A lot of people in the industry believe that flash memory is the future of data storage for personal computers. We all use flash, of course, in digital cameras, in most music players, and on the ubiquitous USB flash memory keys. But flash is just beginning to appear in desktops and notebooks, either as an additional interface between the computer’s main memory (DRAM) and the hard drive; on the hard drive itself as a larger cache; or as a replacement for the hard drive, called a solid state drive (SSD). The concept is simple: hard drives require a motor to physically move a read/write head, while flash memory is a solid state device that offers random access. As a result, flash memory drives should be faster, lower-power, and more reliable.
There’s just one problem: flash memory drives don’t seem to work as advertised yet.
Take the concept of using flash memory to supplement the memory on a PC, which Microsoft calls ReadyBoost in Vista. In my tests, it can make a difference on a machine with 512MB or less, but doesn’t seem to have any positive impact on any machine with 1GB or more of RAM. The overhead associated with managing the memory seems bigger than any real benefits, and most real world tests I’ve run on it show slower performance. It’s not “ReadyBoost” it’s “ReadySlow.” On a 512MB system, you can see some improvements, but why bother? You’ll get much more improvement by adding another 512MB of DRAM enough main memory.
Another idea is to use as a cache between system memory and a hard drive, which Intel calls Turbo Memory, or “Robson.” The idea sounds right, but most system vendors have stayed away, saying they have the same kind of performance issues I saw with ReadyBoost. Now, I’m told that if it’s in place over a number of weeks, there are some improvements, but it’s probably not enough to care about now.
And then finally we come to solid state drives (SSDs), where flash-based drives are supposed to replace hard drives completely. A few weeks ago, I tested one machine – a Dell Latitude D420 -- with an SSD vs. an identical model with a hard drive.
The results were not encouraging. A 32 GB solid state drive added $450 compared to the more standard 60 GB hard drive. It took about 20 seconds less to boot, and 15 seconds less to shut down, which is good, but hardly earth-shattering. Some file operations were faster, some slower, since today’s flash memory typically reads faster but writes slower than a hard drive (though with less startup time, since there’s no drive head to move). Neither performance nor battery life showed a major improvement. Of course, proponents say SSDs should be more durable in a notebook since it doesn’t have any moving parts, but until a lot more people use SSDs in a PC fashion, it’s too early to quantify it.
Since then, I’ve been talking to a number of makers of both PCs and flash memory to ask them why I didn’t see nearly the improvements that flash proponents have claimed. Mostly, the answers say flash is just not ready to replace a hard drive yet. The big issue, they say, is that the operating systems just aren’t tuned for flash drives, but instead treat them just as they would a hard drive, even though their fundamental characteristics are very different. In addition, they say a different interface (the 3 Gbit/sec version of SATA, often called SATA II) will make a big difference as well. And, of course, flash memory, is getting faster over time. When could all this come together? Developers I’ve talked to tell me not to expect it to really make a different until next spring at the earliest. But a year from now, they tell me, flash as PC storage will be ready for prime time.
In the meantime, it remains an interesting idea – particularly if you’re worried about hard drive failures – but not something I’d recommend yet.
No comments:
Post a Comment