SSD Write Endurance 25nm Vs 34nm

[groberts101]

Think what you could do if you were actually trying!



Two things:

1) Probably best to start another thread

2) You might try reading what others have written and respond to people. You still have not responded to Ao1's data about how his Sandforce drive was lifetime throttled to 7MB/s sequential write speed, and it stayed throttled despite several secure erases. Clearly, SE does NOT clear a lifetime throttle condition on Ao1's SSD.

You have no idea how right you are until you read my posts over at OCZ forums in the mega thread titled "TRIM, OP, and everything else Sandforce". The newer "Sandforce write throttling" sticky also has many to choose from as well. lol Seriously though.. it's pretty tough to just tell someone(especially "jaded experts" like some here) that.. "I just know these things". The main puprose was to add some type of credentials for how I got to where I am at this point with my understanding of Sandfocre. Aside from that, I'm a pretty humble dude and you can take it any way you want which seems to be the downside of any forum posting with huge losses in interpretation and attitude.


1)probably won't happen unless someone PM's me and needs the info. Seems to be just a passing debate and to be perfectly honest?.. I'm fine with guys being wrong and believing what you want to. Seems like too much hassle to "debate" about things that have been proven many, many times over at the OCZ forums on a near daily basis. The info is there. Not being snotty at all and just seems like a waste of time if the minds are closed or the cup is already full. Whatever analogy fits there gets the point across I'm sure.

2)I can certainly do that and will try to shed light on what I'm guessing happened with a few different possibilities there. Seems moot at this point though with the drive out of the running now.

EDIT: Holy friggin' crap! I give up trying to multi-quote bits and pieces of certian posts as I'm just hacking this stuff up trying to get it right. Don't know how you guys do it and don't have the patience to learn something new right now unless someone can lend me a clue or I'll just stay simple and try and make my red responses to be clearly just my own responses to each portion of the original post. If that's not good enough then I'll just go back to spectating again. This back and forth stuff is half the reason I didn't respond when the drive dropped out of the lineup in the first place.

PS. @Ao1.. PM me if you want me to clarify some things a bit better and give some idea's as to what most likely happened there with your drive.

since you don't believe my input here I'll leave you with these quotes from Tony himself.

During the time there is no GC map...EG all the nand is reset and free to write to duraclass has no real impact on Nand life...its just writes

When you SE the controller writes without Duraclass, it writes faster BUT it also writes indiscriminately...it does not look at what blocks have high erase count and move to ones that have a low count... it just writes to them.

you can see the clean speed (NO GC MAP) which is what you get now with a clean new drive or after SE and then the functional 1 month in speed which is what you all see now

idle time will restore the drive to day to day speed, SE will restore unmapped speed but you will see a transition from that state to day to day. Plus SE to often and you actually start wearing out the drive.

So, in the end here?.. Sandforce Lifetime throttle = "settled in state". An SE does in fact reset the drive to fresh out of the box speed. Anyone who find's otherwise has not properly sent the ATA secure erase command for the drive to initiate it at the firmware level?.. or there is an issue with the drive and the firmware is corrupted or an RMA is required. My guess is that Ao1 didn't use the proper tool to send the command to the firmware for the ATA SE to be initiated.

[Computurd]

i would be interested to hear what you think can be done to bring his drive back into the testing. maybe he just has a bad drive, that has crossed my mind a few times.

I personally can vouch for Groberts101. i have seen him over @OCZ for, it seems, years, and he is always right in the thick of the testing, technical aspects, and chasing of new info/ideas. We cross paths frequently on several forums, and i consider him a guru. believe me, its a short list of people that i would consider storage gurus. Luckily, most of them are here, and im glad to see groberts here, so lets not beat up the new guy too bad, eh? we will be glad to have him aboard. One doesnt have to know everything to be a guru, they just have to strive to know and learn everything and truly have a passion for this type of stuff, and invest far too much personal time testing and chasing that...well ya know;...stuff. we definitely have a good share of these guys here, thats for sure. so one more, its a party

this truly is a glorious thread, i hope this endurance testing goes on forever

[gojirasan]

I don't understand all this vouching for people stuff. Is this some kind of pissing contest to see who has the most impressive credentials or something? Unless he literally has inside information as in he is a Sandforce employee or at least an OEM customer of Sandforce I don't see how it's relevant. Argument by Authority is a logical fallacy. After all we're not trying to start a religion here. Data is data. Facts are facts. If you've got 'em then I don't care if you are a janitor or Albert Einstein. Just show me the data.

groberts101, if you can't be bothered to post your data that's fine. Ao1's data seems perfectly valid to me and it sounds to me like you have some mistaken ideas about Sandforce controllers. Frankly I don't trust threads on the OCZ forum very much because the moderators there are often pretty busy deleting and trimming information they don't like or don't want to be public. They even tried to hide the whole life/warranty throttling issue at first. I do respect the fact that they eventually admitted to the throttling, but they were quite evasive about it for a while, until Tony decided to spill the beans.

As far as quoting with BBcode, it's kind of like html tags except you use brackets. So you can multi-quote manually by just copying and pasting the text you want. I'll substitute parenthesis for brackets. It's like this: (QUOTE="groberts101")stuff groberts101 wrote(/QUOTE). If you have javascript active you can use a button above the text box to put the quote tags around the text you copied. There may be an easier way, but that's more or less how I do it when I want to reply to just part of someone's post.

[bulanula]

Really close now between the Intel 320 and the old Intel X25-V. Well, at least according to the MWI that is !

[johnw]

groberts101, if you can't be bothered to post your data that's fine. Ao1's data seems perfectly valid to me and it sounds to me like you have some mistaken ideas about Sandforce controllers.

I was just about to post saying something like that, but you beat me to it.

Ao1 described how he did the secure erase, but it seems groberts101 cannot be bothered to read it. He seems too busy listing his credentials and rambling on than in actually reading what other people have done in their own tests. Anway, Ao1's test sounded like a valid secure erase to me.

So the question remains, if groberts101 is so knowledgeable about SF SSDs, why did Ao1's secure erase not reset the lifetime throttling?

[johnw]

I personally can vouch for Groberts101.

groberts101's claim that Ao1 does not know how to secure erase his SSD kind of reminds me of the guy who claimed that you did not know how to properly plug-in a modular PSU cable.

[Computurd]

i will say this: that you SE the drive to regain the unthrottled performance seems to be 'common knowledge' @ OCZ forums. whether that is true or not remains to be seen. I do think that A01 might try an alternative method of SE the drive though, just to be sure. preferably a method that delivers a completion dialogue, not sure if he did receive that or not. I do know that the independent version of the utility works very well. I believe that he used the toolbox version.

As to vouching for him, what i am saying is this: he is what he says he is. He does Beta Test for OCZ> i know this for a fact. they did not give him the new drives to beta test because he was a dude who fell off the turnip truck. his work with the previous gen of SF was sure to be a basis of this. very very very few people are given that opportunity over there, probably less than one percent. so * I imagine* they sure thought enough of his knowledge/qualifications to justify that. and his knowledge that got him there was the work with the very drive/controller we are discussing.
I dont see many people here who have been allowed to beta test pre-release SSDs...especially the newer generation of the same ing device we are talking about....so you would think he would have some idea of what he is talking about.

why did Ao1's secure erase not reset the lifetime throttling?

dunno. could be a bad drive. if i know LOL

im no fan of SF, i just have never thought much of compression, too much smoke and mirrors for my taste. so i do acquiesce to those who know more than me on that.
doesnt mean they arent good drives, but this whole floating point of performance and some of the problems they've had, just not my cup of tea.

[Vapor]

As to vouching for him, what i am saying is this: he is what he says he is. He does Beta Test for OCZ> i know this for a fact. they did not give him the new drives to beta test because he was a dude who fell off the turnip truck. his work with the previous gen of SF was sure to be a basis of this. very very very few people are given that opportunity over there, probably less than one percent. so * I imagine* they sure thought enough of his knowledge/qualifications to justify that. and his knowledge that got him there was the work with the very drive/controller we are discussing.
I dont see many people here who have been allowed to beta test pre-release SSDs...especially the newer generation of the same ing device we are talking about....so you would think he would have some idea of what he is talking about.

What somebody does off of XS, while sometimes interesting and maybe even important, does not excuse behavior on XS. I hope groberts101 will stop disrupting this thread with discussion about himself and will hopefully be able to contribute with information or direction as to what is going on with Ao1's SF-1200 drive.

[johnw]

i will say this: that you SE the drive to regain the unthrottled performance seems to be 'common knowledge' @ OCZ forums.

More like 'common confusion'. In this very thread, there is one OCZ employee claiming that a SE does reset lifetime throttling, and another OCZ employee claiming that OCZ has never said that SE will reset lifetime throttling. Does anybody really know what time it is?

[Computurd]

What somebody does off of XS, while sometimes interesting and maybe even important, does not excuse behavior on XS. I hope groberts101 will stop disrupting this thread with discussion about himself and will hopefully be able to contribute with information or direction as to what is going on with Ao1's SF-1200 drive.

well yes, that would be nice. im not saying anything about behavior or what-not. thats none of my buisness
did seem like he was trying to provide some info in post #504 (some will miss it since its in the quote box) and then got sidetracked.

More like 'common confusion'. In this very thread, there is one OCZ employee claiming that a SE does reset lifetime throttling, and another OCZ employee claiming that OCZ has never said that SE will reset lifetime throttling. Does anybody really know what time it is?

yea thats why i put 'common knowledge' in those quotation thingys.

EDIT: now ive learned the multi-quote. that helps. never used it before myself just copy/paste. never again. there is a YouTube video on it

[Ao1]

What somebody does off of XS, while sometimes interesting and maybe even important, does not excuse behavior on XS. I hope groberts101 will stop disrupting this thread with discussion about himself and will hopefully be able to contribute with information or direction as to what is going on with Ao1's SF-1200 drive.

Thank you Vapor!

The whole point of this thread was to remove speculation and fud. Instead it now seems to have reverted to even more fud, when the facts clearly say something different.

A V2 40GB drive can be picked up for pocket change. If anyone doubts my observations why not pick one up? Ask Anvil for the app and then run it for 7 days. Then we can have a discussion.

If anyone wants to tell me how to clear the throttling on my V2 please post details for the benefit of everyone and I will try whatever is suggested.

@groberts101

First you accuse be of believing fud from Tony, next you quote fud from Tony as being the truth. I am not making observation on what I read. I am making obsevations on what I found out for myself by testing.

What I don't get is this; The DuraClass feature is working EXACTLY as intended, but you do not want to believe it.

Why not join us and do some testing on a V3?

Run some large xfers and monitor I/O performance during the xfers.

Next ask Anvil if you can use his app and see how long it takes to get to a throttled state that cannot be recovered.

Next come and put facts (not speculation) on the table, as I have tried to do in this thread.

EDIT: If you are beta testing for OCZ maybe the throttling options were not implimented? Do you have a V3 that you purchased? If so it would be great if you could use a retail V3 to contribute to this thread.

[Ao1]

I do think that A01 might try an alternative method of SE the drive though, just to be sure.

Comp, I tried a number of SE's using the OCZ Toolbox. I even tried a hdderease in a different PC. After every SE the partition was gone. What elese can I try?

What zads states in post #523 makes perfect sense:

Its lifetime/warranty wear levelling,
Can't be cleared by secure erase or even firmware upgrade, can only be cleared by a base-level manufacturer's firmware load.

Praz implied that in this thread.

Has anyone on the OCZ forum tried to write for 7 days non stop?

[Computurd]

well the only suggestion i would have is to try the independent tool. but i do feel that you used the tool correctly, and probably 99 percent chance that it worked correctly. but there is always that one percent chance. not questioning your abilities/knowledge by any means at all!
its just that i have heard people say that they needed to use the actual SE separately to get a good result. but that is probably a corner case on old boards...dunno dont pay much attention to it.

I dont think anyone has done any extended testing on these drives at OCZ, but im not sure. I dont hang out there anymore, i dont even lurk. a personal disagreement led to me leaving those forums pretty much. every week i will check PMs there though, as there is still interest in the 9260 thread over there.

might also be a bad drive. man i would love to see another test, it just seems unconscionable that they would throttle to that extent. amazing.

Zads seems to be very knowledgeable, and claims to be a SF vendor according to this: (post 523)

Okay I qualify here-
Its lifetime/warranty wear leveling,
Can't be cleared by secure erase or even firmware upgrade,
can only be cleared by a base-level manufacturer's firmware load.

so who knows, surely not I.
As i tried to say up above, i know Groberts and hes a stand up guy, but that doesn't necessarily mean that his interpretation of his and others results are correct.
i do consider him to be very knowledgeable though.

i know i am definitely not standing up and claiming a SE will restore anything with duraclass. i have absolutely no idea

and i agree with the amount of differing opinions it is really muddy water there.

[Ao1]

How do I go about doing a SE with the independent tool? I am willing to try it. I will then restart the endurance test to see what happens.

It would be great if someone else could get a V2 40GB to see if what I found can be duplicated. Then it's not just me saying it.

I can assure everyone on this thread that what I reported is 100% accurate to what I have observed. I have no interest in fabricating what I observed. Why would I report fud and end up looking stupid if (when) I got caught out?

If I observe something that I can't explain I state clearly that I am speculating on anything I put forward as a possible reason.

[Computurd]

good god man, no one is questioning your integrity!
i dont think it is publicly available anymore. let me email it to you
other than your maddening assertions about SSD/HDD effect on gaming, ive never questioned you

YHPM

[Anvil]

Hmmm....

First, Ao1 has tried SEing the drive both using OCZ Toolbox and HDDerase and so there is pretty much 0 possibility that it's not being done the right way.
(as he was able to upgrade the firmware we can pretty much say that the Toolbox was working)

I'm actually surprised that it lasted for a whole week with continuous writing, I do not believe anyone has made such a test using a SF controller, well, not publicly

I'm sure groberts can be an asset on the SF drives, I'm pretty sure that this is uncharted territory though and so we need more SF drives in the test.
All other aspects about the SF controller is beyond this test and has to go into a new/separate thread.
I'll make such a thread if no one else will, however the SF controller makes for a lot of speculations and apparently none of the info that we are looking for will be made public and so whatever we do, it will be speculations and by our own trial and error.
SF should demystify some of the stuff we are trying to figure out, it's our drives we shouldn't need to be asking questions like these.

[Ao1]

Anvil, I'm waiting for the tool from Comp. When I get it I will restart the endurance test, but I will start a new thread.

@groberts101

Welcome to the forum. I hope you will join me in the new thread. It would be great if you could do some testing on a retail V3, or duplicate what I did on a V2 to see if the same thing occurs.

If I could start over I would have checked I/O performance for various sized xfers to quantify the I/O burst throttling effect.

It would be interesting to establish if MAX IO means this particular method of throttling is disabled when compared to a standard version of the V3.

After that is established it would then be interesting to see how long/ how much has to be written on a V3 before throttling kicks in. Does capacity result in different outcomes? i.e. are smaller sized drives throttled more than larger capacity equivalents?

[bulanula]

Anvil, I'm waiting for the tool from Comp. When I get it I will restart the endurance test, but I will start a new thread.

@groberts101

Welcome to the forum. I hope you will join me in the new thread. It would be great if you could do some testing on a retail V3, or duplicate what I did on a V2 to see if the same thing occurs.

If I could start over I would have checked I/O performance for various sized xfers to quantify the I/O burst throttling effect.

It would be interesting to establish if MAX IO means this particular method of throttling is disabled when compared to a standard version of the V3.

After that is established it would then be interesting to see how long/ how much has to be written on a V3 before throttling kicks in. Does capacity result in different outcomes? i.e. are smaller sized drives throttled more than larger capacity equivalents?

Indeed the only reason I did not go for a Vertex 3 is the lack of transparency.

Statements like "There is less throttling on the V3 compared to the V2" are just not good enough when we saw a V2 drive hitting a maximum of 7 MB/s in this thread !

We need facts and figures not just opinions without proof. Eg kind of reminds me of statements like "bulldozer is competitive with 2600k" etc. how competitive and how much less throttling etc.

[One_Hertz]

142TB. 27%. Reallocated sectors up to 11 from 9!

[Anvil]

107,36TB Host writes
MWI 41

Still no changes for reallocated sectors.

So, 107TB, that is more than 3x the warranted 35TB.

[gojirasan]

Do we actually have a p/e spec on the Kingston SSDNow or the Intel X25-V? I know that 5000 p/e cycles is assumed for 34nm, but that is just a general industry assumption. I am wondering what Intel has to say on the matter. That warranty is truly pathetic. Is that the Kingston warranty, the Intel version, or both? Were they really expecting no more than 900 p/e cycles from their 34nm drive? That's a lot of write amplification. IIRC, Intel publishes the same amount of lifetime writes for all of the drive sizes in a model line which of course makes no sense at all.

[Anvil]

"Up to 20GB Host writes per day for a minimum of 5 years"

So, 35TB is the spec for Intel on the X25-V.

Link to Datasheet

edit:

I have drives that have been used for about one year and they still haven't written 1TB, so, 20GB per day is actually quite a bit.
In this case, 20GB per day is 50% of the total capacity

[gojirasan]

Thanks, Anvil. So it's just as I thought. They don't even bother putting a p/e spec in their data sheet. That seems typical of Intel unfortunately. In this case it's not so bad because they are not selling the flash memory separately. I do find it annoying that IMFT doesn't seem to want to make their p/e specs public though. Based on their warranty it seems Intel has a very low opinion of the write endurance of their drives. It seems unlikely they would assume less than 5000 p/e c. for a 34nm chip so it looks like they are assuming a write amplification of about 5.5. Either that or they are, um, bending the truth for some reason. Some people, including IMFT IIRC, say that FUD about write endurance may be reducing the adoption of SSDs, but here we see Intel themselves implying that their drives will only write about 35 TB before suffering a premature death. Bullet, meet Foot.

[Computurd]

So, 107TB, that is more than 3x the warranted 35TB.

interesting to me that they would only warranty 35TB, yet the smart data still tells us there is plenty of endurance left. you would imagine they would have adjusted the smart data accordingly, matching the data with their projected endurance.

either way, i am just floored. So impressed with the performance of these intels.
would love to see a C300 done. seeing the strong relationship they have with intel through IMFT, and the fact that they are using the same nand, i wouldn't be surprised if the performance mirrored that of intel.

[johnw]

would love to see a C300 done. seeing the strong relationship they have with intel through IMFT, and the fact that they are using the same nand, i wouldn't be surprised if the performance mirrored that of intel.

A 64GB Crucial m4 should be able to wear out its flash memory in fewer hours with Anvil's program than an Intel 320, since the m4 can write at about 100 MB/s, as compared to the 40GB Intel 320 which is limited to about 45 MB/s. So a factor of 2.2 faster writes, but only a factor of 1.6 more flash memory.

A 64GB Samsung 470 would be interesting to test, also, but I do not know how Samsung flash compares to IMFT flash. The interesting thing is that a 64GB Samsung 470 can write at 170 - 200 MB/s.