SSD Write Endurance 25nm Vs 34nm

[Christopher]

Lemme drop a lazy update real quick like:

Samsung 830 256GB:
2,073,558 GiB
2,024.96 TiB
9002 PE cycles

Plextor M3P 128GB:
389,057 GiB
379.96 TiB

The 830 has 2,086,667 GiB of LBA writes. The drive only had 270GB written at the start of testing.

[canthearu]

The 520's raw error rate counter is fluctuating. I wonder what that speaks for the drive's health.

That is normal ... especially since this drive isn't doing much reading. As long as the value doesn't collapse, like it did on the mishkin, it is fine!

[Christopher]

The Mushkin went through the roof before death (RRER), dropping below threshold for two weeks before dying. The way that attribute works on SFs is kinda funny anyway, it's hard to decipher other than below threshold = bad. I also speculate that true WA was high, because it died awfully early. The raw error rate would indicate a high amount of wear on the flash, and there were few reallocs.

[canthearu]

The Mushkin went through the roof before death (RRER), dropping below threshold for two weeks before dying. The way that attribute works on SFs is kinda funny anyway, it's hard to decipher other than below threshold = bad. I also speculate that true WA was high, because it died awfully early. The raw error rate would indicate a high amount of wear on the flash, and there were few reallocs.

For your drive, I think WA was normal, but your drive was programmed for 3000 write cycle NAND. Given how cheap the Muskins were, there is a decent possibility that the NAND was a lower qualified 3k version rather then the higher qualified 5k versions. The mushkin still managed twice the MWI before dieing, which isn't too bad.

Edit: Don't forget the m4 and intel 520 would be generally getting the pick of the crop in terms of NAND quality, vs people who don't actually make the NAND themselves, getting whatever is left or whatever they are willing to pay for.

[Christopher]

The Mushkins weren't cheap last Autumn. Either way, I'm of the opinion that true WAF was higher than raw/host would indicate. I've been holding on to a 120GB Chronos D in case the mood to destroy it strikes me.

[Anvil]

Lemme drop a lazy update real quick like:

Samsung 830 256GB:
2,073,558 GiB
2,024.96 TiB
9002 PE cycles
...
The 830 has 2,086,667 GiB of LBA writes. The drive only had 270GB written at the start of testing.

2PiB should happen today

[Anvil]

Arent the new MBPs just using the 830 now?

I just looked at Anand's WWDC coverage, and the Apple SSD he's benching looks a hell of a lot like a 512GB 830.

According to Anandtech it could be Samsung or Toshiba, OWC found a SandForce label on their sample. Link
Need to get hold of the new Toshiba THNS SSD.

---

Windows update rebooted the rig some 8-9 hours ago...

Intel 330 120GB

199.60TB Host writes
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 78
[F1] Total LBAs Written 6540423
[F2] Total LBAs Read 30856 // not much use in this one
[F9] Total NAND Writes 144075GB
POH 523
MD5 OK

---.--MiB/s on avg (~-- hours)

--

Kingston SSDNow 40GB (X25-V)

1033.42TB Host writes (33863102*32)
Reallocated sectors : 05 46
Available Reserved Space : E8 99
POH 9251
MD5 OK

--.--MiB/s on avg (~-- hours)

[B.A.T]

Todays update:

m4
24h retention test complete.
MD5 OK
ASU 4.5 h - OK
No change in smart values.
48h starts now


Intel X25-M G1 80GB
24h retention test complete.
MD5 OK
ASU 4.5 h - OK
No change in smart values.
48h starts now

Intel X25-E 64GB
261.00 TiB
797-30=767 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 98
MD5 =OK
90.36 MiB/s on avg


Mtron Pro 7025 32GB
ASU write 30.7475 TiB
Host write 56,4012 TiB
Total host write 103,9368 TiB
Day 10
MWI= 97
MD5 =OK
47.75 MiB/s on avg


The Mtron dropped out sometime during the night but runs just fine now. Looks like I'm in for a rerun of the first Kingston...
Both drives have written less then normal last 24h due to a power outage when the local powercompany did some maintenace. But all is up and running normal now.

[Cyber-Mav]

Lemme drop a lazy update real quick like:

Samsung 830 256GB:
2,073,558 GiB
2,024.96 TiB
9002 PE cycles

Plextor M3P 128GB:
389,057 GiB
379.96 TiB

The 830 has 2,086,667 GiB of LBA writes. The drive only had 270GB written at the start of testing.

thats a lot of writes, im impressed with the endurance of this drive so far. im considering getting one of these samsungs or crucial m4, leaning towards the samsung from these results.

[Christopher]

Samsung 830 256GB Day 86

(GiB) 2,091,334.05
(TiB) 2,042.31

(Avg) 292.45 MB/s

(B1) Wear Leveling Count: 9078

(B6) Erase Fail Count: 7
(05) Reallocated Sectors: 28672
(POH) 2067

Plextor M3P 128GB Day 15

(GiB) 405,895.64
(TiB) 396.38

(Avg) 272.93 MB/s
(POH) 419




Okay, the 830 is over 2PiB now in LBA writes (I ended up losing 14TB or so over the course of testing). 2,105,221 GiB in LBA writes for the Samsung right now.

[kaktus1907]

nice i suppose Plextor beat 3K p/e already and also V4 which was dead after 394TiB written

[Anvil]

Intel 330 120GB

204.38TB Host writes
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 78
[F1] Total LBAs Written 6697226
[F2] Total LBAs Read 30856 //
[F9] Total NAND Writes 147528GB
POH 534
MD5 OK

123.29MiB/s on avg (~11 hours)

--

Kingston SSDNow 40GB (X25-V)

1034.86TB Host writes (33910501*32)
Reallocated sectors : 05 46
Available Reserved Space : E8 99
POH 9263
MD5 OK

37.01MiB/s on avg (~11 hours)

[Christopher]

nice i suppose Plextor beat 3K p/e already and also V4 which was dead after 394TiB written

The Plextor should be in the 3500PE cycle range right now. It should have 7000 in less than two weeks from now.

[canthearu]

The Plextor should be in the 3500PE cycle range right now. It should have 7000 in less than two weeks from now.

Should still have much life left!

[Christopher]

Hell, the real endurance test doesn't even start to 1PiB it seems. The first 1023TiB are just drive prep.

-----------
iFixit did a teardown on a rMBP, and it has a proprietary form factor 512GB 830 in it. You wont get any arguments from me about the 830's quality.

[canthearu]

Hell, the real endurance test doesn't even start to 1PiB it seems. The first 1023TiB are just drive prep.

Thats the spirit!

Hopefully the Sandisk extreme will do > 1PiB as well, otherwise it will look kind of silly next the Intel 520 and Crucial m4!

[Christopher]

Thats the spirit!

Hopefully the Sandisk extreme will do > 1PiB as well, otherwise it will look kind of silly next the Intel 520 and Crucial m4!

One of the most salient nuggets of knowledge from this endeavor can be discerned simply by looking at the charts. I hope the Plextor becomes the exception to the rule... I sometimes forget about the Plextor, like it's an afterthought. I'll start thinking about it some more as it gets closer to 1PiB.

[canthearu]

I don't really believe that slower drives = more endurance.

It all depends on:

a) NAND quality and drive size.
b) How gently the controller programs and erases the NAND, to improve NAND life.
c) How the controller detects and handles blocks that wear more then others, while not going crazy due to the average wear increasing.

There have been slow drives that die early .... like the Patriot Torqx 2.
There have been fast drives that last a long time .... like the Samsung 830 is proving to.

[canthearu]

A couple of notes:

The MWI on my Sandisk Extreme is following a close course to that plotted by Anvil's Corsair Force 120.

However, the Wear Range Delta on my drive is far below what Anvil's Corsair Force was showing. At 68494 GB NAND written, Anvil's drive was showing a WRD of 39. My Sandisk Extreme, at 67290 GB NAND written, is showing a WRD of only 3 .... I haven't actually seen the WRD on my drive exceed 5.

[B.A.T]

Todays update:

m4
48h retention test ongoing. Finished 23.00 gmt+2

Intel X25-M G1 80GB
48h retention test ongoing. Finished 23.00 gmt+2

Intel X25-E 64GB
267.21 TiB
817-30=787 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 98
MD5 =OK
90.24 MiB/s on avg


Mtron Pro 7025 32GB
ASU write 34.0158 TiB
Host write 62,5824 TiB
Total host write 110,1180 TiB
Day 11
MWI= 97
MD5 =OK
47.66 MiB/s on avg

[canthearu]

Intel 520 60GB - Day 109

Drive hours: 2,562
ASU GiB written: 814,855.74 GiB (795.76 TiB)
Avg MB/s: 93.39 MB/s
MD5: OK

Host GB written (F1): 820,676.06 GiB (801.44 TiB, 26261634 raw)
NAND writes (F9): 580,975 GiB (567.36 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 104 normalised
Avaliable Reserved Space (AA): 100 normalised
Media Wearout Indicator (E9): 1

Sandisk Extreme G25 - 120gb - Day 9

Drive hours: 215
ASU GiB written: 91,004.79 GiB (88.87 TiB)
Avg MB/s: 121.39 MB/s
MD5: OK

Host GB written (F1): 91,685 GiB (89.54 TiB)
NAND writes (E9): 68,388 GiB (66.79 TiB)

Retired Block Count (05): 4
Failure count (AB, AC): 1 program, 1 erase
Raw Error Rate (01): 105 normalized
Media Wearout Indicator (E7): 84 normalized
Wear Range Delta (B1): 3 raw
Reported Uncorrectable Errors (BB): 0 raw

[Anvil]

Intel 330 120GB

212.75TB Host writes
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 76
[F1] Total LBAs Written 6971250
[F2] Total LBAs Read 30856 //
[F9] Total NAND Writes 153564GB
POH 554
MD5 OK

123.41MiB/s on avg (~31 hours)

(it changed to 76 as I was posting)

--

Kingston SSDNow 40GB (X25-V)

1037.11TB Host writes (33984057*32)
Reallocated sectors : 05 46
Available Reserved Space : E8 99
POH 9282
MD5 OK

34.62MiB/s on avg (~31 hours)

Endurance_cr_20120614.png

[Christopher]

Samsung 830 256GB Day 87

(GiB) 2,121,258.49
(TiB) 2,071.54

(Avg) 293.05 MB/s

(B1) Wear Leveling Count: 9208

(B6) Erase Fail Count: 7
(05) Reallocated Sectors: 28672
(POH) 2096

Plextor M3P 128GB Day 16

(GiB) 434,072.90
(TiB) 423.89

(Avg) 273.62 MB/s
(POH) 448

[B.A.T]

Todays update:

m4
48h retention test ongoing. Finished 23.00 gmt+2 today

Intel X25-M G1 80GB
48h retention test ongoing. Finished 23.00 gmt+2 today

Intel X25-E 64GB
275.19 TiB
843-30=813 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 98
MD5 =OK
90.37 MiB/s on avg


Mtron Pro 7025 32GB
ASU write 38.2044TiB
Host write 70,5586 TiB
Total host write 118,0941 TiB
Day 12
MWI= 97
MD5 =OK
47.68 MiB/s on avg

[Christopher]

BAT, I'll do some power consumption testing of my 32GB MTRON today, It uses 50nm Hynix SLC and not the Micron stuff, but power use should be similar. I've never seen any power usage stats for MTRONs, but I bet they don't use much.

I'm just kinda curious. Right now, the 830 is using over 5w and the M3P only 2.5w.