SSD Write Endurance 25nm Vs 34nm

[bulanula]

Wonder how long the Force 3 SSD will last.

[Anvil]

Both drives have been running for 9 hours or more on the new rig and it looks stable.
Due to the capacity/free space on the Force 3 I had to make adjustments to the cleaning part of the test.
Basically, more idle time was needed as the higher capacity drive produces >12,000 files each loop and it wasn't completely stable with such a workload.
So, 250ms added for every 1,000 files deleted and the pause is now 10 seconds.
I might relate the pause to the number of files deleted at some later point.

BTW, the "new" Endurance rig consumes 72W, excluding the monitor but including an 2TB Seagate Green HDD.
(it's an Z68 and I'm using the onboard GPU)

Kingston SSDNow 40GB (X25-V)

331.83TB Host writes
Reallocated sectors : 7
MD5 (n/a, will configure later today)

36.22MiB/s on avg (~9 hours)

--

Corsair Force 3 120GB

B1 16 (Wear range delta)
E6 100 (Life curve status)
E7 97 (SSD Life left)
E9 22944 (Raw writes)
F1 30587 (Host writes)

108.79MiB/s on avg (~10 hours)

[bulanula]

Both drives have been running for 9 hours or more on the new rig and it looks stable.
Due to the capacity/free space on the Force 3 I had to make adjustments to the cleaning part of the test.
Basically, more idle time was needed as the higher capacity drive produces >12,000 files each loop and it wasn't completely stable with such a workload.
So, 250ms added for every 1,000 files deleted and the pause is now 10 seconds.
I might relate the pause to the number of files deleted at some later point.

BTW, the "new" Endurance rig consumes 72W, excluding the monitor but including an 2TB Seagate Green HDD.
(it's an Z68 and I'm using the onboard GPU)

Kingston SSDNow 40GB (X25-V)

331.83TB Host writes
Reallocated sectors : 7
MD5 (n/a, will configure later today)

36.22MiB/s on avg (~9 hours)

--

Corsair Force 3 120GB

B1 16 (Wear range delta)
E6 100 (Life curve status)
E7 97 (SSD Life left)
E9 22944 (Raw writes)
F1 30587 (Host writes)

108.79MiB/s on avg (~10 hours)

Nice power consumption there !

[Anvil]

Yes, SandyBridge is pretty "Green".

I've got monitors that consume more power.

--

Kingston SSDNow 40GB (X25-V)

332.52TB Host writes
Reallocated sectors : 7
MD5 (n/a, will configure later today)

35.01MiB/s on avg (~15 hours)

--

Corsair Force 3 120GB

B1 17 (Wear range delta)
E6 100 (Life curve status)
E7 97 (SSD Life left)
E9 24659 (Raw writes)
F1 32869 (Host writes)

108.84MiB/s on avg (~16 hours)
(avg speed is up from last report)

[Christopher]

The Mushkin's here.




Here's some quick benches of the new Chronos Deluxe.

I'm using it on my Intel DP67BG "Extreme" motherboard, i5 2500K, a Seasonic X650 PSU, 12GB DDR3 and 3 other SSDs in here along with a GTX460.

I have a H67 uATX motherboard, but I can't get the Noctua heat sink backplate off the Intel Mobo, so I'm going to get a new 1155 processor to use as a testing rig that I can just leave in the corner.

IDLE at the wall: 49.5 - 50.1 (It's always plugged in to a Kill A Watt)

My Dell Ultra Sharp 2408wfp uses about twice that.

[Christopher]

The Mushkin was cooking for the first couple loops. Its slowed down to about 120mbs at 46%. It can manage about 100MBs with uncompressible data. I cloned my installation to it, which is about 17.5GB and I'm running with 10GB free, with the drive showing 55GB of capacity. I was expecting it would have 59GB like my other 60s. Maybe it does have RAISE?

So it's not quite as fast as I was hoping, at least in terms of endurance testing. But for my first SandForce drive, it's wickedly fast. Its brutal on normal workloads, and would be great for a system drive.

If it doesn't have LTT, I'm ordering another one.

[Christopher]

Here are the SMART attributes:





It looks like E9 is writes to NAND, while EA and F1 are host writes.

I don't understand what Life Curve Status is, but it (E6) and E7 are MWIs.

C3, C9, and CC are the ECC stuff, which are increasing linearly with writes.

Everything looks good, assuming that the rapid increase in the ECC attributes is normal. I chose the same compression level as the Force 3 is using, and it looks like raw NAND writes are 75%-80% of host writes at this level. At this compression I'm getting 120MBs vs about 100MBs uncompressed.

[sergiu]

Could you run a full span 4K random write for around 500GB both with compressible and incompressible data at QD 32? For incompressible data, I would expect a very high WA, around 10-20 as this would be worst possible scenario. For this kind of test, if you would like to do it, you would need to note the start values for E9 and EA after at least 60GB as it takes a little to fill all cells.

[Christopher]

Could you run a full span 4K random write for around 500GB both with compressible and incompressible data at QD 32? For incompressible data, I would expect a very high WA, around 10-20 as this would be worst possible scenario. For this kind of test, if you would like to do it, you would need to note the start values for E9 and EA after at least 60GB as it takes a little to fill all cells.

hmmm....

Well, given the speed of writes with this drive with little or no compression, I'm considering putting it in my SATA II laptop, then just leaving it on for the next six months (eight? eleven? fourteen??). I suppose with a little guidance I could try it in a few days, since I'm going to SE the drive, then install Win7 on it, possibly in a few days. I just want to get some uninterrupted writes down on it for now.

[Anvil]

Kingston SSDNow 40GB (X25-V)

333.25TB Host writes
Reallocated sectors : 7
MD5 (n/a)

34.31MiB/s on avg (~21 hours)

--

Corsair Force 3 120GB

B1 18 (Wear range delta)
E6 100 (Life curve status)
E7 96 (SSD Life left)
E9 26546 (Raw writes)
F1 35379 (Host writes)

108.88MiB/s on avg (~22.5 hours)
(avg speed is up from last report)

SSDLife_2011_09_21_23_52.png

[Christopher]

I see the Force 3 and the Mushkin have exactly the same percentage of host to nand writes.

Both are exactly 75%

[Christopher]

Here we go: My first official update

Mushkin Chronos Deluxe 60
F1 Host Writes 4669
E9 NAND Writes 3578
E6 Life Curve 100
E7 Life Left 0
Average 121.10MBs last 8 hours.




A monumental occasion.

[Anvil]

Kingston SSDNow 40GB (X25-V)

334.45TB Host writes
Reallocated sectors : 7
MD5 (n/a)

34.65MiB/s on avg (~32 hours)

--

Corsair Force 3 120GB

B1 21 (Wear range delta)
E6 100 (Life curve status)
E7 95 (SSD Life left)
E9 29657 (Raw writes)
F1 39520 (Host writes)

108.88MiB/s on avg (~33 hours)
(looks to be stable at about 108.88MiB/s)

edit:

I just noticed that there has been developments on several attributes

05/C4 2 (Retired Block Count)
AB/AC/B5/B6 -> 1

SSDLife_SMART_2011_09_22_11_01.png

[bluestang]

M225->Vertex Turbo 64GB Update:

341.59 TiB (375.58 TB) total
986.91 hours
7324 Raw Wear
116.68 MB/s avg for the last 16.37 hours (on W7 x64)
MD5 OK
C4-Erase Failure Block Count (Realloc Sectors) at 4
(Bank 6/Block 2406; Bank 3/Block 3925; Bank 0/Block 1766; Bank 0/Block 829)

[Christopher]

I've had some movement overnight as well.

Attribute 01, RAW READ ERROR RATE, had decreased by 2.

I always thought this was one of those totally made up attributes.




Check out the spiffy new picture.

UPDATE

After the MWI finally dropped to 99, the RAW READ ERROR RATE jumped back up to 94.

I've tried to do some research as to what RRER is (at least as far as SSDs go), and haven't found any real answers.

[Christopher]

When is a good time to start MD5 testing?

I was under the impression that it wasn't really needed until later in the life cycle of the drive, maybe once lots of reallocations occur, or in the even that there ECC corrections are less than ECC errors maybe?

[bulanula]

Yes, MD5 testing is also very useful !

[Anvil]

I've had some movement overnight as well.

Attribute 01, RAW READ ERROR RATE, had decreased by 2.
..
After the MWI finally dropped to 99, the RAW READ ERROR RATE jumped back up to 94.

Yes, that is a strange one, I'm seeing the same behavior.

There are other attributes that mirror the same "Raw value". (C3/C9/CC)

When is a good time to start MD5 testing?

Much too early, we'll figure it out when we start seeing some trends.

[Christopher]

Much too early, we'll figure it out when we start seeing some trends.

I understand, but you said you were "setting it up" for the Force 3, which is why I was confused.

It's not really something that needs to be done at 98%, unless something terrible is happening "under the bonnet".

Good news! I finally hit 98%.... I might make it through MWI before the apocalypse, assuming the apocalypse isn't three months from now.

Yes, that is a strange one, I'm seeing the same behavior.

There are other attributes that mirror the same "Raw value". (C3/C9/CC)

My Raw read error rate has plummeted down to 86, but the RAW value is also (C3/C9/CC) as you pointed out.

I think that this is something completely different than what it looks like. If this number(01) starts approaching 0, perhaps this is some precursor to throttling? I noticed it jumped back up after stopping the endurance test for a few moments so it's not likely, but it'd be nice to know what's going on. The Raw Read Error Rate sounds pretty self explanatory for mechanical HDDs, but isn't really needed for SSDs and maybe it's one of those attributes that's always included but had many uses.

This Mushkin was taken out of the box and then almost immediately started endurance testing. I'll start including 01 in updates along with power on hours, and maybe some pattern will make itself known between the two SF2281s.

EDIT
Now it's jumped back to 94.
What the hell?
I was hoping there was some correlation with another attrib. like power on hours, but sadly there is not.

[Christopher]

Tonight's Update:

Mushkin Chronos Deluxe 60
F1 Host Writes 13088
E9 NAND Writes 10086
E6 Life Curve 100
E7 Life Left 98
Average 123.77MBs.



Finally past the 10,000GB NAND write mark.

[One_Hertz]

Nothing special happening here. 4082 reallocated sectors. 437.5TB. Erase fail count down to 98 from 99. Average speed 44.26MB/s and still rising. It was 41.8MB/s when I began the test.

[Anvil]

Kingston SSDNow 40GB (X25-V)

337.10TB Host writes
Reallocated sectors : 7
MD5 OK

32.19MiB/s on avg (~15 hours)

--

Corsair Force 3 120GB

01 94/50 (Raw read error rate)
05 2 (Retired Block count)
B1 23 (Wear range delta)
E6 100 (Life curve status)
E7 93 (SSD Life left)
E9 36446 (Raw writes)
F1 48573 (Host writes)

107.79MiB/s on avg (~15 hours)
Uptime 140 hours. (power on hours)

SSDLife estimates lifetime to be 1 month 20 days.

I understand, but you said you were "setting it up" for the Force 3, which is why I was confused.

It's not really something that needs to be done at 98%, unless something terrible is happening "under the bonnet".

EDIT
Now it's jumped back to 94.
What the hell?
I was hoping there was some correlation with another attrib. like power on hours, but sadly there is not.

As I was moving to another rig I had to pick a new file for MD5 testing as the Kingston has been doing MD5 for quite some time.
No errors reported on the Kingston and so MD5 testing might have been started a bit early.
Unless there is something strange going on with the Force 3 I might not enable MD5 testing for another 2-300TB's

The 01 (Raw read error rate) on the F3 is back at 94 for me as well, will try to find out what is going on, it will most likely change the next few days.

[Christopher]

I watched 01 jump back and forth quite a bit. Since the RAW is the ECC numbers, I guess it just changes based on part of the RAW number. I just kept refreshing the smart data with CDI, and it jumps around from minute to minute so.

My guess is what's important is the RAW ECC numbers that a HDD doesn't have, so the interpretation is wrong. I wish I knew what it was, but it looks like some worthless appendage from mechanical drives.

[Anvil]

It looks that way, it has now changed to 82. (15-20 minutes ago it was 94)

I'll still report it for a few more days in case it starts making sense. (can't see any connections to other things though)

[bulanula]

The Intel 320 really is putting up a fight. Speed is rising !?