SSD Write Endurance 25nm Vs 34nm

[Christopher]

My drive doesn't ever get a chance to idle since it's either writing 120+ mbs or disconnected. I only have 17 percent static data anyway, as I used a fresh Win7 install on it for when I was using it in a laptop. Today WRD has advanced to 11, so perhaps it will go back up to 20+.

[Christopher]

Mushkin Chronos Deluxe 60 Update, Day 18

05 2
Retired Block Count

B1 11 Going up
Wear Range Delta

F1 168250
Host Writes

E9 129720
NAND Writes

E6 100
Life Curve

E7 32
Life Left

Average 128.23MB/s Avg
RST drivers, Intel DP67BG P67

415 Hours Work (23hrs since the last update)
Time 16 days 7 hours

12GiB Minimum Free Space 11720 files per loop

SSDlife expects 7 days to 0 MWI

[bulanula]

Most likely controller failure for the Corsair drive. Too bad because it even had LTT removed. So I guess with the Mushkin disconnecting, we won't get to see how SF does perform !? Maybe SF is just not suitable for this kind of testing. Or maybe Anvil's drive will give us some results without the controller crapping out.

[Ao1]

I read a note that wear levelling via static data rotation can potentially impact performance, so one could assume that idle time is not required, however I also read that the SF controller is supposed to maintain the wear delta within a few % of the maximum lifetime wear rating of the NAND.

Vapor indicates a value of 3 in post # 1,899, which seem about right to how the SF is supposed to behave. AFAIK Vapor’s SF has run continuously. So how could the F3 get to 58? If the P/E is 5K or 3K the difference between least and most worn is huge and would take a lot of writes (not just idle time) to rebalance.

3 = [(166.7) / 5,000] x 100 - Vapor
11 = [(550) / 5,000] x 100 – The highest I saw before aborting
58 = [(2,900) / 5,000] x 100 - Anvil

It seems there are two triggers for static data rotation:

• Length of time that data remains static before it is moved
• The maximum threshold between least and most worn blocks

If SF works on the first trigger then it would be reasonable that the delta would get quiet high if the drive is written to heavily. That does not seem to match up with what Vapor reports.

[Christopher]

Most likely controller failure for the Corsair drive. Too bad because it even had LTT removed. So I guess with the Mushkin disconnecting, we won't get to see how SF does perform !? Maybe SF is just not suitable for this kind of testing. Or maybe Anvil's drive will give us some results without the controller crapping out.

I don't think you should worry about results with the Mushkin or the Force 3. It's not like total drive failures are common, but I'm not really surprised one drive (the F40-A) died from something other than wearing out the NAND. The chances of it happening to another drive in the near future is minuscule.

I've bought a new motherboard, it should be here this later this week. While I think it's ridiculous to have to buy hardware that works with the SF2281 rather than just having a drive that works with anything (you know, like it's supposed to), I'm committed to figuring out what the hell is going on. I can always reboot once a day as this should stop the drop outs from occurring, but that's not a very good solution (I'm not always going to be here to babysit the Mushkin). I've made some other changes this weekend, and as a result, I'm already past my own personal consecutive time between drive failures. If I can make it to 40 - 50 hours I may just be on to something.

The new motherboard is a Maximus IV Gene-z, so that makes one P67, H67, and Z68 board I'll end up testing on with the only hardware difference being motherboards. The P67 and H67 were basically the same at no more than 30 hours of running. Either the Mushkin lasts no more than 30 hours with the new motherboard, or it just works. Like I said, in the meantime I'm trying something else, but it would be hilarious if the Mushkin worked for the next 3 days until the Gene-z gets here. But it would also be hilarious if people have been swapping out the wrong components the whole time...

[Anvil]

Kingston SSDNow 40GB (X25-V)

382.22TB Host writes
Reallocated sectors : 11
MD5 OK

36.12MiB/s on avg (~21 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 64 (SSD Life left)
E9 141282 (Raw writes)
F1 188176 (Host writes)

106.62MiB/s on avg (~21 hours)

power on hours : 561

--

Wear Range Delta is slowly decreasing, nothing else of importance except for that it's still running.

Ao1,
We'll just have to monitor the Force 3, the next few days should tell us the trend although it looks like a slow process.
Some parameters a surely related to the capacity and free space.
Maybe the continuing creating and deleting of "fresh" files disturbs the advanced logic on the SF controller.

[Anvil]

..
I've made some other changes this weekend, and as a result, I'm already past my own personal consecutive time between drive failures. If I can make it to 40 - 50 hours I may just be on to something.
...

What changes and are you running a no "Power save" setup?

I'll know in 12 hours or so, the speed is OK for an 3Gb/s port.

[Christopher]

What changes and are you running a no "Power save" setup?

I'll know in 12 hours or so, the speed is OK for an 3Gb/s port.

All power saving technologies enabled, same RST 10.6 drivers. It's pulling 46.9w from the wall right now.

34 hours straight... still too early to tell, but still a record for me.



I don't want to jinx myself by proclaiming that I've found a fix, so the most I'll say at the moment is it SHOULD have already crashed, or SHOULD crash any moment...

[Anvil]

Lets hope it works out and is reproducible!

[bluestang]

M225->Vertex Turbo 64GB Update:

466.85 TiB (513.31 TB) total
1245.34 hours
9609 Raw Wear
118.46 MB/s avg for the last 64.76 hours (on W7 x64)
MD5 OK
C4-Erase Failure Block Count (Realloc Sectors) from 6 to 7.
(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)

[Anvil]

The boot drive just disconnected, a Force GT on fw 1.3.

I've got a screenshot from 30min ago so avg speed etc is pretty close, the other values are the current ones, taken from CDI.


Kingston SSDNow 40GB (X25-V)

382.91TB Host writes
Reallocated sectors : 11
MD5 OK

35.33MiB/s on avg (~27 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 64 (SSD Life left)
E9 143026 (Raw writes)
F1 190497 (Host writes)

MD5 OK

106.65MiB/s on avg (~27 hours)

power on hours : 568

[Ao1]

Here is a summary of B1 data posted on SF drives. There is not enough info on what happened with Vapor’s SF drive. Vapor also did a lot of testing on compression before getting started on the endurance app, which would have made a difference; plus it has modified f/w.

Anyway, when the endurance app is running all the time B1 appears to only increase. I’m wondering if SynbiosVyse’s drive failed due the fact that B1 got too high. The difference between least and most worn blocks exceeds the P/E cycle capability.

130 = (6,500 / 5,000] x 100
130 = (3,900 / 3,000] x 100

Edit: Looking at SynbiosVyse’s posts he mentions he had hardly any static data (a few MB?) yet B1 was still able to significantly increase above the target SF value.

[B.A.T]

Update:
m4
641.5496 TiB
2331 hours
Avg speed 90.95 MiB/s.
AD gone from 245 to 240.
P/E 11165.
MD5 OK.
Still no reallocated sectors


Kingston V+100
The drive is up again but I need to restore the log before I can restart the test.

[bulanula]

The real hero here is the M4. Too many problems and headaches with the SF. I wonder if any SSD will beat in in the future. Probably the C300

[Anvil]

Here is a summary of B1 data posted on SF drives.

Nice compilation, maybe Vapor has more info on his drive(s)

It looks like mine hit some boundary at 59, it's still at 55. (the drive is not back online yet)

[Christopher]

Now I'm starting to get nervous... 45 hours straight without disconnects.

[One_Hertz]

well, I give my 320 another week or two. Reallocations are rising at the rate of about 150 per day now and the SSD is slowing down a bit. I had a 2 day setback due to a failed part in my testing rig (ram).

4694 reallocated sectors. Reserved space is at 16. Erase fail count is at 86. 487.5TB written.

[Christopher]

Mushkin Chronos Deluxe 60 Update, Day 19

05 2
Retired Block Count

B1 10
Wear Range Delta

F1 179537
Host Writes

E9 138428
NAND Writes

E6 100
Life Curve

E7 27
Life Left

Average 127.74MB/s Avg
RST drivers, Intel DP67BG P67

440 Hours Work (25hrs since the last update)
Time 18 days 8 hours

12GiB Minimum Free Space

SSDlife expects 6 days to 0 MWI


47.36 Consecutive Hours, A new record!

About two minutes after I posted the update, I received an email from Mushkin about a new FW release. The new 3.30 release is already available or will be soon for drives that use standard SF FW.

I think I have to skip this FW after finally finding stability. If the Mushkin can make it another 24hr, I think it's stable in my slightly revised system.

[Christopher]

well, I give my 320 another week or two. Reallocations are rising at the rate of about 150 per day now and the SSD is slowing down a bit. I had a 2 day setback due to a failed part in my testing rig (ram).

4694 reallocated sectors. Reserved space is at 16. Erase fail count is at 86. 487.5TB written.

How long have you been getting 150 reallocations a day for?

If you really think it's going to die soon, maybe you should post more frequent updates (if possible).

[deathman20]

The real hero here is the M4. Too many problems and headaches with the SF. I wonder if any SSD will beat in in the future. Probably the C300

Indeed the champ continues. Watch they miss labeled some SLC NAND in that drive.. ooops

[Christopher]

Indeed the champ continues. Watch they miss labeled some SLC NAND in that drive.. ooops

Maybe we have a ringer here...

Maybe Crucial actually gets the best NAND out of the IMFT venture. Maybe Intel get the bottom of the barrel, the dregs of NAND production.

I think the C300 and the M4 are going to peter out around the same point, but the C300 will just last longer due to it's slowness. That point might be 20K or 30K PE cycles though, so don't think it's happening next Thursday at 04:37 GMT. I think it's going to take most of a year to kill it.

It's hard not to be impressed by the M4 though. I'm really looking forward to the 320 giving up the ghost soon, but maybe it's just another false alarm.

[One_Hertz]

How long have you been getting 150 reallocations a day for?

If you really think it's going to die soon, maybe you should post more frequent updates (if possible).

It was 150 for the last day. It is speeding up a lot so actually I think this has a few days left. Reallocations are now at 4719. It rose by 25 over the last 3 hours. Wont make it over 500TB probably.

[Christopher]

At 500TB the 320 40GB will have written it's own capacity 12,500 times -- not too shabby. The M4 has almost done that 10,000 times. Perhaps the difference between async and sync nand is more important for longevity than just 25nm vs 34nm. The nand in the 320 should be really similar to the Micron *AAA 25nm async, while the M4 uses *AAB sync nand. Undoubtedly the IMFT sync nand is higher quality, not just faster.

[Anvil]

well, I give my 320 another week or two. Reallocations are rising at the rate of about 150 per day now and the SSD is slowing down a bit...

Both 40GB Intel drives have proven to be exceptional value, 500TB is not bad at all although I'm aiming/hoping for more

--

Running on the X58, Power savings Enabled, C-States Disabled.
Boot drive is the Corsair Force GT 120GB FW 1.3.2, online for 16 hours.

Kingston SSDNow 40GB (X25-V)

384.81TB Host writes
Reallocated sectors : 11
MD5 OK

33.99MiB/s on avg (~12 hours)

--

ASRock Extreme4 Z68
Power savings Enabled, C-States Disabled

Corsair Force 3 120GB

01 89/50 (Raw read error rate)
05 2 (Retired Block count)
B1 54 (Wear range delta)
E6 100 (Life curve status)
E7 63 (SSD Life left)
E9 146890 (Raw writes)
F1 195641 (Host writes)

104.21MiB/s on avg (~12 hours)
As a result of Power savings the avg is down a bit. (normal)

power on hours : 583

--

As a test I have increased the delay while deleting files, it is now 500ms per 500 files.
So far it looks OK, the result is that it is able to write the set amount of random writes every time.

I have made the pause a user setting so if the drive has no issue with deleting thousands of files it can be set as low as .5 seconds and by doing that one can sort of regain the time lost caused by the added delay while deleting.

[Ao1]

It is on its way to me via USPS.

Hi One_Hertz, did you find anything interesting on John’s SSD? Are you able to determine the P/E cycle count at a block level? (I’d be really interested to find out how the wear was distributed over the blocks).