SSD Write Endurance 25nm Vs 34nm

[canthearu]

i don't think i'll have a problem with the 500gb for 3-5 years.

For now, I'd still get the 830 ... it is cheaper and has beetter NAND. It is also well proven in the field.

[Christopher]

For now, I'd still get the 830 ... it is cheaper and has beetter NAND. It is also well proven in the field.

I won't argue with that!

But here in the states, the 830 is getting harder to find at such economical prices. I think the 830, while slower, isn't noticably outclassed by the 840 Pro.

[Demo]

Wow! 6PB!! I wonder if the forceGT can even come half as close to that!

[turbox997]

I know canthearu is testing the Samsung 840, but is anyone on here testing an 840 Pro by any chance??

[B.A.T]

Here are todays update:

m4
2053.8234 TiB (2 PiB)
8440 hours
Avg speed 72.11 MiB/s.
AD 241 to 202.
P/E 35367
C3 614011911 to 1786518967
01 239518 to 957059(1)
05 0 to 2048 (Reallocated sectors count)
AA 0 to 1 (Grown failing block count)
C4 o to 1 (Reallocation event count)
CE 90
MD5 OK.
Reallocated sectors : 00


Intel X25-E 64GB
1.42 PiB
4903-30=4873 hours
Reallocated sectors : 157 to 255 to 72
Available Reserved space: 56 to 43
MWI= 91
MD5 =OK
76.03 MiB/s on avg


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

OCZ Vertex 1 120GB
Will start after an update to the latest firmware.

And that was that. 2PiB done and the first reallocated sector done. Wonder how much further it can hold out.

[Christopher]

And that was that. 2PiB done and the first reallocated sector done. Wonder how much further it can hold out.

That's the equivalent of 8PiB on the 830! Excellent.

[UranusFX]

Cheers for good work!
There is also quite intense durability IMFT(Micron)?s NAND and of course SAMSUNG.
The durability of the NAND itself is a problem that becomes overspec

[canthearu]

That's the equivalent of 8PiB on the 830! Excellent.

Your drive might still make 8PiB yet!

[JR.]

Hi,

This sounds interesting - http://www.bbc.co.uk/news/technology-20579077

Regds, JR

[B.A.T]

Here are todays update:

m4
2063.1434 TiB
8487 hours
Avg speed 71.71 MiB/s.
AD 202 to 196.
P/E 35534
C3 1786518967 to 2266091151
01 957059 to 1402073(1)
05 2048
AA 1
C4 1
CE 90
MD5 OK.


Intel X25-E 64GB
1.43 PiB
4950-30=4920 hours
Reallocated sectors : 72 to 226
Available Reserved space: 43 to 39
MWI= 91
MD5 =OK
75.79 MiB/s on avg


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

OCZ Vertex 1 120GB
Will start after an update to the latest firmware.

And today the m4 has endured a year of torture

[Zaxx]

Hi,

This sounds interesting - http://www.bbc.co.uk/news/technology-20579077

Regds, JR

I hope they'll provide the oven and crucible to handle hours of baking at 800C/1,472F...

J/K...lol

Does make ya wonder if a 2.5in. enclosure could handle that kinda heat even if only applied to a few cells at a time. ATM it's looking like nand rebust-ness isn't too much in question a la the 6PB 830. I'm beginning to suspect it's the controllers that can't handle the milage. hmmm...

[Christopher]

I hope they'll provide the oven and crucible to handle hours of baking at 800C/1,472F...

J/K...lol

Does make ya wonder if a 2.5in. enclosure could handle that kinda heat even if only applied to a few cells at a time. ATM it's looking like nand rebust-ness isn't too much in question a la the 6PB 830. I'm beginning to suspect it's the controllers that can't handle the milage. hmmm...

Nah, the controllers themselves are pretty rugged. They have higher Tmax ratings that the flash itself, but the firmware is a different story. Obviously, higher temps skew Vth, but a super short burst of high heat might be the ticket to making that stuff last longer.

I do sometimes wonder about the drives using overclocked ARM silicon... that could be an issue long term.

[Christopher]

Samsung 830 256GB Day 259

(GiB) 6,445,077
(TiB) 6,294
(PiB) 6.18


(Avg) 298.63 MB/s over the past 3400+ hours

(B1) Wear Leveling Count: 27,858

(B5) Program Fail Count: 0
(B6) Erase Fail Count: 40
(B7) Runtime Bad Block Count: 831
(05) Reallocated Sectors: 3,403,776

Read Failures: 791

[canthearu]

Samsung 840 Day 19

Drive Hours: 454
ASU GiB Written: 164,332.41
Avg MB/s: 104.97 (147.59 hours)
MD5: OK

Wear Leveling Count (B1): 1315 raw (1 normalized)

Reallocated blocks (B3,05): 0 (0 sectors)
Failure count (B5, B6): 0 program, 0 erase
Uncorrectable Error Count: 0
ECC Error Rate (C3): 0

I may be here for a while with this drive .... damn samsung and their conservative engineering

At least the drive wasn't rated on the ragged edge of its true endurance.

[sergiu]

If Samsung 840 survives past 5000 cycles, then it will be a true wonder.

[Christopher]

I don't know if you guys remember, but several months back one of my X25-Es died after a FW update (I think we were talking about X25-E FW and I found out the first FW revisions were bugged). After the lock up, just having the drive connected was enough to lock a system up, and after contacting Intel support, I couldn't get my hands on newer OEM-only FW.

The good news is, I took it out of the gear closet yesterday, and I was able to SE it with RHEL. It works fine now. I'd tried this a couple times over the past few months, but for some reason it worked this time. If only I could do that to one of the dead MTRONS.

[Anvil]

Kingston SSDNow 40GB (X25-V)

1407.30TB Host writes 46114405*32)
Reallocated sectors : 05 768 -
Available Reserved Space : E8 86 (86)
POH 12860
MD5 OK

31.80MiB/s on avg (~382 hours)

--

I'll catch up tomorrow night and make those charts during the weekend.

[Christopher]

Samsung 830 256GB Day 261

(GiB) 6,488,443
(TiB) 6,336
(PiB) 6.22


(Avg) 298.66 MB/s over the past 3400+ hours

(B1) Wear Leveling Count: 28,044

(B5) Program Fail Count: 0
(B6) Erase Fail Count: 44
(B7) Runtime Bad Block Count: 913
(05) Reallocated Sectors: 3,739,648

Read Failures: 869

[canthearu]

Does make ya wonder if a 2.5in. enclosure could handle that kinda heat even if only applied to a few cells at a time. ATM it's looking like nand rebust-ness isn't too much in question a la the 6PB 830. I'm beginning to suspect it's the controllers that can't handle the milage. hmmm...

There wouldn't be any problem with the current form factor of NAND. Baking operations would be realtively rare (once every thousand writes or so) and would only affect a few thousand NAND cells at a time. As long as the casing around the heater elements can handle the heat, it would be dissipated in milliseconds.

Maybe secure erase will become "erase and bake"

[B.A.T]

Here are todays update:

m4
2084.5611 TiB
8601 hours
Avg speed 70.59 MiB/s.
AD 196 to 183.
P/E 35935
C3 2266091151 to 4206476112
01 1402073 to 3554123(1)
05 2048 to 12288
AA 1 to 6
C4 1 to 6
CE 90
MD5 OK.


Intel X25-E 64GB
1.43 PiB
4950-30=4920 hours
Reallocated sectors : 226 to 255 to 1 to 216,
Available Reserved space: 39 to 31
MWI= 91 to 90
MD5 =OK
75.22 MiB/s on avg


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

OCZ Vertex 1 120GB
Will start after an update to the latest firmware.

[Zaxx]

There wouldn't be any problem with the current form factor of NAND. Baking operations would be realtively rare (once every thousand writes or so) and would only affect a few thousand NAND cells at a time. As long as the casing around the heater elements can handle the heat, it would be dissipated in milliseconds.
Maybe secure erase will become "erase and bake"

It does have a nice ring to it...lol

I see your point re: the rarity of an actual bake operation. They could add some thermal pads and use an aluminum enclosure problem solved...that is errr, after the tech is fully developed.

[canthearu]

Samsung 840 Day 25

Drive Hours: 599
ASU GiB Written: 217,131.14
Avg MB/s: 104.45 (292.10 hours)
MD5: OK

Wear Leveling Count (B1): 1737 raw (1 normalized)

Reallocated blocks (B3,05): 0 (0 sectors)
Failure count (B5, B6): 0 program, 0 erase
Uncorrectable Error Count: 0
ECC Error Rate (C3): 0

Speed Slowly Dropping it seems.

[Hopalong X]

Speed Slowly Dropping it seems.

I'm going to assume you OP'ed the drive at least the amount suggested by Samsung Magician v3.2.
Have you been running the Magician Performance Optimization occasionaly?
This does GC and manual TRIM and will/could improve read and write performance.

Just a suggestion.

[canthearu]

I'm going to assume you OP'ed the drive at least the amount suggested by Samsung Magician v3.2.
Have you been running the Magician Performance Optimization occasionaly?
This does GC and manual TRIM and will/could improve read and write performance.

Nope, no over-provisioning nor explicit performance optimisation ... but it isn't really required on a drive with working TRIM.

These things help most on drives that are have demanding 4kb I/O loads on near full drives, and virtually no difference on desktop loads (ASU included, as it runs a desktop type load pattern) with working TRIM.

[Christopher]

Alright, people. I have an announcement to make...

The 830 is on life support.

It probably would have made 7.5PiB if I had just let it be... it had been running for 3775 hours continuously, without a power cycle. Unfortunately, I had to move to a new apartment, and so powering down the system was unavoidable. I had everything choreographed, and I was able to get the system powered back up at the new place in only 10 minutes. Sadly, this was too long.

I'll post some more information later, but it seems as though the <10 minutes was enough to cause all sorts of uncorrectable errors. For the record, the system detects the drive, but attemps to access it result in long hangs, and the partition and static data are toast.

It's a crying shame, too, because I figure the drive could have continued until all spare blocks were consumed. It was approximately 30% of the way through all of them, and so another month or two would be all I needed to kill it properly.