SSD Write Endurance 25nm Vs 34nm

[Ao1]

Has anyone looked at SMART atributes on the 830 using GsmartControl? (Ref this post that revealed "hidden" info)

[Anvil]

I'll test one of my 830's shortly, just need to find the drive and download/install gsmartcontrol.

[Anvil]

[Ao1]

Thanks Anvil If you hover your mouse over an attribute in the GsmartControl GUI does a pop up appear with more info? (I think it is a Java based script, so Java might need to be installed for it to appear).

Edit: I see that B1/ 177 is supposed to update continuously. Hmmm.

Does SSD Life predict the MWI?

[B.A.T]

Todays update:
Kingston V+100
It dropped out again. I'll restart it later today.


Intel X25-M G1 80GB
190,9270 TiB
19953 hours
Reallocated sectors : 00
MWI=150 to 148
MD5 =OK
47.22 MiB/s on avg


m4
156.4539 TiB
570 hours
Avg speed 80.87MiB/s.
AD gone from 18 to 09.
P/E 2749.
MD5 OK.
Reallocated sectors : 00


Almost at MWI=1% now

[Vapor]

Hey everyone

Had to deal with a family situation since early October. I flew to Connecticut and took care of a lot of things there for the past few months (the 10 day power outage was fun, not) and have only been at my place in Michigan for ~14 total days since early October (and those were awfully busy days). Fortunately everyone is okay (enough) now and I'm back in Michigan indefinitely. In the next few days I hope to try to catch up on the charting and everything. I may just veg out and play video games though...it will be nice to just relax for once.

To those who emailed me while I was away, thank you for your support and I'm sorry I never returned emails (I brought a laptop, but its battery and keyboard don't work, so I was basically PC-less for past few months).

I haven't so much as browsed XS since early October but this thread still looks very active so I bet I have a ton to catch up on in this thread alone

At some point along the way, the PC testing my C300 and SF-1200 nLTT restarted. The SF-1200 nLTT is no longer recognized and the C300 still is (but it hasn't been in-testing for an unknown amount of time). The SF-1200 may just need some TLC to get recognized again (maybe even just a simple power cycle).

Hope everyone is doing well

[Anvil]

Welcome back Vapor

Great to see that you are doing well!
We have missed boh you and your charts and there have been numerous request for the C300 in particular.
(and then there is the attachment situation...)

Anyways, looking forward to an update on the charts (when you are ready!)

[B.A.T]

Welcome back Vapor! You have been missed

[bluestang]

Welcome back Vapor!

I think the vegging out for a little while sounds well deserved

[Anvil]

Kingston SSDNow 40GB (X25-V)

599.49TB Host writes
Reallocated sectors : 05 19
Available Reserved Space : E8 99
POH 5422
MD5 OK

33.63MiB/s on avg (~29 hours)

--

Corsair Force 3 120GB is on a short retention test, it's 29 hours into the "test".

Next update will be next year
(the X25-V almost made 600TB during 2011, by midnight it will be about 130GiB short of 600TiB)

[B.A.T]

Last update of the year. 6 month of testing is over and 12 new can start
Kingston V+100
317.7995 TiB
7995 hours
Avg speed 24.53 MiB/s
AD still 1.
168= 1 (SATA PHY Error Count)
P/E?
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
193,7977 TiB
19970 hours
Reallocated sectors : 00
MWI=148 to 147
MD5 =OK
45.87 MiB/s on avg


m4
161.5540 TiB
588 hours
Avg speed 78.99MiB/s.
AD gone from 09 to 06.
P/E 2837.
MD5 OK.
Reallocated sectors : 00


Happy new year everyone

[bulanula]

Happy new year to everyone and welcome back Vapor !

Good news to hear the C300 will hopefully be back in action so we can see if it is indeed better than the M4.

[Ao1]

Happy New Year Everyone.

Anvil, it would be cool to a have a graph that is normalised to reflect differences in the (raw) drive capacity and theoretical P/E cycle capability. This would highlight the ability of the controller to limit write amplification and/ or benefit from compression.

For example:

• Kingston (Intel) X25-V 40GB x 5,000P/E = 200,000 GB. P/E expired at 180,000 GB
• Corsair F3 120GB x 3,000P/E = 360,000 GB. P/E expired at 438,000 GB

The F3 has a distinct advantage, although in this case it was using compressible data.

As another example:

• Kingston (Intel) X25-V 40GB x 5,000P/E = 200,000 GB. P/E expired at 180,000 GB
• Kingston (JMF618) V100 64GB x 5,000 = 320,000 GB P/E expired at 196,000 GB

A huge difference, that is not readily evident from the graph.

It also seems that the X25-M G1 drive is really suffering due to a lack of TRIM.

By the way I have asked Samsung support if they can confirm if 177 is the P/E count. I’m not holding my breath for an answer. Regardless I’ve got one on order, although it will be annoying if I can’t determine when the theoretical P/E count expires.

The more I think about the more I think data retention is the true Achilles heel of SSD rather than write endurance.

Vapor, was your SF drive left unpowered? I can’t remember how much you had written past MWI 10, but as you mention that the drive no longer works presumably it lost its ability to retain data in just under 3 months.

[Anvil]

Happy New Year!

@Ao1
Will have a look at a that graph.
The Intel G1 is really looking bad, I noticed the extremely low P/E count at MWI exhaustion, would be interesting to see what a G2 would do without TRIM.

[Ao1]

This is not normalised, but it shows the difference between theoretical and actual writes required to deplete the P/E cycles.

I can’t recall what Vapor used for compression, but presumably the workload was a lot less compressible than the 46% used for the F3.

I’ve guessed the 830 actual writes, but for sure Samsung SSD’s are really bad at controlling WA, followed by the JMicron/ Toshiba controller. The impact of a lack of TRIM on the Intel G1 drive is also savage. It would be interesting to test a G2 without TRIM.



Now...how to work out normalised values for actual write capacity.....

[Ao1]

OK, a normalised chart. I’m not sure how accurate this would be. I normalised the theoretical write capacity based on all drives having a capacity of 120GB with 5,000P/E cycles. I then multiplied the difference between non and normalised values of the theoretical write capacity with actual write values to get a normalised actual write value.

[Anvil]

Kingston SSDNow 40GB (X25-V)

602.48TB Host writes
Reallocated sectors : 05 19
Available Reserved Space : E8 99
POH 5449
MD5 OK

33.21MiB/s on avg (~56 hours)

--

Corsair Force 3 120GB

01 120/50 (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 617211 (Raw writes) ->603TiB
F1 821510 (Host writes) ->802TiB

MD5 --

--.--MiB/s on avg (~-- hours)

power on hours : 2374

--

The F3 has just started testing again having been disconnected for 56+ hours and the X25-V is about to have a short data retention test. (48 hours)

[Anvil]

OK, a normalised chart. I’m not sure how accurate this would be. I normalised the theoretical write capacity based on all drives having a capacity of 120GB with 5,000P/E cycles. I then multiplied the difference between non and normalised values of the theoretical write capacity with actual write values to get a normalised actual write value.

It looks to be an interesting angle of view and the result looks fine to me

[B.A.T]

First update of the year.
Kingston V+100
318.0055 TiB
1687 hours
Avg speed 25.89 MiB/s
AD still 1.
168= 1 (SATA PHY Error Count)
P/E?
MD5 OK.
Reallocated sectors : 00


It dropped out again last night. It's the second time in 24 hours.

Intel X25-M G1 80GB
198,2124 TiB
19998hours
Reallocated sectors : 00
MWI=147 to 146
MD5 =OK
52.72 MiB/s on avg


m4
168.9476 TiB
616 hours
Avg speed 78.09 MiB/s.
AD gone from 06 to 02.
P/E 2965.
MD5 OK.
Reallocated sectors : 00


Almost there. Now the fun begins

[Christopher]

Damn, I'm glad to be home. I'll unpack, then post an update after doing a manual power cycle of the 830 to get current PE values.

I will check out gSmartControl with the 830. Then I'll be prepping a new drive to start testing.

I still would very much like to know what happened to the 830... it's behaviour is not at all what I would expect to happen. I'm hoping a firmware update can fix the problem, but I don't know that Samsung will release one anytime soon.

Good to see you back Vapor!

[Christopher]

Samsung 830 64GB Update, Day 27
FW:CXM01B1Q

GiB written:
175509.38 ASU
176717.9 Total

Avg MB/s
63.65

PE Cycles
11046

Reallocated Sectors
32768
16 Blocks




Okay -- about the PE cycle count -- it initially increased over the first 10 days, but very slowly. Before the power cycle I just performed, PE cycles were the same as its been for two weeks at 6742. It's now at 11046 after the power cycle. I've been away for the past two weeks, so I was unable to power cycle the system.

[Anvil]

The SMART attribute you refer to as P/E cycle count is not the P/E cycle count, the value has been decreasing from 100 down to 1 and it's raw value has been increasing.
The value could be working more like a MWI counter but it's raw value is something completely different.

Have you been logging with SMARTLOG?

[Christopher]

The SMART attribute you refer to as P/E cycle count is not the P/E cycle count, the value has been decreasing from 100 down to 1 and it's raw value has been increasing.
The value could be working more like a MWI counter but it's raw value is something completely different.

Have you been logging with SMARTLOG?

Yes, I have. What is it if not PE cycles? It went from 100 to 1 (I shut down the system after about 10 days). It looks a lot like PE cycles, and the Samsung Magician Software refers to it as a PE cycle attribute. Of course, you can't log the PE cycles since the drive has to be power cycled.

[Anvil]

If the raw value is a P/E cycle counter then WA is close to 4.
GIB written / Capacity in GiB should be close to 3000 and also the P/E count if WA was 1

It is weird that the counter does not change during "normal" usage.

edit:
Could you expand the raw value column next time, Total Count of Write Sectors is not fully shown

[Anvil]

Kingston SSDNow 40GB (X25-V)

Is 22 hours into a data retention test!

602.48TB Host writes
Reallocated sectors : 05 19
Available Reserved Space : E8 99
POH 5449
MD5 OK

--.--MiB/s on avg (~-- hours)

--

Corsair Force 3 120GB

01 94/50 (Raw read error rate)
05 2 (Retired Block count)
B1 49 (Wear range delta)
E6 100 (Life curve status)
E7 10 (SSD Life left)
E9 623290 (Raw writes) ->609TiB
F1 829599 (Host writes) ->810TiB

MD5 OK

106.76MiB/s on avg (~22 hours)

power on hours : 2395

--