Yes, the C300 uses our 34nm NAND, but it’s important to look at more than process node—not all 3xnm NAND is the same. For example, we’ve been manufacturing 34nm NAND since 2008; it took months of refinement before we reached the reliability and performance we felt was necessary to build an SSD worthy of the Micron brand. It’s true that the design effort does get more difficult as the process shrinks, but design maturity and refinement can go a long way toward improving durability and performance on any process node. We’re confident that our 34nm NAND is among the best on the market.

For a little more explanation on how NAND matures during development, see Kevin’s video on Enterprise NAND. He’s talking about how we’ve been able to increase endurance over time for an enterprise part, but the same principle applies to general NAND development.

Our NAND adheres to the JEDEC specification for data retention, which is 10 years of retention at up to 10% of the maximum specified number or program/erase cycles, and one year after max cycling, at 55°C. (Retention would be significantly better at “room” temp.) This is a baseline number; SSDs built on these retention specs will do equal to or better than this, depending on how sophisticated the controller is and on actual drive temperature.

Unfortunately, I don’t have an SSD retention spec to give you. As I noted, there are many, many variables that affect the eventual retention spec. While it’s true that we could define a set of specific use case conditions and come up with a number that would likely be true for some users, doing so would make it very easy for any competitor to tweak the test to their favor and market against us—claiming that they have better retention. And because most people just look at the spec, not the test methodology, most people would believe it. This lack of a testing standard for various SSD specs is an issue we’ll see crop up more as SSDs become more mainstream. Once the industry agrees on standardized measurements for performance, reliability, and endurance, it’ll be easier for consumers (and manufacturers) to make a true head-to-head comparison. But again, we expect the data retention to be more than sufficient in actual SSD use.

As for real-world use, I just mentioned putting it in a drawer to create a vivid picture that what I was asking for was unpowered data retention. This does have a real-world use. Someone may have a computer in storage for a long time and then take it out and expect it to work. My question is, how long before the data decays and it cannot be used?

Surely someone at micron can give a reasonable estimate of the unpowered data retention time for the C300?

We do spec NAND retention to our customers, but SSD data retention is dependent upon total bytes written and multiple other factors that make it difficult to give a simple, definitive spec. That said, we’re very confident in the quality of our NAND and the durability that it lends to our SSDs. No one recommends using an SSD as archive media (throwing it in a drawer for years), but when it’s used as a normal drive in real-world conditions, data retention is really a non-issue for the C300.

Thank You

You’ll notice that Crucial has responded on their forum, but I think the answer bears repeating here:

“Yes, we are aware of the issue Anand has encountered, and have been working to understand the conditions and parameters he was using to recreate the scenario in our own labs. We’ve done thousands of hours of testing on the C300 and have sent the drive to dozens of independent testers, receiving phenomenal response on how it has performed. We are confident in the high quality and unmatched performance of the C300.”

In short — our engineers are on top of this.

I do a lot of video transcoding, which is actually more CPU intensive, but not having a noisy 15k SATA drive thrashing away under my desk is wonderful.

I also recently ordered an SSD for one of my production database servers, as some of my queries were taking too long on a Fujitsu MAS Ultra-SCSI.

Now, I could have sat down and laboriously reworked my already-optimized SQL for 20-30 hours, but after testing an SSD as the data drive, the problem queries sped up by a factor of 10. Problem solved.

http://forum.crucial.com/t5/Solid-State-Drives-SSD/Serious-problem-with-RealSSD-C300/td-p/10599