It’s known amongst tech enthusiasts that being the first to upgrade will always incorporate some additional level of risk. The problem with brand-new technology (especially that which expands the current envelope) is unforeseen complications that can develop. Or, new limitations that previously were never a concern before, or even on the radar, but can suddenly become a major problem.
Solid-state drives are not a new technology, and NAND flash memory has been around for well over twenty years. What is new, however, is the recent explosive growth in SSD performance. While it would have taken many years before SATA 2.0 (3Gbit/s) began to be a problem for spinning disk drives, SSDs have quickly maxed out the capable performance by delivering 260-270MB/s read speeds for a few years now.
That is partially why extreme SSD “drives” have elected to take the form of an add-on card in order to use the PCIe bus, bypassing the bottleneck. To solve this, the SATA-IO organization released the SATA 3.0 (properly known as “SATA 6Gbit/s”) specification that should give flash-based drives the potential to stretch their NAND chips just a bit. The Crucial RealSSD C300 is the first SSD to include full support for this faster SATA interface.
Anand Lal Shimpi of AnandTech is no stranger to the perils of working with bleeding-edge technology, having experienced the early drive performance degradation problems to having an early prototype Vertex 2 drive fail during testing. Recently, he has had the “privilege” to add yet another SSD to his wall when his Crucial RealSSD C300 appeared to fail during testing of AMD’s 890GX, that company’s first SATA 6Gbit/s capable chipset.
After Crucial had time to take a look at the “dead” SSD they determined that the firmware tables had in some unknown fashion developed errors, causing the drive to react extremely slowly during the boot process… slowly enough that the drive would not be detected during POST. While this means any user data stored on the drive could still be recovered, it would necessitate returning a brand new $800 drive with your precious data to Crucial for recovery.
If that was all then perhaps it wouldn’t be so bad as most new technology does have early teething problems. Once new technology reaches the mainstream, most, if not all of the major kinks have been already worked out of it. Unfortunately, Anand also discovered some unusual performance aspects with TRIM. For some curious reason the C300 seems to be one of a few recent SSDs that don’t recover properly after used LBAs are TRIMed. While read performance remained steady, write performance would drop from 180MB/s to a flatlined 20MB/s. The last time we have observed this behavior was with early JMicron controller-based drives, which at the time could not handle data fragmentation all that well.
Fortunately, Crucial has stated that it was able to replicate the problem and devise a solution that it’s implementing into the RealSSD C300 drives. If this sounds at all familiar, then that might be because Intel ran into the exact same problem with its first-generation X25-M series SSDs and had to release new firmware to resolve the problem! It all just goes to show, it isn’t called living on the “bleeding-edge” of technology for nothing.
Write performance does not look good. While parts of the drive can still write at around 180MB/s, the last 60GB of the drive are limited to about 20MB/s. A quick format across the drive should invoke the TRIM instruction for all LBAs and tell the C300 that it none of the data on the drive is needed and those blocks can be recycled immediately. Performance should restore to new (constant ~200MB/s across all LBAs). Unfortunately, it doesn’t.