SSD Write Endurance 25nm Vs 34nm

[Christopher]

The 256GB 830 is 4x the capacity of the 64GB M4, and does 4x the M4's average speed. Of course, the M3P is 2x the capacity and almost 4x the M4's average speed, so at this point, I'd say the Crucial with 18K PE cycles on it will still probably outlive the M3P.

[B.A.T]

Time for a retention test on that m4 *mwhahahaha*

But grats on the 1PB ... I didn't think the m4 could make it.

Thank you!
The first m4 showed signs of failure from 700 TiB but this one has been nice all the way. I had plans to test retention after 1 PiB and I think I'll try some 6, 12 and 24h after the weekend.

The 256GB 830 is 4x the capacity of the 64GB M4, and does 4x the M4's average speed. Of course, the M3P is 2x the capacity and almost 4x the M4's average speed, so at this point, I'd say the Crucial with 18K PE cycles on it will still probably outlive the M3P.

Have you calculated the WA on those drives? The m4 got at the moment a very low WA and I think that is what's took it this far.

[canthearu]

Have you calculated the WA on those drives? The m4 got at the moment a very low WA and I think that is what's took it this far.

Sure the difference is huge between a drive that does write amp of 5 and a drive that does 1.5 ... however there shouldn't be much difference between a drive that does 1.1 write amp and a drive that does 1.05.

My suspicion is that NAND manufacturing has improved over the last year on the 25nm node, and with newer firmware, have combined to make the NAND more resistant to write wear/bit fade. On the other hand, it could simply be luck ... it wouldn't take much to screw up a drive and make it panic lock.

[Christopher]

The 830 is right at 1.10, the M3P I'm not so sure about, but I have no reason to believe its higher than 1.1ish. Practically every drive with TRIM should be able to manage 1.1 with endurance test loads.

[Anvil]

Intel 330 120GB

168.17TB Host writes
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 82
[F1] Total LBAs Written 5510665
[F2] Total LBAs Read 30844 // not much use in this one
[F9] Total NAND Writes 121390GB
POH 430
MD5 OK

123.08MiB/s on avg (~17 hours)

--

Kingston SSDNow 40GB (X25-V)

1023.47TB Host writes
Reallocated sectors : 05 46
Available Reserved Space : E8 99
POH 9157
MD5 OK

36.02MiB/s on avg (~17 hours)

[kensiko]

FYI, there is a 11.5 RST Beta driver out, TRIM in RAID 0 or 1 array.
http://www.station-drivers.com/page/intel%20raid.htm

[Anvil]

11.5.0.1171 beta does not look to bring TRIM to raid yet.
I've tested on W8 which was where one thought it would start working, there is still something missing.
(tested on a P8P67 (rev1), orom 10.5.0.1034, it might require an updated orom, so, if there is an updated bios with 11.5... OROM I'll give it a try)

[Christopher]

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

Plextor M3P 128GB Day 12
312,528.18 GiB
305.20 TiB

272.31 MB/s Average

322 Hours

Rellocated Event Count 0

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

Samsung 830 256GB Day 82

1,967,199.61 GiB
1,921.07 TiB

292.22 MB/s Average
1969 Hours

8648 Wear Leveling Count
MWI 1

7/0 Erase/Program Fail
Used Reserved Block Count: 14/28672 sectors
---------------------------------------------
Attachment 127463

That's 2,005,063GiB in LBA writes.

[Anvil]

Kingston SSDNow 40GB (X25-V)

1.00PiB Host writes
Reallocated sectors : 05 46
Available Reserved Space : E8 99
POH 9167
MD5 OK

33.33MiB/s on avg (~9 hours)

2012-06-09-2245.PNG

[B.A.T]

Congratz Anvil!! A little over a year of writing for the poor little ssd... A true accomplishment to the toughness of the Intel controller and 34nm nand.
Might be time for the P/E graph again with all this milestones being passed.

Todays update:

m4
1029.0958 TiB
4321 hours
Avg speed 74.77 MiB/s.
AD gone from 24 to 21
P/E 17730.
CE 26
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
700.68 TiB
23630hours
Reallocated sectors : 142 to 152
Available Reserved space: 75
MWI= 50 to 44
MD5 =OK
39.40 MiB/s on avg


Intel X25-E 64GB
235.72 TiB
718-30=688 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 98
MD5 =OK
92.53 MiB/s on avg


Mtron Pro 7025 32GB
ASU write 21.4589 TiB
Host write 39,1834 TiB
Total host write 86,7190 TiB
Day 6
MWI= 99 to 98
MD5 =OK
47.04 MiB/s on avg


A lot of things going on at the same time now. 1 PiB for the m4 today. The G1 passed 700 TiB and are going through a retention test tomorrow and I had a deja-vu today.
The Mtron pulled a Kingston V+100 on me and dropped out earlier today. Lets just hope it'll not repeat that to much otherwise I'll get really tired of restarting my endruance rig.
On the other hand the mwi dropped from 99 to 98 after 21.4589 TiB of ASU write. To be more accurate I need it to drop another to 97 just to be sure how much host write it takes to make it drop 1%.

I was thinking of testing retention on the m4 but then I might miss the chance to get it into a read only state. What do you think?

[[XC] Oj101]

Oh how I hope that Sammy makes it to 2 PiB I was surprised when it hit 1.5.

I must say I'm not sure what to think of the X25-E already dropping its MWI to 98, shouldn't SLC outlive MLC by a LOOOOOOOOOOOOONG shot? As far as I'm aware, the X25-E is rated at 100,000 P/E cycles. That's something like 72 GB/day for a century.

[B.A.T]

It took 150 TiB to drop 1%. that is 14.6 PiB not taking into account WA without trim. I think it'll last a very long time but we just have to wait and see if the drop is going at the same rate or if it changes.

Edit:
100K*80GB=8PB
MWI shows better than the spec.

[[XC] Oj101]

This is what happens when you browse the forums late at night, I read it as 23.572 TiB :/

[B.A.T]

An easy mistake to make At least the X25-E looks better now then?
I'm curious to see if mwi follows P/E of 1000=80 TB or like now with 150 TiB.

[canthearu]

I was thinking of testing retention on the m4 but then I might miss the chance to get it into a read only state. What do you think?

I'm absolutely certain that there is no read only mode for SSDs. (my vertex 4 wasn't readonly ... just crashing on write/erase activity)

[B.A.T]

I think the Samsung 470 reached a read only mode before it died.

[canthearu]

I think the Samsung 470 reached a read only mode before it died.

http://www.xtremesystems.org/forums/...=1#post4932126

Nah, the drive could still perform some writes, but because of the condition of the NAND, the controller kept crashing.

Thus it was not returning the correct errors for read-only mode. There is a specific error to return to say the device is readonly, and crashing the controller isn't the correct method

[Christopher]

I'm working on finding the correct reference material again, but drives becoming read only is for the most part a decision the controller makes. Most drives only become read only once all/most of the reserve blocks are gone after the controller write protects the drive. We never get to see this behavior because the drives are all dying from something else.

I found one instance again --

From a Micron Product PDF for the P400e:

Total Spare is the block count used to substitute (replace) bad blocks before the drive
goes into write protect mode. It is a fixed value.


PercentageLifetimeUsed is calculated as:

PercentageLifetimeUsed = max(PercentageEraseCountUsed,
PercentageSpareBlockUsed)

It's not something that happens organically, it's something the controller decides to do in the event there are no reserve blocks left. Also, I believe SF does this as well. So while Read Only is possible, it's practically never going to happen. Unless maybe the X25 V starts dropping several blocks every day for the next couple of years.

[Christopher]

Won't be long now. LBA writes are 13TiB ahead or so, but either way 2PiB isn't far off.

[B.A.T]

That is one quick writer

[canthearu]

When the 830 reaches 3PiB, that is when things will probably start to get a bit perillous!

[Zaxx]

Man that Sammy is smokin'. Just makes me wonder just how much the 512MB 830 could rack up. Can't imagine someone sacrificing that big of a drive tho...I couldn't.

[2f_tfe]

Thank you!
The first m4 showed signs of failure from 700 TiB but this one has been nice all the way. I had plans to test retention after 1 PiB and I think I'll try some 6, 12 and 24h after the weekend.

well done

the retention test is a good idea, don't forget to update to fw 0309 (or 000f)
to prevent the "5K hours bug"

bye
luke

[Anvil]

Intel 330 120GB

172.83TB Host writes
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 81
[F1] Total LBAs Written 5663293
[F2] Total LBAs Read 30852 // not much use in this one
[F9] Total NAND Writes 124755GB
POH 452
MD5 OK

122.95MiB/s on avg (~11 hours)

The 330 has been on and off as I needed that port for updating fw on the Vertex 4's.
It was idling for a few hours while the X25-V joined as a member of the PiB club

--

Kingston SSDNow 40GB (X25-V)

1026.18TB Host writes
Reallocated sectors : 05 46
Available Reserved Space : E8 99
POH 9180
MD5 OK

37.31MiB/s on avg (~11 hours)

--
Endurance_PEsummary_20120609.png

Endurance_cr_20120609-10.png

[Anvil]

I was thinking of testing retention on the m4 but then I might miss the chance to get it into a read only state. What do you think?

I'm planning on having a retention test next weekend (Friday-Sunday) on both drives, it will be the first for the 330, the V has had quite a few tests but it's been a while since last time. (probably at 800GiB)

An SSD getting into a true Read-Only state is what we have been "told" or told our-selves would happen, it does not look to happen in the real-world.
Even if it did, data-retention would make unplugging the drive a high risk sport.

On the other hand, if there is a chance that someone implemented a Read-Only state I'd say the odds are good on Intel and Micron.

So, it's a tough one, I'm set on performing a test next weekend, it may end in "tears" though
(if it makes it through the test there will be no more retention testing on the V)