Sick of SSD life questions? Lets squash this rumor.

[LeftSide]

I'm sick of seeing "I'm afraid if I put my 'browser cache / virtual memory' on my SSD, it will eat up all my erase cycles!"

Lets put an end to this rumor. I have an old Kingston 40gb SSD, it was used in a raid 0 (with no trim), for 1 year and then decommissioned to my wife's computer as a stand alone C drive. It has 2.53TB of writes and the media wear-out indicator is at 0. I have never done any tweaks to windows, or moved any browsing cache, virtual memory, or anything else off the drive. I have used it non-stop for 2 years with no indication of wearing out. I have never once worried about it.

Has anyone ever used an SSD heavy enough to actually wear out the cells in a non-server environment?

[Accord99]

If you haven't seen it already, there's a Xtremesystems forum thread where people go through and try to see how much data can be written to SSDs before they die. There's a 256 GB Samsung 830 that's nearing 4 petabytes of writes and is still alive.

http://www.xtremesystems.org/forums/...=1#post5132834

Even drives that have died tended to last longer than the # of cycles their flash memory were rated for.

[sequoia464]

That Kingston/Intel 40 GB drive has done extremely well in the testing over there also.

I have a couple of those drives still, trying to figure out a use for them.

[Blain]

Originally Posted by LeftSide

I'm sick of seeing "I'm afraid if I put my 'browser cache / virtual memory' on my SSD, it will eat up all my erase cycles!"

I never get tired of redundant threads here at the old AT forums.
What kind of traffic would there be without them?

[LeftSide]

Originally Posted by Accord99

If you haven't seen it already, there's a Xtremesystems forum thread where people go through and try to see how much data can be written to SSDs before they die. There's a 256 GB Samsung 830 that's nearing 4 petabytes of writes and is still alive.

http://www.xtremesystems.org/forums/...=1#post5132834

Even drives that have died tended to last longer than the # of cycles their flash memory were rated for.

Wow, I never realized the drives would last that long! That is crazy.

[Cerb]

Originally Posted by LeftSide

Has anyone ever used an SSD heavy enough to actually wear out the cells in a non-server environment?

No, only thumb drives, and only from several years ago, when WA <20 was a rarity for anything but large files.

WA varies by drive, but 1.1-1.5 seems to be a typical range for non-Sandforce, these days. Since it takes awhile to find out for any given drive based on actual use, I like to pessimistically go with a <=2 assumption. As long as rated cycles stays high enough, and you aren't trying to re-purpose it in a transaction-heavy server, it's worrying about nothing. There are far more likely failure modes than too many writes.

[hhhd1]

Originally Posted by LeftSide

Has anyone ever used an SSD heavy enough to actually wear out the cells in a non-server environment?

Never.

There is actually allot of people right now with SSDs, I wonder how they are doing.

[hhhd1]

I strongly agree with this topic, I was about to create a thread asking for some statistics from real people around here about their wear leveling count.

[Smoblikat]

I have an OCZ vertex 1, if anyone has any info on reliability thatd be nifty.

[Vic Vega]

This thread is in contrast to Anand's own articles where he has claimed to brick SSDs inside of a week.

[vss1980]

Originally Posted by Vic Vega

This thread is in contrast to Anand's own articles where he has claimed to brick SSDs inside of a week.

Well, keep in mind you only need to get one NAND element to fail before you could technically write off the drive as being faulty and having failing reliability..... I suppose you could theoretically exploit a weak spot of a device if you know how, or through blind luck, extreme amounts of benchmarking, etc.

As an example, some of the wear leveling on early high capacity micro SDHC cards was abysmal (read as effectively non-existant) hence I have a 16GB card of which the first 2GB are essentially useless and unreliable - the rest works like a champ for now. My mistake, using a journaling file system on that section.... <facepalm>...

[Mark R]

Originally Posted by LeftSide

Has anyone ever used an SSD heavy enough to actually wear out the cells in a non-server environment?

I've seen a few threads in other forums where people with small-capacity early Indilinx drives have worn them out.

The 1st gen indilinx controller had disgraceful write amplification - typically in the region of 30-70x (compared to 1.2-1.5 for a contemporary Intel controller, and even less for Sandforce).

If you had a small drive, like a 32 GB indilinx drive, you're basically only looking at 2 TB of writes before "remaining life" reaches 0. That can easily be achieved in a year of light enthusiast use (and judging by the few threads I've read, this has happened on a number of occasions). In a couple of cases, the drives reportedly died a few months later.

[jwilliams4200]

Originally Posted by Vic Vega

This thread is in contrast to Anand's own articles where he has claimed to brick SSDs inside of a week.

I don't remember Anand mentioning wearing out the flash in an SSD. Can you link to the article?

[Emulex]

there is a website that is testing consumer drives in server workloads. it is horrific the results. google it

[VirtualLarry]

Originally Posted by Mark R

If you had a small drive, like a 32 GB indilinx drive, you're basically only looking at 2 TB of writes before "remaining life" reaches 0. That can easily be achieved in a year of light enthusiast use (and judging by the few threads I've read, this has happened on a number of occasions). In a couple of cases, the drives reportedly died a few months later.

I used to use a 30GB OCZ Agility drive on my main rig, which is running DC 24/7. After less than six months, IIRC, it had 1.9TB worth of writes, and was at ~74-75 SSD Health, according to SSDLife. I think it estimated a 2-year lifespan for that drive, about.

[MaxPayne63]

Originally Posted by LeftSide

Has anyone ever used an SSD heavy enough to actually wear out the cells in a non-server environment?

I have an 80gb x25-M as my system drive that some random ssd utility I have says is good until 2052. I've been thinking about getting a larger drive for Steam etc but I don't want to get locked in to a 40 year upgrade cycle.

[LeftSide]

Originally Posted by hhhd1

I strongly agree with this topic, I was about to create a thread asking for some statistics from real people around here about their wear leveling count.

We could do something like this.

SSD :
Controller:
Size:
Power On Hours:
Lifetime Writes:
Life Curve Status:
SSD Life Left:
Trim enabled? (y/n):
OS:
Is this the OS drive? (y/n):

SSD : Kingston hyperX
Controller: SF-2281
Size: 120GB
Power On Hours: 3974
Lifetime Writes: 2.26TB
Life Curve Status: 100
SSD Life Left: 100
Trim enabled? (y/n): Yes
OS: Windows 7 x64
Is this the OS drive? (y/n): Yes

[LeftSide]

Originally Posted by VirtualLarry

I used to use a 30GB OCZ Agility drive on my main rig, which is running DC 24/7. After less than six months, IIRC, it had 1.9TB worth of writes, and was at ~74-75 SSD Health, according to SSDLife. I think it estimated a 2-year lifespan for that drive, about.

What is DC?

[fuzzymath10]

Originally Posted by VirtualLarry

I used to use a 30GB OCZ Agility drive on my main rig, which is running DC 24/7. After less than six months, IIRC, it had 1.9TB worth of writes, and was at ~74-75 SSD Health, according to SSDLife. I think it estimated a 2-year lifespan for that drive, about.

I think a somewhat recent firmware (1.7?) helped reduce the WA on my Onyx quite significantly. My health went down to ~20 before the update over about 2 years, and now it has stabilized after the update.

[Blain]

Originally Posted by hhhd1

I strongly agree with this topic, I was about to create a thread asking for some statistics from real people around here about their wear leveling count.

Thread creation is the life blood of these forums... Go for it!

[Soulkeeper]

SSD : Crucial M4
Controller: Marvell 88SS9174
Size: 128GB
Power On Hours: 7511
Lifetime Writes: 2670 GB
Life Curve Status:
SSD Life Left: 99%
Trim enabled? (y/n): Yes
OS: Slackware
Is this the OS drive? (y/n): Yes


SSD : Plextor M3
Controller: Marvell 88SS9174
Size: 128GB
Power On Hours: 3662
Lifetime Writes: 1366 GB
Life Curve Status:
SSD Life Left:
Trim enabled? (y/n): Yes
OS: Slackware
Is this the OS drive? (y/n): Yes



it's worth noting that the smart data varies wildly between drives
the only one that is likely to be on the vast majority is "Power_On_Hours"
Also Lifetime Writes is filesystem specific and erased when you create a new filesystem (which i've done on both my drives).

[ShintaiDK]

SSDs will always outlast HDs. Unless you bought sandforce

My oldest SSD in my HTPC now. X25-M 80GB G1. 10.45TB writes. 99% reserved space still free. 24764hours ontime.

Its only people trying to hold on to a hope in HDs that will claim otherwise.

[Blain]

Originally Posted by ShintaiDK

SSDs will always outlast HDs. Unless you bought sandforce

You've painted with a pretty broad brush there.
Can I extend my non-sandforce SSD warranty through you?

[BrightCandle]

In many ways I am disappointed with the reliability of SSDs. Solid state in my mind should be approaching the reliability of CPUs or RAM not competing with Hard disk drives. The lack of moving parts should mean they are incredibly reliable devices once they have been quality tested to work at all.

So why do we see failure rates in the single digit percentage points within a year? These aren't failures where the drive has run out of writes, these are technical failures of parts. Until they get to the bottom of why it is so many parts fail long before they should its irrelevant how long the drive will survive being written to.

Lets say an SSD could only survive 1TB of writes and then it would stop working. That is a very short period of time for many people but at least if that was how they behaved you should get a read only drive at the end of it that gave you a warning before it reached the problem point so you could get a replacement. That is a massively better scenario than any HDD because the data is still safe. But it doesn't appear that is what is happening for most people that have SSDs stop working.

The life thing doesn't matter, its the really high failure rates that are the problem, these things should be more reliable than HDDs and yet they aren't.

[Blain]

Originally Posted by BrightCandle

Solid state in my mind should be approaching the reliability of CPUs or RAM not competing with Hard disk drives.

SSD's have a more direct line to potential voltage fluctuation (from power supplies), than do CPU's or memory.
Even memory has a less than stellar reputation for quality and reliability.