How stable are the mainstream intel SSD's?
The have excellent reviews except a few saying how they break a few weeks or months in. Would it be to much to expect a year of high performance from one? What are the chances of it breaking... 1 in 10 after a year?
80 Gig Intel G2, Oct 2009, Not one problem (desk top - I5-750, win 7 64 Bit)
Patriot Torqx 128 Gig, Aug 2009, Not one problem (Toshiba Laptop A305, Win 7 32 Bit)
WD 128 Gig, received 1 week ago, not installed yet ( not sure where to install - choise, Wife's I3-540, My E6400, or Toshibia A025 laptop). Probably would not recommend, No current support from WD, ie firmware, software equiv to intel Toolbox.
Got an 80GB X25-M G2 ~October/November 2009. Running great as an OS drive with Server 2008 R2 (Yes, it's a CAD/Rendering workstation, NOT a server; running a Server OS cause I got a free key via MSDNAA).
This isnt a very good way of getting anything close to real data to make a decision.
Go to the Intel site and read their warranty. Last time I checked it was something like they guarantee writing 100G/day for 5 years - or something like that. I havent kept a mechanical drive for 5years - they get replaced before that, and I expect the same here.
When it does fail it will typically be increasing soft errors and you will notice that.
So there isnt going to be a lot of failures on the drive side.
But apps and the OS still corrupt the file system. People blame the drives many times when it is not the drive at issue.
Warranty is not a great indicator of how long an "electronic" device will last.
Currently the best indicator is from owners, unfortunatly very small population, and comparing the reviews such as on Newegg. several points - Need a fair Number, Must take with a grain of salt, Also remember that about the same percentage of people (Better adjective available) who break the item/don't know how to properly install is about the same across the board.
Warranties for SSDs vary from 3 yrs to "Lifetime". I bought a 128 gig patriot Torqx with a Lifetime warranty. Will it out last the 3 -> five year warranty of OCZ, or Intel - I dough it.
How long it will really last "100G/day for 5 years" is an educated WAG. Intel has not conducted a 5 year test with a represenative sample units for their G2 model - They were only introduced mid 2009. They (as others do) run a stress test and then apply the results to a model for a predictive failure rate. I'm NOT saying that it will not last that long, infact it may last longer.
Warranty is based more on costs, ie Nr of units that must be kept "on self" and nr of years repair capabilities must be inplace. A good example is you can buy an identical item with different warranty length - Longer warranty, more you pay - But the item is identical.
How long it will really last "100G/day for 5 years" is an educated WAG.
Actually, I believe the "100GByte/day for 5 years" applies to the SLC G1 units. The figure I've seen for the MLC G2 units is "20GByte/day for 5 years", still a very respectable number. My Windows 7 system is running at less than 5GB/day, so that would give it a 20 year life expectancy.
These figures are not guesses, but are extrapolated from the known wear rates of the flash memory cells, the algorithms that the drive uses to spread the writes over different cells, and an understanding of what the I/O workload looks like in a typical system. Intel's process in arriving at these figures is fairly well documented and I have a lot of faith in it. See this Anandtech article for a detailed description.
Siminlal - thanks
My "educated WAG" was loosely applied. Very few electronic systems are really tested with x # of units for the time period. As you say they are more often based on extrapolated data and/or from empirical data based on a model, or an algorithm.
As to warranties I think I'm pretty close to correct.
Very few electronic systems are really tested with x # of units for the time period.
I understand exactly where you're coming from - things like the life expectancy of CD/DVD/Bluray media, for example, are testing using "accelerated aging", which is engineering jargon for "wild-ass guess" . But the number of writes that flash memory cells can accept is a well-established value, easily tested without having to wait for 5 years. It's kind of like giving an estimate for how long a toner cartridge is going to last - it's easy to calculate if you know the black coverage of the average page.