SSD Write Endurance 25nm Vs 34nm

[gojirasan]

Don't confuse what Secure Erase is doing at the NAND level with what is happening at the firmware level. OCZ has never stated Secure Erase will reset DuraClass parameters.

Hasn't someone from OCZ stated exactly that in this very thread? And Tony also stated it multiple times in his very informative thread on the topic over at the OCZ forum. I was very surprised and frankly disappointed that ao1 was not able to reset his drive with a secure erase. I was thinking that I could avoid any life throttling by just imaging my drive and secure erasing it every few days.

I'm not much of a gambler, but with people expecting a petabyte of writes from these puppies it makes me think we should be placing bets. Anything more than 6000 p/e c. from a 25nm drive or 10,000 p/e c. from a 34nm would surprise me greatly. Although I suppose it is possible that the wizards at Intel have some kind of magic up their sleeves. They've got a lot of big brains over there.

[johnw]

Hasn't someone from OCZ stated exactly that in this very thread?

OCZ has stated whatever they want to have stated at the time that they are making their claims. Get with the program!

[johnw]

that doesn't mean that it will catch all sorts of "errors" of course.

LOL. That is why the suggestion for testing some data with checksums was made.

[Computurd]

OCZ has never stated Secure Erase will reset DuraClass parameters.

firmware update does? a destructive FW flash definitely would i imagine. im afraid to update the FW on this here vertex LE, its on the original and still fast as heck. but im wondering if the newer FW might have some more draconian/stringent Duraclass parameters written into them.

@ Hertz
I challenge you to a Duel! LOL. i would throw a 500 GB SLC Fusion IO in the air, and as you dive to save "your precious" i would pounce!


@ gojirisan-welcome to the storage area you seem very knowledgeable, good to have ya around!

[Anvil]

83.01TB Host writes
MWI 54

No other changes

(Windows Update did it's job and so there are about 6? hours last night where the app didn't run)

@johnw

CRC is automatically done by Windows when reading files, we are only writing and thats why I suggested adding a few reads per round of writes. It would tell a lot more than doing nothing at all

--

Anyways, MD5 is implemented and we just need to agree on some sort of procedure for performing the test.
The app can copy an MD5 checked file from another drive and perform the MD5 test when the copy job is done, we only need to tell the app what the checksum is and it will do the job automatically.

[Ao1]

My bet is on 250TB max.

[bluestang]

firmware update does? a destructive FW flash definitely would i imagine. im afraid to update the FW on this here vertex LE, its on the original and still fast as heck. but im wondering if the newer FW might have some more draconian/stringent Duraclass parameters written into them.

I don't think a normal non-destructive FW flash will. However, I do believe a destructive FW flash definitely would. I don't think I've seen OCZ supply a destructive flash to a consumer for Sandforce, only Indilinx and that reset everything. Maybe Tony or Ryder would be so kind to supply you with one (if even possible) strictly only for the purpose of this test.

Speaking of Indilinx...I wonder how a Barefoot controlled 50nm NAND SSD would hold up in this test???

[johnw]

@johnw

CRC is automatically done by Windows when reading files,

Do you have a reference for that? I don't think NTFS does checksums on files. The only filesystems I can think of that do checksums are ZFS, btrfs, and NILFS.

Were you referring to internal checksums done by the SSD, with the SSD returning a read error if they do not match? I think those can only detect a certain number of errors per block. If there were too many errors, it is possible that the internal checksums can miss it. The whole silent data corruption issue.

[One_Hertz]

Do you have a reference for that? I don't think NTFS does checksums on files. The only filesystems I can think of that do checksums are ZFS, btrfs, and NILFS.

Were you referring to internal checksums done by the SSD, with the SSD returning a read error if they do not match? I think those can only detect a certain number of errors per block. If there were too many errors, it is possible that the internal checksums can miss it. The whole silent data corruption issue.

This has been my understanding also. NTFS trusts the device to do the reading and writing correctly.

[aintz]

boxing + muay thai + mma

2 bad i am unagi.

i moved my cache back to the ssd. are you happy now, if it dies with in the next year ur fusionio belongs to me.

[bulanula]

Wonder how the C300 would do. Anyone?

[Computurd]

my money says its on par with intel. C300 is UBER.

[Anvil]

This has been my understanding also. NTFS trusts the device to do the reading and writing correctly.

You may be right, it might rely on the device reporting crc errors and just reports it up the chain to the "user".
I've had several HDDs where Windows reported CRC erorrs, still got 1 of them. (ST7200.10 320GB)

This was the initial info I found on crcdisk.sys
"CRCDISK.SYS is described as Disk Block Verification Filter Driver. This file carries out CRC (Cyclic Redundancy Check) on all data, which is read off the disk."

--

85.78TB Host writes
MWI 52

--

What do we do on the MD5 stuff, to early yet? or should I just come up with a test we can try?
- Would mean, pointing to a local file on another drive which will be copied and checked + reporting
- Can be part of the "loop" or performed every n of the loop
Or, do it manually?

[One_Hertz]

113.5TB. 41%.

I wouldn't mind doing it manually once in a while since it is a pretty straight forward process, just time consuming to do. Big file + MD5 check. Move big file to SSD + do another MD5 check. Should match if all is well.

I have some more advanced methods I can use too with some of the tools I have here. Write LBA number pattern into every LBA and then read it back to see if everything is perfectly correct. I'll probably just do MD5 though.

[sergiu]

113.5TB. 41%.

I wouldn't mind doing it manually once in a while since it is a pretty straight forward process, just time consuming to do. Big file + MD5 check. Move big file to SSD + do another MD5 check. Should match if all is well.

I have some more advanced methods I can use too with some of the tools I have here. Write LBA number pattern into every LBA and then read it back to see if everything is perfectly correct. I'll probably just do MD5 though.

Writing each LBA sector would be better than a plain MD5. It would show us at the end how many errors were encountered at bit level.

[One_Hertz]

Writing each LBA sector would be better than a plain MD5. It would show us at the end how many errors were encountered at bit level.

Correct, but I am not expecting any errors. If my MD5 doesn't match, then I will do a more thorough test.

[Anvil]

OK, I'll probably create an option for MD5 testing anyways, not much work left to get such a procedure working.

Pattern testing sounds interesting

[johnw]

What about the static data you have on the SSDs? If you read it back, could you compare it bit-for-bit against what it should be?

If not, what about replacing the static data with a file that you keep another copy of so you can compare it later?

[bulanula]

Glad you guys picked up on this fact.

There will be lots of data loss I am sure !

[johnw]

There will be lots of data loss I am sure !

I'm not sure. In fact, I doubt it. But it would be interesting to test it to know for certain.

[Anvil]

88.44TB Host writes
MWI 51

Re-allocated sectors, unchanged at 6.

[Anvil]

Wonder how the C300 would do. Anyone?

I've found some info on SMART for the Crucials and so it's possible to get "host writes" for the C300 and the m4.
("AD" counts the number of times "all cells have been written to", so a count of 1 = 64GB "Host writes" for the 64GB drive)

The spec says 72TBW, I'd presume it'll do a lot more

[bulanula]

Still it would put my mind clear because I am trying to decide on a SSD. If someone could test a C300 with 34nm I would be very glad !

[MadHacker]

Still it would put my mind clear because I am trying to decide on a SSD. If someone could test a C300 with 34nm I would be very glad !

do u wish to donate one for testing?

[Computurd]

lol send it to me! i have 8 256 GB C300, but they are on an open loan from Crucial, i dont think they would appreciate me killing one! actually maybe i should ask permission to do so. i will see what i can do