[Corsair Extreme x64] SSD write performance degradation / Health is dropping fast

[toyo]

This SSD was a gift, I have strong doubts I will be able to upgrade it very soon, that's the most important reason for me wanting it to last as long as possible. Especially since I have other priorities when it comes to upgrading, like CPU/MB/GPU/PSU/RAM.

Anyway, I found out I have SSDLife installed in my previous OS on the HDD, here's the estimated life for the Corsair:

I obviously have strong doubts it will last that long, but I might install CS5 and Ableton in the end on it... who knows ? :P

[Maxhedroom]

Toyo...I forgot all about this SSD endurance thread at Xtreme Systems. Been a long time since I checked it...interesting stuff. ;)

http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

[toyo]

Since you posted... I'm at 94%, soon 93% (I can feel this, deeply in my guts :P). I decided to do my best and chronicle the life of this SSD for the Internetz. Please check out how the estimated life went down... from 2020 to 2013. I can almost see myself making it to summer 2012 with the SSD alive, if I take good care not to write anything on it :) Also, my NAND is Samsung 51 nm, model: 916 K9HCG08U1M PCB0. I can't find much info on it., I'll take a look on your thread anyway, thanks.

Interpreting SMART Data on Corsair SSDs

[Maxhedroom]

Thats the same nand as my Ocz Apex as well as all the initial barefoot drives...later they went to 34nm. The 51nm has a p/e of 10,000...seems like a ton now but back then wear-leveling wasn't as good as it is now. My Apex doesn't have any garbage collection or trim even so the high p/e of the 51nm is a must...but the barefoot has decent wear-leveling and trim as well as aggressive garbage collection...thats what makes the sudden aggressive falling health of your drive a head-scratcher. Good idea would be maybe to post in that thread with your screens shot and a shot of Crystal Disk Info as well. They may be able to enlighten you more. Not a whole lot of info resources regarding SSD durability cuz it hasn't been near long enuff for even the 1st generation drives (2 1/2 years old approx.) to start exhausting their nand.

Edit: Whoah! Just noticed your total lifetime writes haven't even hit a TB yet!! Something is definitely waay off...the worse drive in that thread didn't start going down hill until ~60TB...latest chart on this page...

http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm/page47

[toyo]

Edit: Whoah! Just noticed your total lifetime writes haven't even hit a TB yet!! Something is definitely waay off...the worse drive in that thread didn't start going down hill until ~60TB...latest chart on this page...

http://www.xtremesys...-Vs-34nm/page47

The drive is a gift and a refurbished one, but that should have 0 influence on the lifetime, refurbished units should perform identically compared to normal units, but I have to mention it for the sake of complete information. I'm not sure, but I think it's refurbished because the drive did not perform correctly and gave errors (I don't know what kind). It was repackaged as new, stamped with a "REFURBISHED" sticker and returned, then it magically found it's way to me :P

When I got the drive, it was already at 98%/400 GB written (approximately).

Corsair would not have ANY opinion on why this is happening, they just plainly REFUSE TO ANSWER.

So you see, it's either I find what's going on on my own, or nothing.

[Maxhedroom]

Ahhh...I see. Unless it was returned my the original owner somehow for a premature wear issue and wasn't screened out by Corsair since it would bench test just fine and receive a QA 'pass' to be resold. Totally speculation ofc. Only thing I can imagine trying would be maybe...maybe...a destructive fw flash of the latest fw. No idea if Corsair even offers such a thing tho. :( Like I mentioned, it may be worth a few mins to post up your issue with the screenies in the endurance thread.

[toyo]

Like I mentioned, it may be worth a few mins to post up your issue with the screenies in the endurance thread.

All I know is that the drive gave constant errors during use and was returned. I can say that it functions without error now. I intend to post in the endurance thread today, I was just to lazy to subscribe to yet another forum, I hate doing that.

I will also try to find out about the destructive flash thing, I didn't see anything like it on Corsair's forums, although I read about it on the net.

EDIT: Let's proceed to 93%. I told you I could feel it coming :P

Interpreting SMART Data on Corsair SSDs

-------------------------------------------------------------------------

No registration/posting on xtremesystems.org for me, ever. They ask for money for subscriptions! Yes, it's just 1$, but even if it was a cent, I'd still not pay for something that is just a frame for members to express themselves. Members build 99% of a forum, not admins and mods, they mostly maintain it.

I hate when people ask for money like that. It completely compromises the quality of their work and their image.

Best thing is to put a Donate button, and whoever wants to push it, it's really great, but never, ever force people into paying. These idiots always forgot that not everybody lives in US/UK/France/Germany etc., there are net surfers from poor African/Asian/European countries too. You probably won't believe it, but that 1 USD might pay for food for your kid today.

[toyo]

Got response from Corsair employee:

You are not being ignored. Read the sticky post here about SMART info as it relates to SSDs. My advice is, if your drive is stable, ignore SMART and enjoy your drive.

I told the guy I can only enjoy tech when knowing how it works and I would like to know a proper way of estimating drive health instead. Full response below:

Thank you for answering. I read your SMART sticky, it's exactly what you can find on the web - SMART is not reliable, but good as a guide.

I'm fully aware that the drive health might not be correctly estimated (but why Corsair has approved a faulty estimation in their drives? It's the SSD that calculates the parameter, and this parameter is not one of those "Unknown" ones).

I am willing to completely ignore SMART, if you provide me a better tool for this job. I think drive health should be properly documented, especially since this is totally possible with NAND, before putting things on the market.

Also, there is this pattern of 1%=100 Write Cycles/40 GB. This cannot be an odd occurrence. Somebody made it so, so there is a reason before it. I would like to know it, if possible.

All these SMART values were put there with a purpose, and are only read by the SMART utilities, so if they are unreliable, it's the SSD's fault. Somebody created the algorithms for these calculations based on some real data, I think, so the typical health of this SSD/NAND is not unknown and it was estimated. I would like to know these values, everybody deserves to know these things.

As for enjoying the drive, this is how I enjoy technology, learning about it and trying to use it at full potential. I'm not comfortable using this SSD blindly, I would like to know what I'm doing and be in control.

I would be very happy if Corsair would give me a proper tool to estimate the life my SSD has left. I would also welcome details on the NAND type and its life. This drive can be considered old, and disclosing internal testing data about it should not be an issue.

No, this is not correct. First, we have not "approved" anything relating to SMART utilities. The only specifications we approve are listed on our product page.

Second, you cannot state what SMART parameter any given utility is using to generate these information you are quoting. Many utilities will report different values for the same parameter. They are NOT RELIABLE.

Then what is that SMART parameter that all SMART utilities I have used said it's the SSD health, and it went down from 98% to 93% in such a short time? Who put it there if not Corsair (or your collaborators, Indilinx, Elpida or Samsung - but still your responsibility, as seller)? Why is it going down 1% every 100 write cycles/40 GB? How come every application reads the same values/parameter, are they ALL WRONG?

Again, please provide a better tool instead. The specs listed on the product page are almost nothing and are certainly not helpful in my case.

I'm sorry, but your posts try to avoid giving a technical answer and also avoid addressing the 1%=100 Write Cycles=40 GB issue. Please, don't tell me all these apps were wrong and their creators idiots.

Even if the apps are wrong, it's because Corsair would not correctly document each SMART parameter publicly. These apps are just reading stuff from the SSD, they don't invent it.

OK, this guy is saying we are all idiots and everything SMART reads is wrong and unreliable. He won't provide better info though, just adopting the classical attitude: "Everything is well, please go back to sleep!"

I need to remember never to buy Corsair products, ever. I will continue to document the drive health and this discussion too, the Internet need to know how bad some businesses are supporting their products.

If you cannot discern the point I am trying to make to you from the information in the sticky post, then I cannot help you. We do not provide SMART monitoring utilities. I cannot tell you how any given SSD reports SMART data. Most likely, this is proprietary info from each but you are welcome to ask Indilinx, Intel, Samsung, SandForce etc. I cannot tell you how any given utility interprets that data. I did not write the utilitiesl. We do not dictate to any SSD controller maker how THEIR controller reports SMART data.

I got it now, thank you.

You're selling (I mean Corsair, obviously, this doesn't have anything personal with you), but don't really know what you sell, or if you do, you won't disclose to the clients the truth about their gear and rather keep them in darkness. This is a very shady business practice and I'll make sure the Internet knows how Corsair support treats their customers.

I will also keep documenting the health, I mean, "unknown" parameter and other SMART data until the drive dies, here (if the thread will not be closed or deleted somehow) and in other forums on the Internet too.

I'm sorry, but people that give you money deserve to know the details about what you're selling. It's not like I am asking for electronic circuitry schemes and how the components are made, isn't it? I'm so tired about the "proprietary" charade, it's not like we want to steal from you, we support the businesses that we love.

I will not bother you from now on, thank you for all your answers.

So, this is it. There will be no info about the SSD and why it reports these values. Instead, Corsair is not taking responsibility for what they sold, and, just like children, is sending customers to ask Indilinx or other component makers.

If you plan on buying Corsair, make sure you know what you're getting into.

No, it's not a shady practice and I'd appreciate you not coming into our forum and accusing Corsair of such activity. Our forum is not here for uninformed or abusive statements to be made and especially when the statement is wrong or made out of ignorance or omissions of fact.

And please stop making uninformed statements about the usage of the term proprietary. For Corsair and any other vendor selling SSDs, the proprietary information about any product they buy and resell is covered by NDAs. That stands for NON DISCLOSURE AGREEMENT if you are not familiar with the term.

Even if the information you want existed, I could not give it to you. I could get myself and Corsair into legal trouble. Not going to happen.

I am still at loss here. I don't understand what part of this you don't get. The data you are grilling me over is unreliable. I have proven that to be a FACT. Go to any other SSD vendors site and you'll find the same information or, lack thereof.

SMART has long been in use by the HDD industry and is much more standardized in its application to HDDs. However, much of the information reported by an SSD does not use or comply with common SSD SMART reporting. SMART IS NOT RELIABLE ON AN SSD, PERIOD. YOU CANNOT RELY ON THE INFORMATION YOU ARE SEEING. I am out of analogies or examples to attempt to explain this to you.

If you are going to go about trumpeting about this on the internet, be sure to link the sticky on SMART reporting.

I will obviously share this over the Internet. We live in a share culture now, and we don't appreciate all this info being kept secret. This a deprecated way of life and future will make sure it will die, as hiding such information for commercial purposes is harmful to customers. In the end, information will still reach the public.

I will continue to document the SMART data until the drive dies. I have a strong feeling it is totally erroneous, and it is a shame nobody cares to put some order into these matters. If the SMART data is unreliable, how about making a reliable way of checking your drives available? Is this too much to ask?

I will also be sending a letter to Indilinx, as I suspect it's their controller that's responsible, and will post the answer here if they care to respond.

I will also respect your desire, and link to the SMART sticky on this site.

As per Corsair's desire, I will be making available a link to the SMART sticky on their site. To make sure we are fair, I will post it on a few places on the thread.

[toyo]

If you plan on reading the entire thread from the Corsair forums, it is here: Estimated lifetime for an Extreme x64 SSD?

I almost feel this Yellowbeard guy trying to drag me into an offtopic discution about world issues like economics and stuff, and as much as I would like to walk this way, I'll be staying focused on tech problems.

I also wrote Indilinx in an attempt to find out info about the Health parameter. I suppose they won't probably bother to answer, although I have sent a very polite and documented email.

Fighting with the windmills, myself...

[Maxhedroom]

Sounds like you got a real tool for a rep. Doesn't matter what smart reader is used, the controller still reports the same inherent data. One last avenue would be to post on Ocz posing as a 60gb Vertex owner....just give the numbers...no scrennies. Ask about possible causes of sudden and rapid health deterioration. I promise you'll get a whooole lot more attention and probably a good explanation...good or bad. ;)

Btw, your experience is the number one reason that Ocz is the top SSD vendor....and I hate to say but Corsair's ssd support is about as good as it gets...a lot of companies don't even offer tech support at all. :blink:

[toyo]

If I lie, I will lose all my credibility when it will be found (someone will notice and correlate the three threads in the end). As much as I desire to find out what the heck is happening, I would like to avoid lying at all cost.

If you personally know somebody there and can send him a PM with the thread link here/data/pictures or something, please do it. You could even be honest and post that you have a friend with a different drive that's deteriorating fast and is in dire need of SSD expertise, and a certain OCZ competitor would not help him :)

[Maxhedroom]

I'll see what I can do...Ive been a regular there for a couple years now and I'm in good with a lot of the resident SSD Gurus that also run other brand names as well as Ocz drives. Gimme a few days, I'll see what I can come up with. ;)

[toyo]

I'll see what I can do...Ive been a regular there for a couple years now and I'm in good with a lot of the resident SSD Gurus that also run other brand names as well as Ocz drives. Gimme a few days, I'll see what I can come up with. ;)

Thank you very much for being my partner in crime in this topic and offering to helping me! Take your time and do as you think it might be best...

[Maxhedroom]

No prob...don't mind at all...and I'm also just curious as to why it's behaving so wierd. If my Vertex 2s started doin that, I'd be flippin' and $hittin' all over the place... :fear:

[toyo]

Hehe, I'm pretty much at peace with the situation regarding my SSD, since this was a gift. I can't do anything about it I think, so I'll be trying to take care of it for as long as I can as I'm aware that it will not be easy to replace.

The reason I'm going through this fight is because I want the information to be made public and be available to the SSD community, which I know is starved of any real endurance data. I mean, common, people are killing their SSDs just to see how long they would last? Wouldn't it be easier if the companies would provide a better way to analyse their gear, and release their information on NAND life etc.? At least the older drives should have been tested enough by now to have lots of available data that cannot commercially hurt companies any more, especially since each one of them used the same controllers&NAND.

So this is it. I want this SMART data properly documented, or another reliable way to analyse the status of your SSD. I want the clients to have the power of knowing their SSDs, not being confuse and in the dark all the time with no reliable information about their expensive gear available.

I know that I will most probably not achieve this, but we live in the days of the Internet, and maybe more people will voice their opinion and demand this data to be made public in time. Who knows?

[toyo]

Check this out:

- the drive subtracts 1% Health every 100 Write Cycles.

- a simple calculation: 100 cycles x 100 Health = 10.000

10.000 (P/E cycles) is exactly the number which I heard that the Samsung 51 nm NAND is supposed to have. P/E cycles = Program/Erase, so it should be a write, AND an erase.

It might be very possible these guys just took the rough number of average P/E Cycles Samsung made available for their NAND, and considered it as the life of the product without any additional testing :frusty:

Somebody needs to give them a good beating if this is the truth.

[robalm]

My ssd still rocks after 15.35TB writes

[Maxhedroom]

I bought all 4 of my drives at about the same time and since they all share the traffic in raid0, I'm only at about 1.5 - 2TB on each drive and all still in perfect health.../me knocks on wood. Below is my smart reading using the Ocz Toolbox and also a screenshot of a CDM run with 3 of my drives but that was on my old Core2 Duo/P35 ICH9R rig. Within the next couple weeks I'll manage to round up all four for a good shake down on my new rig....muahaha...

[toyo]

Yeah, lucky you.

I really need to find out somehow if all the Barefoot drives count the SSD life like this, 100 Write Cycles=1% Health. I can't find info on it anywhere. It would certainly help in anybody has an OCZ Vertex or other Barefoot and can monitor it for a few days until it completes 100 Write Cycles. Really easy to find out.

If other drives count the health the same way, then I need to know why the write cycles are going up so fast. Remember, I took off the SSD every write-heavy app/feature that I could, but still there are a few (4-8) GB/day written, from Windows itself and from the Opera features that cannot be moved, plus Kaspersky IS.

So:

- is it normal for any Indilinx Barefoot drive to calculate its lifetime like 1%=100 Write Cycles (it makes sense, given the estimated life of Samsung NAND, 10.000 P/E cycles)?

- is it normal for the Write Cycles (Cell Wearing Cycles in SSD Status) to grow so fast? Each 20-40 GB written, 100 Write Cycles are gone. The Health is going down very, very fast, especially when compared to other drives.

So, if any of you have an Indilinx Barefoot drive, please help me with this, monitor it for a few days and post.

[Maxhedroom]

I think the easiest way I can help is to invite one of my fellow ssd fanatics to reg and join this thread. ;) I even know one fella who is still running a 6 x Vertex 30GB array and has never even used trim...lol

[toyo]

Please do... Corsair are not sharing ANY info at all and Indilinx did not even bother to respond to my email. It's not like I ask for fabrication secrets or something, I only want to understand the damn Health parameter, and related Cell Wearing/Written data relation.

[Maxhedroom]

Mk...I shot him a pm on the Ocz forum. I imagine he'll pop in before long. ;)

[groberts101]

Hi toyo, Just dropping you a quick line to help ease some of your concerns after receiving a PM from Zaxx over at the OCZ forums. He has shared his time with many others and it's earned enough of my respect to take some of my own and share with who he thinks is worthy of it. However, it will be brief(well.. anything under 5000 words is "brief" for me) and quite blunt. Many appreciate my no holds barred attitude and I can assure you no offense is intended. I run 8 of the Vertex 30GB drives spread about my home in various systems(no more 6 drive array), have 3 on loan to friends and family, and am quite experienced with the inner workings/limitations of the barefoot controllers having owned and tested more than 30 so far. I currently run a 6 drive Vertex 2 raid0 array on my primary workstation(gfx/vid machine) and also beta-test the newest Vertex 3 for OCZ. I don't know everything(won't let my wife hear me admit to that though).. but I do know a lot about a little. Arrogance aside.. here goes.

First off here is the speed concerns you initially posted of. Again.. it will only be a brief overview with side comments not necessarily intended to be open for debate or further teachings on the matter. Kind of a "take it or leave it" type thing as I'm known for when my time is short.

Empty quick formatted drives being tested as spare will always be fastest. Having an OS installed and reducing the fresh block availabity will always slow ANY SSD down(though Sandforce is less impacted by this). To quantify configs/settings/drivers effects on performance it's always best to secure erase between the tests to refresh the drive. This will not wear the drive out prematurley(I've done it literally hundreds of times on all of mine) and should/must be done to quickly quantify results and can actually save time and repeated testing/tail chasing in the long run. This gives an exact/proper baseline to work from and eliminates many variables that can skew results and leave you frustrated.

Now, that being said.. it appears to me that with the reboot being the main factor in the drive suddenly slowing down(without very much prior testing causing too full/dirty of a drive?).. I would guess that it's a system/bios throttling issue occurring there. Go into the bios and disable all c-states and speedstep. Don't need to live with them, but they surely add to the number of throttling variables involved with any system that uses them. These settings have to do with the CPU/bus throttling of any system and will have varying degree's of impact based on the particular way the mobo mfgr has implemented them. Standards are there but comparing a laptop to a PC will be the easiest way to show you how differently they can be implemented based on power saving priorities. "Green" is not always the best way to get maximum performance and compromises will and must be made. Also set the W7 power options to high-performance and to also never let the hard-drive shut down. Should also never let the system sleep when you are doing dedicated logoff idles to allow GC to do its thing(overnights on ocassion are always best). If you do feel the need to sleep the system during logoff idle recovery?.. then you must switch the bios to S1 sleep mode rather than the typical defaulted S3 mode since S3/S4 will cut power to the drive and negate any GC from occurring. GC must have constant drive power and low activity(hence the "no sleep/S1 and logoff idle" rec's).

Now for your bigger concern here. And here comes the blunt part.

You will NOT get this proprietary info from any mfgr(MANY before you have tried) and to be perfectly honest?.. most don't give a flying hoot about these numbers anymore anyways(besides the ones who ask the same "why my drive dying so fast" questions. The drive in question here goes for under $100 now and can write WELL into the Terabytes of data before you exhuast the PE/c of that 51nm nand. Yellowbeard is not an engineer and was very forthright in his explaination(the little I read) to let you know what the other mfgrs are saying as well. DISREGARD the smart data as it's simply a calculation based on the current usage levels. Test the P' outta' the drive and use it to watch streaming media?.. watch it drop like a rock and don't freak out. Go out of town for a weeks vacation while the drive idles at the login screen and then use it lightly after your return?.. the life-remaining stats slow down and average out better to leave you feeling all warm and fuzzy.

As Zaxx already mentioned there has been more than a few threads around the internet where people were in a similar uproar over the lifespan calculations of these drives and guess what happened in the end?.. the drives reached 0% and then the counter completely reset and they went about their business. They may have not been completely satisfied that their drive would last forever.. but they realized that what they were being told by the moderators/staff/and other members was in fact reality. Less than a handful of the drives actually died but there is no actual proof that they were completely burnt since these things have been known to panic lock on ocassion anyways.. however odd that conicidence may be(which most don't realize is not actually the same as "dead") and can simply be destructive flashed back to life(at least OCZ's Indy drives can) with totally reset smart data. Also consider that there have been variuos firmware adjustments made through the years to actually reset these smart counters to reduce the amount of "scares" related to drives moving from 50nm nand to 34nm and now 25nm and smaller. OCZ had a hell of a mess when they reduced the PE/c rating of the nand and left the counters alone with a barage of posts related to lifetime calculations.

So my advice to you is to use the P' outta that thing and quite worrying about a drive that is already obsolete in the scheme of things SSD related. OS/app's, pagefile(though limiting it to a static size of 1024/1024 is always good if you have sufficient ram of 6 gigs or more) and whatever you can fit on there without going beyond about 70% or so of available space would be perfectly fine. Otherwise?.. get rid of it and move to another SSD if you can't use it for the purpose that it was intended for and/or it's stressing you out this much. Going on a "smart data consumer advocacy campaign" will not be worth the time and effort and totally misses the point of using an SSD in the first place,.. which is TIME SAVED.

I say buy em'.. burn em'.. and upgrade em'. Simply call it planned obsolescence as it's always been intended to be disposable tech anyways. Plus, it's very likely that by the time that thing dies?.. it will be half the size and half the speed of similarly priced SSD's that you newer PC will be able to make use of. Good luck with it and hope I helped even just a little. :rolleyes:

[toyo]

Welcome to nsaneforums, groberts101 :) Thank you for taking the time to register here, especially since you did this completely related to this issue.

You are correct - from your point of view. My point of view is that even if this was a 10 USD drive, it must last as long as possible. What is disposable tech to you, might be the only drive for another. Also, SSD purpose is different for each client, for you it's time saving, for me it might be the only place to host my OS until a better one.

The manufacturers are still guilty of hiding essential information from the public and not properly documenting SMART data. I don't intend to campaign for this data to be released, my purpose for now is to find out simple things, like if all Barefoot drives are subtracting 1% health/100 Wear Cycles, and how many cycles/written GB are usually consumed.

I am fully aware about your other suggestions, thank you. From what I've read, this drive does NOT have GC implemented - only TRIM. Yes, this is the latest Corsair FW, v2.0, this is based on Indilinx FW 1819. I understand a newer Barefoot FW is available (2142), but Corsair seems not willing to make it available.

EDIT: Corsair says that FW 2.0 is having both TRIM and GC.

[toyo]

92% images below. Since this is happening so fast, and the health drops each 100 wear cycles, I will probably post updates only 5% or so. Only 25 GB written for this particular 1%...

Just one year until I consume all the P/E cycles, and I didn't even install the "work" apps, or any non-essential app to be honest :sad: Will the SSD die after this? I doubt it, but remains to be seen though.