Solid state drives?

Anyone seen this yet? It's basically a raid controller with four SSDs on the card all designed to go into a PCIe slot:

Yesterday, we had the honor and privilege to actually test this insane new product from OCZ called Z-Drive. Now I'm not talking about taking a screenshot or two and calling that "article", I'm talking about hours of work and trying to extract the maximum possible performance out of a device that's just... insane. That's the word that's the best possible description really. Can you imagine having transfer rates in the >900MB/s region, more then 700MB/s sustained average reads and writes over 600MB/s by using any kind of storage-technology just l ike that? Yeah, it took me awhile to figure some stuff out but right now... There you go.
 
How many chips it takes depends on the bandwidth available per chip ... and making the chips higher bandwidth isn't really inherently expensive either. Flash presently has ridiculously low bandwidth and bus widths, there is a whole lot of room to increase that without substantially increasing the cost (ie. going from 1% control logic to 5%).
 
Won't last long unless patents throw up too many hurdles.

Competition is fierce and there is nothing fundamental which makes these speeds expensive.

I generally try not to project too far into the future. The fact that prices will come down is obvious barring some unforseen catastrophe.

How fast and by how much however is hard to tell.

For example while MLC chips have come down drastically in price, SLC still seems to be holding onto the a high price range. And even with new control logic, MLC based devices can still run smack into the stuttering/system halting mess that plagued previous devices.

However, that said, the new control logic for some of the newer devices does a much better job of coping and prolonging usage before the onset of the problem. Enough so, that it may be possible for most users to never run into it in daily computing.

It's unfortunate that SLC chips are still prohibitively expensive to create (compared to MLC chips). As their garbage collection happens in a fraction of the time that it takes for an MLC chip.

Regards,
SB
 
I think I have experienced a few of these stalls, now that I have put some mileage on an mlc-based ssd in OSX. The occurrances don't seem to really follow any predictable pattern. It just happens out-of-the-blue every few days or so. Everything just grinds to a halt for maybe 30 sec, and then resumes as if nothing ever happened. It's a little annoying, but not necessarily intolerable (to cause me to abandon the whole ssd as a system drive trial). Arguably, it still beats having the system paging heavily all of the time to a traditional hdd just to do simple things, after a significant amount of uptime. Most notably, the menubars always respond in realtime and sound effects stay reasonably synchronized with UI actions (as opposed to them finally happening as far out as 10's of seconds when the action that triggered it has long come and gone).
 
Yeah, I can "mostly" live with my 1.5 gen SSD in my tablet PC since it only serves light duty. Browsing, e-mail, and the occassional document. But even with that light load, sometimes it'll just grind to a halt. Installing a large windows update for example where the HD is accessed (simultaneous read and write) frequently and constantly.

Newer drives such as the vertex series here should virtually eliminate that for casual uses.

Once the price comes down on them some more, I may pick one up just to see how far you can push one until it runs into the problem.

Regards,
SB
 
Last edited by a moderator:
INTEL has released a firmware upgrade for an SSD performance problem which it initially said did not exist. Even now it claims that the SSD fragmentation issue unlikely to affect average users, but it has issued a firmware upgrade for its X25-M anyway. It was claimed by PC Perspective reviewers that the drive suffers from fragmentation resulting in performance degradation over time. Intel said it could not replicate the results but soon found it was the only organisation on the planet which couldn't.
http://www.intel.com/support/ssdc/index_update.htm
 
Back
Top