SSD Write Endurance 25nm Vs 34nm

[Anvil]

Kingston SSDNow 40GB (X25-V)

541.67TB Host writes
Reallocated sectors : 05 14
Available Reserved Space : E8 99
POH 4913
MD5 OK

33.55MiB/s on avg (~48 hours)

--

Corsair Force 3 120GB

01 94/50 (Raw read error rate)
05 2 (Retired Block count)
B1 45 (Wear range delta)
E6 100 (Life curve status)
E7 10 (SSD Life left)
E9 512854 (Raw writes) ->501TiB
F1 682612 (Host writes) ->667TiB

MD5 OK

106.49MiB/s on avg (~168 hours)

power on hours : 1948

WRD decreased by 1 from 46 to 45.

--

@Christopher

I expect 4096 could be the smallest unit for reporting.

Nothing to worry about.

[Christopher]

Anvil,

I figured that out. I'm most pleased that SSDlife now shows proper attribute names. I now have context, and will include smart values in updates (since I know what they are now).

Reallocations have increased to 8196 within the past 30 mins.

Having PE count is handy too.

[Christopher]

Samsung 830 64GB Update, Day 4
FW:CXM01B1Q

GiB written:
45390.03

Avg MB/s
107.57

Per-day average:
9078

PE Cycles
72

Reallocated Sector Count
8196 (2 blocks)

120 hours

[johnw]

Umm... the Samsung's had some major movement... maybe?
Reallocated sectors went from 0 to 4096

Note that the normalized value decreased from 100 to 99 (threshold 10) when the raw value increased from 0 to 8192. With only one decrement of the normalized value, any extrapolation will be very rough, but current estimate is that the SSD can have 89 more of those RAW increases by 6144 (average of 4096 and 8192) before it has any trouble.

[B.A.T]

Todays update.
Kingston V+100
283.3777 TiB
1288 hours
Avg speed 28.60 MiB/s
AD still 1.
168= 1 (SATA PHY Error Count)
P/E?
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
115,5559 TiB
19507 hours
Reallocated sectors : 00
MWI=45 to 38
MD5 =OK
42.72 MiB/s on avg


m4
33.2744 TiB
122 hours
Avg speed 82.57 MiB/s.
AD gone from 85 to 81.
P/E 584.
MD5 OK.
Reallocated sectors : 00

[Christopher]

Note that the normalized value decreased from 100 to 99 (threshold 10) when the raw value increased from 0 to 8192. With only one decrement of the normalized value, any extrapolation will be very rough, but current estimate is that the SSD can have 89 more of those RAW increases by 6144 (average of 4096 and 8192) before it has any trouble.

Yeah, I figured that out.

Look at the SMART data in the update I posted today. It incremented by 4096 twice (I believe these are actual sectors). Retired blocks equals 2 now. But SSDlife and CDI didn't show attribute names. Now SSDlife does since switching to the installer version, so that helps. I assumed that 4096 equals 2MiB, but that doesn't really fit since 100 x 2MiB is only 200MiB, of which it can lose ~180MiB? I'll have to do more research, but I know that 1) reallocated sectors increase in increments of 4096 and 2) retired blocks increase correspondingly in increments of 1. I basically watched it increase from 4096 to 8192 and from 1 to 2. So that's why I put 8196 = 2 blocks in the update.

The case of the Samsung 830 is also very interesting. We will see how it compares to the 470 but I bet it will last much less considering the worse NAND it has. The controller is not much improved either, I bet. Samsung controllers always have been very proprietary and pretty simple but they do not easily fail like the SF ones.

I don't think the 32Gbit 27nm Toggle NAND on the Samsung is all that awful. It's probably better than Toshiba's Toggle NAND. Yes, the controller is a black box -- but so is every other controller. Samsung drives do have the distinction of being one of the main OEM drive manufacturers. It's not like Apple is putting Intel drives in Macs (though they are putting some fairly awful performing Toshibas in there). The 830 is OEM equipment for several high end laptops such as Dell Precisions and the like. That's always a good thing as that usually means everyone gets in some validation.

[bulanula]

The case of the Samsung 830 is also very interesting. We will see how it compares to the 470 but I bet it will last much less considering the worse NAND it has. The controller is not much improved either, I bet. Samsung controllers always have been very proprietary and pretty simple but they do not easily fail like the SF ones.

[Anvil]

Kingston SSDNow 40GB (X25-V)

542.97TB Host writes
Reallocated sectors : 05 14
Available Reserved Space : E8 99
POH 4925
MD5 OK

43.32MiB/s on avg (~10 minutes)

--

Corsair Force 3 120GB

01 90/50 (Raw read error rate)
05 2 (Retired Block count)
B1 47 (Wear range delta)
E6 100 (Life curve status)
E7 10 (SSD Life left)
E9 516166 (Raw writes) ->504TiB
F1 687021 (Host writes) ->671TiB

MD5 OK

113.91MiB/s on avg (~10 minutes)

power on hours : 1961

I had to install some updates and so the computer was restarted.

[Christopher]

Samsung 830 64GB Update, Day 6
FW:CXM01B1Q

GiB written:
54505.70

Avg MB/s
107.91

Per-day average:
9084GB

PE Cycles
87

Reallocated Sector Count
8196 (2 blocks)

146 hours

[B.A.T]

Todays update.
Kingston V+100
285.4463 TiB
1311 hours
Avg speed 25.61 MiB/s
AD still 1.
168= 1 (SATA PHY Error Count)
P/E?
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
119,6193 TiB
19529 hours
Reallocated sectors : 00
MWI=38 to 33
MD5 =OK
51.45 MiB/s on avg


m4
39.7725 TiB
145 hours
Avg speed 80.41 MiB/s.
AD gone from 81 to 78.
P/E 693.
MD5 OK.
Reallocated sectors : 00

[Christopher]

So after I d-flashed my Agility 60 for the second time, it's started exhibiting some very strange behavior... It's acting exactly like the Chronos Deluxe did as a system drive before the 3.3.2FW "fix". The same pause, then BSOD, then it needs a full manual power cycle. It's most curious... I mean exactly like a SF2281 hang/BSoD/reboot then manual power cycle. On the positive side, at least I don't have to d-flash it again (maybe?). I had always thought of my Indys as pretty bullet proof, but this one's getting kinda flakey.

[Anvil]

Kingston SSDNow 40GB (X25-V)

547.68TB Host writes
Reallocated sectors : 05 14
Available Reserved Space : E8 99
POH 4966
MD5 OK

33.45MiB/s on avg (~41 hours)

--

Corsair Force 3 120GB

01 85/50 (Raw read error rate)
05 2 (Retired Block count)
B1 52 (Wear range delta)
E6 100 (Life curve status)
E7 10 (SSD Life left)
E9 527666 (Raw writes) ->515TiB
F1 702329 (Host writes) ->686TiB

MD5 OK

106.46MiB/s on avg (~41 hours)

power on hours : 2002

WRD is increasing rapidly, I expect it will start decreasing again within a few more days.

--

@bluestang

What is the status on the M225?

[badatgames18]

Kingston SSDNow 40GB (X25-V)

542.97TB Host writes
Reallocated sectors : 05 14
Available Reserved Space : E8 99
POH 4925
MD5 OK

43.32MiB/s on avg (~10 minutes)

--

Corsair Force 3 120GB

01 90/50 (Raw read error rate)
05 2 (Retired Block count)
B1 47 (Wear range delta)
E6 100 (Life curve status)
E7 10 (SSD Life left)
E9 516166 (Raw writes) ->504TiB
F1 687021 (Host writes) ->671TiB

MD5 OK

113.91MiB/s on avg (~10 minutes)

power on hours : 1961

I had to install some updates and so the computer was restarted.

EDIT: nvm.. answered my questions lol

[devsk]

bluestang: do you load the M225 again to see if it loads? May be there is a miracle waiting to happen...

[Christopher]

Samsung 830 64GB Update, Day 7
FW:CXM01B1Q

GiB written:
63610.84

Avg MB/s
107.91

Per-day average:
9087GB

PE Cycles
87

Reallocated Sector Count
8196 (2 blocks)

PE Count:
101

170 hours






If you look at the SMART data, notice that the Reallocated Sector count is at 8196. It went from 0 to 4096, then from 4096 to 8192. So look at the used reserved block count... 4. Does that make each block equal 2048? If a page is 4K, 128 pages equals 512K. Not sure how the math works out, but 8196 x 512 bytes = 4MiB.

It sounds really basic, but I really have no clue how this part works out. I know Indilinxes use (or used?) 4K pages, 128 pages equals one block, so each block would equal 512K or 100,000 512 byte sectors.

I keep trying various Google searches for specific Samsung 830 reallocated sector info, and to my dismay the most relevant article is... page 119 of this thread.

[B.A.T]

Todays update.
Kingston V+100
287.8517 TiB
1338 hours
Avg speed 25.45MiB/s
AD still 1.
168= 1 (SATA PHY Error Count)
P/E?
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
124,5284 TiB
19557hours
Reallocated sectors : 00
MWI=33to 26
MD5 =OK
51.40 MiB/s on avg


m4
47.4961 TiB
173 hours
Avg speed 80.85 MiB/s.
AD gone from 78 to 73.
P/E 828.
MD5 OK.
Reallocated sectors : 00

[Christopher]

Why is the M4's wear leveling count at 828 now? The Samsung is only at 101. Perhaps they calculate differently?

[B.A.T]

I was wondering about the same thing.
In my case it's easy to calculate. 64 020 803 584 byte*828 =48,21 TiB
With an WA of 1.01 it looks correct. Yours on the other side......

[minpayne]

Is there no MWI for the Samsung 830? I've also read somewhere that the temperature of that SSD is a bit high, hence not recommended for small laptops where cooling might be a problem.

[Christopher]

I was wondering about the same thing.
In my case it's easy to calculate. 64 020 803 584 byte*828 =48,21 TiB
With an WA of 1.01 it looks correct. Yours on the other side......

Yes. Perhaps it's a straight average? Maybe it's pe cycles / 10? Perhaps it's actually in Klingon.

I don't see how it's possible to have over 1000 times the drive's capacity with slightly over 100 PE cycles. Maybe it's magic.

Is there no MWI for the Samsung 830? I've also read somewhere that the temperature of that SSD is a bit high, hence not recommended for small laptops where cooling might be a problem.

For real.

It gets pretty warm, but so did the Mushkin Chronos Deluxe I was testing before the Samsung. The toggle mode NAND models use a good bit of power, both from the NAND and controllers. The both should have thermal pads to conduct heat into the chassis. My Samsung 830 is actually hanging out of the back of my test system, so it's temp readings are from it's own heat and not the system. I should think it would be fine in a smaller laptop chassis as it only gets hot when you are writing to it as fast as you can for days on end. Laptops tend to get hot anyway, so it's more likely that a laptop heats up the Samsung. Indilinx drives and Intel drive don't get appreciably hot, but then, they use very little power. All of the 6gbps controllers will use more juice anyway, but the toggle mode NAND (Toshiba and Samsung) seem to use the most power on a per-device basis. But boy, are they fast.


Concerning SSD power use, newer drives like the Samsung are using as much power as 2.5" HDDs. But like a SB processor, they don't use much power at idle, then use larger amounts of power while completing work, then return to idle. Because they are so much faster than mechanical drives, they spend one twentieth the time doing the same amount of work at max power which means they can get back to idle as fast as possible. So the drive might heat up for a minute or two at a time, but shouldn't ever get problematically hot unless you're doing what I'm doing.

EDIT
If you look at the whole wear leveling values for the 830 you get 97/97/101. It's 73/73/838 for the M4. Maybe I should add some more static data to see what happens.

[bluestang]

Hey guys, just got in back to work today...was out sick yesterday.

I'll try and see if I can get the M225 back today.

EDIT: No go. Can't seem to be able to D-flash FW to any version. Keeps saying "ERROR: Erase Failure\High" then "data corrupt" "see good.bin and bad.bin". Don't think the NAND will let itself get flashed. Or should I say the FW won't let the NAND get flashed with the high erase counts (PE).

I need RyderOCZ to supply me a special FW to overlook the high writes and just flash! LOL

[Anvil]

I've also read somewhere that the temperature of that SSD is a bit high, hence not recommended for small laptops where cooling might be a problem.

It's not impossible, if the drive is packed with NAND it will produce heat (my Intel 600GB can get hot) and the Samsung is on the small side, really small.
I haven't checked my Samsungs yet but will do.

Yes. Perhaps it's a straight average? Maybe it's pe cycles / 10? Perhaps it's actually in Klingon.

I don't see how it's possible to have over 1000 times the drive's capacity with slightly over 100 PE cycles. Maybe it's magic.
...
EDIT
If you look at the whole wear leveling values for the 830 you get 97/97/101. It's 73/73/838 for the M4. Maybe I should add some more static data to see what happens.

You should keep some 15-20GB of writes per loop (or thereabouts) on the 64GB.
I'm using 12GB free space as a minimum on both drives.

The PE counter (177) looks OK, it's just on a different scale, if you get the chance you could try catching/creating a screenshot when the attribute changes and compare to F1.

[minpayne]

It's not impossible, if the drive is packed with NAND it will produce heat (my Intel 600GB can get hot) and the Samsung is on the small side, really small.
I haven't checked my Samsungs yet but will do.

My Intel 600GB is a lot cooler than my M4 512GB, when both running the Endurance testing in AST at around 80-100MB/s fill. The coolest one is the Intel X25-E 64GB - my hand cannot even sense that it is powered on under stress test.

I'm now thinking about getting a Samsung 830 for my laptop but I'm concerned with the heat I've read so far.

[Christopher]

Anvil,

I was using 12GB min free, but from 16 down to 12 average writes dip into the 77MB/s range after a few loops. That's one of the reasons why I was thinking of adding some more static data.

[Christopher]

Samsung 830 64GB Update, Day 8
FW:CXM01B1Q

GiB written:
70706.50

Avg MB/s
107.86

Per-day average:
9088

PE Cycles
116

Reallocated Sector Count
16384 (8 blocks)

194 hours





I was watching last night when the PE cycles hit 110. The CV/worst value hit 96/96 as well. That would mean it should hit 0 at 2750 Wear Leveling Count.

1 in the raw value roughly equals 609GiB, so 2750 * 609GB would equal 1,676,219.83GB

Yup -- 1.68PB to PEexhaustion? I don't think so. I'll keep trying to figure out what it means.