, which resulted in an insufficient reserve block count. I think that is why the drive failed, but why, with hardly any static data, was SynbiosVyse’s drive unable to level out the wear.

@ Anvil, a summary in the first post of each drive being tested along with the NAND P/E specs would be really helpful. Also keys findings and calculation methods would be great.

SynbiosVyse WRD (B1) & MWI (E7)


Anvil/SynbiosVyse/ Christopher

[Ao1]

Everything is still going ok, MD5 is passing, just windows started ing at me (most likely due to the reserve space SMART attribute going below threshold).

At least it failed gracefully. That gives me a bit more faith about exceeding the MWI. Will it scape 500TB before it pegs out?

[B.A.T]

Here are yesterdays update because xs was down.:
m4
657.6115 TiB
2382 hours
Avg speed 90.81 MiB/s.
AD gone from 236 to 231.
P/E 11432.
MD5 OK.
Still no reallocated sectors


Kingston V+100
148.1625 TiB
585 hours
Avg speed 76.52 MiB/s
AD gone from 54 to 48.
P/E?
MD5 OK.
Still no reallocated sectors

[Anvil]

@ Anvil, a summary in the first post of each drive being tested along with the NAND P/E specs would be really helpful. Also keys findings and calculation methods would be great.

Thanks for the chart!

I'll try to gather what I can find this weekend


Kingston SSDNow 40GB (X25-V)

389.68TB Host writes
Reallocated sectors : 11
MD5 OK

34.57MiB/s on avg (~26 hours)

Force GT "up time" ~35 hours.

--

Corsair Force 3 120GB

01 90/50 (Raw read error rate)
05 2 (Retired Block count)
B1 57 (Wear range delta)
E6 100 (Life curve status)
E7 59 (SSD Life left)
E9 160317 (Raw writes)
F1 213510 (Host writes)

104.17MiB/s on avg (~61 hours)

power on hours : 635

[bulanula]

Really interested to see the results of the Samsung.

[bluestang]

M225->Vertex Turbo 64GB Update: Milestone reached...

488.58 TiB (537.20 TB) total
1290.04 hours
10005 Raw Wear
119.06 MB/s avg for the last 20.48 hours (on W7 x64)
MD5 OK
C4-Erase Failure Block Count (Realloc Sectors) from 7 to 8.
(Bank 6/Block 2406; Bank 3/Block 3925; Bank 0/Block 1766; Bank 0/Block 829; Bank 4/Block 3191; Bank 7/Block 937; Bank 7/Block 1980; Bank 7/Block 442)

[Christopher]

Here is a graph that shows the relationship between WRD (B1) & MWI (E7) for each of the SF drives being tested (excluding Vapor’s as there is not enough info). Anvil’s drive is holding up significantly better than SynbiosVyse & Christopher’s drives.

Well, the Mushkin is 25% faster than the Force 3 but only half the capacity. Remember, the Force 3 is 120GB(111GB) while the Mushkin is 60GB(55.5GB).

The 32nm Toggle nand in the Mushkin is 5000pe rated and the async nand in the two Corsairs is 3000pe rated.



EDIT
B.A.T.,
is the M4 running with no static data?

You have 100 percent free space there.

[B.A.T]

Nope. SSDLife Pro is not showing the correct info. This is a screenshot during the test at the end of the loop. I got around 40GB of static data.


New update
m4
665.5526 TiB
2407 hours
Avg speed 90.81 MiB/s.
AD gone from 231 to 227.
P/E 11563.
MD5 OK.
Still no reallocated sectors


Kingston V+100
154.8108 TiB
610 hours
Avg speed 76.34 MiB/s
AD gone from 48 to 39.
P/E?
MD5 OK.
Still no reallocated sectors

[Anvil]

Kingston SSDNow 40GB (X25-V)

391.10TB Host writes
Reallocated sectors : 11
MD5 OK

33.91MiB/s on avg (~39 hours)

Force GT "up time" ~46 hours.

--

Corsair Force 3 120GB

01 92/50 (Raw read error rate)
05 2 (Retired Block count)
B1 56 (Wear range delta)
E6 100 (Life curve status)
E7 58 (SSD Life left)
E9 163829 (Raw writes)
F1 218183 (Host writes)

104.16MiB/s on avg (~74 hours)

power on hours : 648

[Christopher]

The Mushkin was down for several hours today as I switched out the Intel DP67BG mobo for the new Asus Maximus IV Gene-Z.


Mushkin Chronos Deluxe 60 Update, Day 22

05 2
Retired Block Count

B1 12
Wear Range Delta

F1 209632
Host Writes

E9 161646
NAND Writes

E6 100
Life Curve

E7 14
Life Left

Average 129.02MB/s Avg
MSAHCI drivers, Asus M4g-z

511 Hours Work (24hrs since the last update)
Time 21 days 7hours

10 GiB Minimum Free Space

SSDlife declines to speculate on the death of the Mushkin

[bulanula]

AFAIK the NAND in the C300 is 34nm which is rated for 5000 program cycles. It is also ONFI 2.0 compliant and synchronous.

[Christopher]

I'll be out of town until Sunday night, so I won't have any updates until probably late Sunday night.

I hope the 320 doesn't die while I'm gone...

[Ao1]

Well, the Mushkin is 25% faster than the Force 3 but only half the capacity. Remember, the Force 3 is 120GB(111GB) while the Mushkin is 60GB(55.5GB).

The 32nm Toggle nand in the Mushkin is 5000pe rated and the async nand in the two Corsairs is 3000pe rated.

Thanks for confirming that. The differences in MWI and B1 are huge. It annoying that the F3 and Mushkin experienced issues, as it’s now hard to know why the outcomes are so different.

Here are comparative B1 and MWI values at ~146K writes to NAND:

Anvil: MWI: 63, B1: 54
(Corsair F3 120GB (25nm) SF2281)

Christopher: MWI: 22, B1: 9
(Mushkin Chronos 60GB (34nm toggle) SF2281)

SynbiosVyse: MWI: 10, B1: 102
(Corsair F40-A 40GB (25nm) SF1222)

[bluestang]

M225->Vertex Turbo 64GB Update:

495.21 TiB (544.49 TB) total
1303.67 hours
10126 Raw Wear
118.57 MB/s avg for the last 16.21 hours (on W7 x64)
MD5 OK
C4-Erase Failure Block Count (Realloc Sectors) at 8.
(Bank 6/Block 2406; Bank 3/Block 3925; Bank 0/Block 1766; Bank 0/Block 829; Bank 4/Block 3191; Bank 7/Block 937; Bank 7/Block 1980; Bank 7/Block 442)

[One_Hertz]

499TB. 3% wear leveling area left... As far as I understand, once this is zero, the SSD becomes read-only.

[Christopher]

Thanks for confirming that. The differences in MWI and B1 are huge. It annoying that the F3 and Mushkin experienced issues, as it’s now hard to know why the outcomes are so different.

Here are comparative B1 and MWI values at ~146K writes to NAND:

Anvil: MWI: 63, B1: 54
(Corsair F3 120GB (25nm) SF2281)

Christopher: MWI: 22, B1: 9
(Mushkin Chronos 60GB (34nm toggle) SF2281)

SynbiosVyse: MWI: 10, B1: 102
(Corsair F40-A 40GB (25nm) SF1222)

Yes, but as Anvil pointed out, when the drive disconnects it's effectively off. It can't wear level, it can't do anything because it's off.

The Mushkin hasn't spent more than 70 - 90 minutes doing anything other that endurance testing -- or disconnecting.


To be sure, the Toggle NAND equipped models are fantastic (though the Micron Sync NAND equipped models are pretty good too).

[Ao1]

Maybe, but then I’m at a loss to understand why the B1 values are so high then. Let’s look at it another way

Intel gave Tech Report the following formula:

Cycles = (Host writes) * (Write amplification factor) * (Wear levelling factor) / (Drive capacity)

From post #1927 it looks like the theoretical P/E count expires when the MWI get to 10 with SF drives. SymbiosVyse posted the following when he got to MWI 10:

05: 0
B1: 102
E7: 10%
E9: 146,240
EA/F1: 195,264

3,000 = (F1) * (E9/F1) * (X) / (40)
3,000 = (195,264) * (0.75) * (X) / (40)

3,661/3,000 Wear Levelling Factor = 0.81

Intel's X25-E has a Wear Levelling Factor of 1.1, so 0.81 is not bad at all.

Let’s see how your drive pans out when you get to 10.

I found a Micron TN on wear levelling here btw

[Anvil]

Kingston SSDNow 40GB (X25-V)

393.43TB Host writes
Reallocated sectors : 11
MD5 OK

33.44MiB/s on avg (~60 hours)

Force GT "up time" ~67 hours.

--

Corsair Force 3 120GB

01 90/50 (Raw read error rate)
05 2 (Retired Block count)
B1 55 (Wear range delta)
E6 100 (Life curve status)
E7 57 (SSD Life left)
E9 169535 (Raw writes)
F1 225775 (Host writes)

104.16MiB/s on avg (~94 hours)

power on hours : 670