SSD Write Endurance 25nm Vs 34nm

[Anvil]

@stevecs
What workload are you talking about?
If there are loads of small random writes then you need to buy a drive like the X25-E or the new 710 Series.
The 710 Series is priced at ~$700-$750 per 100GB and is available in 100GB, 200GB and 300GB capacity.
(price based on 4K NOK ex.vat for the 100GB and a 5.5 exchange rate)

[stevecs]

Actually, I /WAS/ waiting on the 710 series for this, it's just that I'm antsy and wanted something sooner considering it seems they keep pushing things back in release schedule right now I can't even find a solid release date for the 710's. The E's are such old technology and frankly don't hold up really to that type of workload I would have (4KiB writes generally and then block reads as that write cache is flushed to the main storage system). So not too friendly for SSD's but the latency is the key here so it's either SSD's or battery-backed ram;

[Anvil]

I've heard within a month or so, as the price has surfaced the last few days I expect there is hope for a such a timeline.

[johnw]

The Intel 710 series is SATA 3Gbps.

[johnw]

127.012 TiB, 355 hours, sa177: 1/1/10535

136.389 TiB, 380 hours, sa177: 1/1/11233

Average speed reported by Anvil's app has been steady at about 112MB/s.

The other two unknown SMART attributes, 178 and 235, are still at 72/72/276 and 99/99/2, just as they were when the SSD was fresh out of the box.

[Anvil]

The Intel 710 series is SATA 3Gbps.

Didn't look at the interface, guess we should get used to Intel releasing 3Gb/s drives

[B.A.T]

Eveningupdate:
Due to an power failure my pc got shut down this evening. When I startet up again Anvils app startet from 84 TiB (from when I updated to the latest ver) Anvil is helping me to fix it so the log is correct. This evening you just have to enjoy my smartdata.
1635 P/E used ~95 TiB. Speed before the power failure was 88.36 MiB/s, AD is down form 48 to 46. No MD5 errors before the shutdown.

I'll start up again when it's ok.

[Anvil]

Every loop is saved in a second table so BAT's issue was fixed.

--

159.08TB Host writes
MWI 13
Reallocated sectors, 6.
MD5 OK

[bulanula]

Do you guys think the Intel 520s will use Sandforce controllers ???

ATM Sandforce I is to be avoided because of cold start bug and other problems people are having with these drives suddenly dying and write throttling as well. Also, the NAND used is of arguable quality and the various bait and switch ( 25nm and Spektek etc. ) methods used by vendors is not reassuring.

I don't know about SF II but I doubt much has improved since I see loads of people having issues there as well.

If you want the most usable SSD for daily use then I think this thread has shown that you cannot go wrong with the Intel, Crucial and the Samsung.

What do you guys think ???

Maybe someone can setup an automatic reboot and re-secure erase and repeat etc. script so we can test the secure erase endurance too without needing manual input ??? It should not be too hard to do on Linux with an old machine etc. Too bad I don't have the SSD or the cash to dedicate for one because I could get this old machine and the script ready. Maybe someone else can try on their own or help me do this ???

I think it really is important we also test this aspect of endurance as this really pushes the SSD to its limits as it is like writing to the whole capacity of the SSD in about 10 seconds and uses one NAND cycle on all of the SSD's cells etc. !

[zalbard]

Intel should stick to Intel controllers... I don't want anything SF based at this point.

[boostdfd3s]

Has anyone started testing a 240GB Vertex 3 yet?

[Computurd]

Has anyone started testing a 240GB Vertex 3 yet?

lifetime write throttling would make it impossible.

[johnw]

Has anyone started testing a 240GB Vertex 3 yet?

If you have one with 2500 power-on hours, I'd be glad to test it for you.

[lowfat]

Has anyone started testing a 240GB Vertex 3 yet?

Who would ever waste a $400 drive just to see how quick it would die?

[dr_sharp]

Who would ever waste a $400 drive just to see how quick it would die?

Don't ask questions! Just appreciate those willing to help!

[B.A.T]

Morningupdate:
My problem from last night was effectively fixed by Anvil and now everything is going as normal.
97.2497 TiB. 332 hours (the downtime was 2 hour) Wear Leveling Count and Percentage of the rated lifetime used has gone from 46 to 44.
Avg speed reported from anvils app is 90.42 MiB/s. Looks like the 2 hours break made my speed improved some.
MD5, no errors.

[bulanula]

Does nobody think the secure erase endurance testing is as important as what we are doing now ???

Actually, I think it is more important because it stresses the SSD and its cells much more than the continuous writing we are doing now etc.

[Anvil]

@bulanula

No, at the moment this is the test that's going on and it will continue to run for quite some time, so, all my resources are bound to this test method for quite some time.
You don't need to ask every other post, I do read every word that is written in this thread and you are the only one pushing that test "pattern".
As One_Hertz mentioned, it will most likely be a manual process and that would make the test time-consuming.

--

160.46TB Host writes
MWI 12
Reallocated sectors : 6

MD5, no errors.

[bulanula]

As One_Hertz mentioned, it will most likely be a manual process and that would make the test time-consuming.

I am pretty sure that with an old machine running Linux I can make this work easily without need of any manual input so that is not a problem as far as I'm concerned. I just don't have the bucks for a SSD ATM so maybe someone else can do this test ???

[sergiu]

I am pretty sure that with an old machine running Linux I can make this work easily without need of any manual input so that is not a problem as far as I'm concerned. I just don't have the bucks for a SSD ATM so maybe someone else can do this test ???

I just want to point that, even thou the test would be interesting, it does not have any relevance compared to a real life test. Testing random 4k writes would simulate a heavy OS paging while testing continuous erases does not have similar real usage pattern. And this might be implemented in a different way from one manufacturer to another. For example, if I would produce SSD firmware, I would definetly add an "if" statement that would skip erasing blocks that are already erased. Writing a few MiB/GiB of data before each erase also does not give you any warranties that you will hit all blocks because of the way different controllers choose to cluster pages.

[johnw]

I am pretty sure that with an old machine running Linux I can make this work easily without need of any manual input so that is not a problem as far as I'm concerned. I just don't have the bucks for a SSD ATM so maybe someone else can do this test ???

Rather than posting about it twice a day, maybe you could use the time to earn some extra money and do it yourself.

[Vapor]

C300 Update

50.51TiB, 83 MWI, 850 raw wear indicator, 61.8MiB/sec, MD5 OK.

[Vapor]

Updated charts


Host Writes So Far




(bars with a border = testing stopped/completed)


Raw data graphs

Writes vs. Wear:


MWI Exhaustion:


Writes vs. NAND Cycles:



Normalized data graphs
The SSDs are not all the same size, these charts normalize for total NAND capacity.

Writes vs. Wear:


MWI Exhaustion:



Write-days data graphs
Not all SSDs write at the same speed, these charts factor out write speeds and look at endurance as a function of time.

Writes vs. Wear:


MWI Exhaustion:

[zads]

Do you guys think the Intel 520s will use Sandforce controllers ???

ATM Sandforce I is to be avoided because of cold start bug and other problems people are having with these drives suddenly dying and write throttling as well. Also, the NAND used is of arguable quality and the various bait and switch ( 25nm and Spektek etc. ) methods used by vendors is not reassuring.

I don't know about SF II but I doubt much has improved since I see loads of people having issues there as well.
...

Sandforce really depends on the SSD manufacturer (OCZ, Kingston, Patriot, etc) to enforce their own quality testing and production procedures. While this limits Sandforce's financial risk and investment costs for production and such, that sword also cuts the other way- it opens the door for these drive manufacturers to release improperly validated designs that are then back-associated to Sandforce's quality themselves.

But in all honesty, knowing all that I know, if I wanted to buy an SSD today,
I would still buy OCZ Sandforce-based SSD drives (3 year warranty)..
they are typically best bang for your buck.
And I wouldn't be one of those that complain about getting a 25nm vs 34nm drive..
As with any type of mass storage, just make sure you back up your vital data..

Has anyone started testing a 240GB Vertex 3 yet?

Not "Vertex 3" but does 256GB SF-2582 count?

Who would ever waste a $400 drive just to see how quick it would die?

OEMs

Does nobody think the secure erase endurance testing is as important as what we are doing now ???

Actually, I think it is more important because it stresses the SSD and its cells much more than the continuous writing we are doing now etc.

Its quite the opposite, actually.
Secure Erase stresses the cells less than continuous writing,
unless you are talking about drive that has Mil-spec Secure Erase features (and none of you have that)

[johnw]

136.389 TiB, 380 hours, sa177: 1/1/11233

145.757 TiB, 405 hours, sa177: 1/1/12031

Average speed reported by Anvil's app has been steady at about 112MB/s.

The other two unknown SMART attributes, 178 and 235, are still at 72/72/276 and 99/99/2, just as they were when the SSD was fresh out of the box.