One thing that erks me is half these articles including many professional tech news websites... is they get the failure rate completely wrong.

The SSD failure rate is one of the so called 'fad talks' about why not to get an SSD drive for an OS.

As if these tech news ppl were brilliant at math, they take the MTBF or mean time before failure, quoted to be 2 million for some core SSD or 1.5million... they divide by 24 hours and 365 days and go "OH it will last 150-200 years" then proceed to crack about the 2 year warranty compared to the uncovered other 198 years...

The MTBF is in a sense a failure rate, or ratio for failure. It has almost nothing to do with direct time before failure.

So, although I'm ranting.. it is for the greater good of people understanding what the MTBF means...

the MTBF is actually the reciprocal of the time before a part in the device fails. so it'd be 1/(% of part a part failing).

For a 2 million offering, function of % non failure is

% chance that someone will work in T hours is

P(T) = e^(-T/MTBF)

Percentage the SSD will last for 5 years with a MTBF of 2mil is = e^(-5*24*365/2,000,000) = 97.83%

"Is this a SLC or MLC device?"

OCZ aren't saying. Has to be MLC at that price, though.

"If you read the fine print on these, they are nowhere near as fast as the other OCZ SATAII SSD's that are for sale: they specifcally mention transfer rates in Mb/s instead of MB/s."

You're putting way too much faith in the ability of marketing to know the difference between "b" and "B".

Anyway, from the press release:

"Core series SSD drives are available in capacities of 32GB, 64GB, and 128GB and deliver incredible 120-143 MB/s 80-93 MB/s read/write speeds and seek times of less than 0.35ms, making the Core series up to 10x as fast on a seek-time basis and up to 40% faster on a R/W basis that the best performing 2.5” HDDs on the market, all while consuming 50% less power."

Clearly they mean megabytes, not megabits, otherwise it would be slower than a HDD, not faster.

A little math shows that the worries about wearing out flash is pretty silly.

Take the smallest 32GB device and assume it uses the cheap parts that last for "only" 100,000 writes. The wear levelling insures that even if you wrote the same block over and over again it will move that block around. It will move other data out of the way into blocks that are more "worn" as necessary to even out the wear. You tell the drive to write block 234 but internally it remaps the block numbers you tell it to its own representation, so it can move data around with you knowing or caring.

With a little math, you can show that if you wrote 10MB/sec every second for a year on that 32GB device, you'd have an average of 10,000 writes per block. So if the wear levelling works as advertised (and its a simple principle, I don't see why not) then it would last 10 years at that rate.

And who the hell is going to write 10MB/sec every second for 10 years on a laptop? If you raise your hand and say "me", then maybe you should stick with hard drives for now. The other 99.9999% of us can make good use of an SSD.

Recently in: theInquirersters' article about Intel bundling SSD Card with Atom appeared & Cost Was Low. i ask, Is Above products Price as Good as Manufacturer State? its mere ~100 mb/s unit, yet it is nicely done in metal. However, Intels' LOSS Leader to Sell Atom Means You'd Get Atom Almost FREE, with SSD Card.
drashek

jacob : actually a algorithm can save you then.
what it can do it move data that is hardly ever changed to sectors that have seen heavy write traffic, and use the newly freed up sector for heavy writing for a while.

If you read the fine print on these, they are nowhere near as fast as the other OCZ SATAII SSD's that are for sale: they specifcally mention transfer rates in Mb/s instead of MB/s.

> Now, if the disk is fairly full, no re-mapping
> algorithm can save you from realizing bad areas

That isn't true. The trick is to move existing data around, so basically writing a little bit more to avoid exhausting one area. Writing only to free space would be a recipe for disaster, writes need to be spread around the whole disks.

That said, I don't know what wear leveling algorithms are actually used, and it's not written on the box.

Come Inquirer. Is this suppose to be a IT site or not?

Is this a SLC or MLC device?

Makes a huge difference.



Toms review is complete bullshit (like much of what they do), their testing methodology is beyond flawed.

Simply put they carried on running HDD test until the battery ran out, so the nice fast SSD's completed more tests than the HDD, these tests would also have been using CPU power as well, is it any surprise that they came to the wrong conclusion by using wrong testing.

The OCZ SSD's are just re-badged Samsung SSD's, so the price point will be very similar.

Andy

what the mean time to "can't write any more" would be: high quality flash chips can live through about a million rewrites to the same cells. Cheaper flash chips can sustain about 100,000 rewrites. Now, if the disk is fairly full, no re-mapping algorithm can save you from realizing bad areas after several says of intensive work - e.g OS cache.